To explore the application potential of airborne point cloud and UAV visible light image in tree species identification and classification， a single-tree scale tree species classification and recognition method based on UAV hybrid fusion of multi-modal features and decision was proposed. Firstly， Kendall Rank correlation coefficient method and Permutation Importance （PI） were used for feature selection， and Efficient Low-Rank Multi-Mode Fusion Algorithm （LMF） was used to fuse the selected point cloud and visible image features. Ensemble learning was introduced to input point cloud， image， and fusion features into eXtreme Gradient Boosting （XGBoost）， Light Gradient Boosting Machine （LightGBM）， and Random Forest （RF） base classifiers integrated by Stacking. Finally， the meta classifier， Naive Bayes， is used for decision fusion. The experimental data show that the independent test accuracy of the proposed algorithm is 99.4%， which improves 22.58% compared with the Random Forest classifier by traditional feature concatenate fusion. In addition， the Kappa coefficient also increased by 28.54%. The comparison experiment with Convolutional Neural Network（CNN） shows that the proposed algorithm has obvious advantages in small sample training and better generalization ability.

In recent years， in order to improve the classification accuracy of ground objects， break through the technical system of single sensor， and make up for the limitations of single data source application， multi-source remote sensing data fusion has become a research hotspot concerned by many scholars in the field of remote sensing. The fusion technology of optical image and LiDAR point cloud data of hyperspectral remote sensing technology provides a feasible scheme to improve the accuracy of ground object recognition and classification at the technical level， breaks the technical upper limit of single sensor， and provides a new solution for the integrated acquisition of target three-dimensional space-spectral information. At the same time， it lays a foundation for the research of hyperspectral LiDAR imaging technology. This paper reviews the development history of LiDAR and hyperspectral imaging data fusion， discusses the main fusion methods and research progress at the feature level and decision level， introduces the commonly used feature level fusion and decision level fusion methods in detail， summarizes the latest research algorithms and discusses their challenges and future development and application prospects. Finally， the future development of LiDAR and hyperspectral imaging data fusion is prospected systematically.

The Gaofen-6 medium resolution wide-width camera （GF6-WFV） is designed with two red-edge bands， which has the potential to monitor chlorophyll a concentration in water. In this study， six typical lakes in eastern China， including Guanting Reservoir， Luhun Reservoir and Baiyangdian Lake， were selected as the study area， and measured spectrum and chlorophyll a concentration data were obtained from 141 sampling points. Based on the measured data， the parameters of four kinds of commonly used semi-empirical inversion models of chlorophyll a concentration were optimized and the model accuracy verified， and the optimal inversion model was selected. The results show that the red edge band Ⅰ （B5：710 nm） and red band （B3： 660 nm） are newly added in GF6-WFV data. Which construct a two-band ratio 2BDA model with high inversion accuracy， correlation coefficient square （R²） is 0.89， the Mean Relative Error （MRE） is 34.71 %， and the Root Mean Square Error （RMSE） is 13.29 mg/m³. The results show that the chlorophyll a concentration in water body can be effectively retrieved by using GF6-WFV image data. The inversion model of chlorophyll a concentration in water body established in this paper based on multi-lake and multi-temporal data has good applicability in typical lake repositories in eastern China.

From text analysis to image interpretation， the Topic Model （TM） consistently plays a pivotal role. With its robust semantic mining capabilities， topic model can effectively capture latent spectral and spatial information from Remote Sensing （RS） images. Recent years have seen the widespread adoption of topic models to address challenges in RS image interpretation， including semantic segmentation， target detection， and scene classification. Thus， clarifying and summarizing the present application status of topic models in remote sensing imagery is pivotal for advancing remote sensing image interpretation technology. This paper initially presents the foundational theory of topic models， followed by a systematic overview of their typical applications in remote sensing imagery. In addition， experimental comparisons and analyses are performed across various typical remote sensing image interpretation tasks， illustrating the extensive applicability of topic models in the realm of remote sensing and the efficacy of distinct topic models in enhancing our comprehension of remote sensing imagery. Subsequently， we have outlined the limitations of topic models and explored the potential and prospects of integrating them with deep learning.

Forest is a valuable non-renewable resource， but the ecological environment of forest is seriously threatened by many natural or man-made factors such as fire， flood， and deforestation interference. Accurate grasp of forest resource changes can provide effective information for forest resource management and protection. In the task of forest change detection， traditional machine learning change detection methods have difficulty in capturing deep semantic information due to large differences in forest categories and tree species， and suffer from poor adaptability of extracted features， weak recognition ability， and pseudo-change due to seasonal phases. We propose to build a deep learning model with Siamese neural networks for forest change detection experiments. Three different feature extraction methods， ResNet50 （Residual neural network）， CBAM （Convolutional Block Attention Module） and SE （Squeeze and Excitation） with different lightweight attention mechanisms are used as backbone feature extraction modules， respectively. All three backbone feature extraction networks are trained based on pre-trained weights， which improve change detection by fusing the extracted multi-scale feature maps so that the coarse and fine details of information in different feature maps complement each other. It also has the advantage of sharing weights with the same number of parameters. Taking Jiande Forest Farm in Zhejiang province as the experimental area， two phases of GF-2 images in 2015 and 2020 are acquired to construct a forest change detection dataset with a resolution of 1m. The results of Siamese neural network change detection are compared with the true change labels （Ground truth）， where the backbone feature extraction network SE-ResNet50 has the best combined results with Precision （0.91）， Recall （0.78） and F1-score （0.83）， which is better than mainstream change detection models FC-Siam-conc， FC-Siam-diff. It is proved that Siamese neural networks can accurately capture forest changes in the task of forest lad change detection from high-resolution remote sensing images， and provide a new forest change detection method for forest resource management departments.

Mangrove forests are among the ecosystems with the highest net primary productivity in the world， and they play an important role in the study of global climate change and the evolution of coastal zone geography. Rapid and accurate acquisition of the spatial distribution of mangroves on a large scale is vital for effectively managing and exploiting mangrove resources. Landsat satellite images have become an important data source for extracting large-scale and long-period mangrove distribution information. Yingluo Bay and Pearl Bay along the coast of Guangxi， China are selected as the study sites in this study. Landsat-8 OLI images are used to construct five indices to extract the distribution of mangroves， including Normalized Difference Mangrove Index （NDMI）， Combined Mangrove Recognition Index （CMRI）， Modular Mangrove Recognition Index （MMRI）， Mangrove Index （MI） and Mangrove Vegetation Index （MVI）. This study compared the efficiency of different indices used for mangrove extraction to determine the optimal mangrove extraction index. Optimizing the mangrove distribution information extraction is proposed by combining Normalized Difference Water Index （NDWI） index. The aim is administrator improve the remote sensing classification accuracy of mangroves. It is also applied to the extraction of coastal mangroves in Guangxi. The results showed that： Mangrove distribution can be effectively extracted based on Landsat-8 OLI satellite images and index method. By comparing the extraction accuracy of five indices of mangroves， we found that the MVI has the best extraction effect and the CMRI has the worst extraction effect. The combination of NDWI can better optimize the extraction accuracy of mangroves， and the optimized MVI applied to Guangxi coastal mangroves showed the best extraction results with an overall accuracy of 97.10%. The research strategy and the range of mangrove index thresholds in this paper can provide reference and decision support for large-scale mangrove distribution extraction.

A brief review has been conducted on the progress of typical spaceborne and airborne polarimetric Synthetic Aperture Radar（SAR） systems at home and abroad， for which the implementation of radiometric and polarimetric calibration accuracies has been focused and surveyed. First the general requirements for the polarimetric SAR data calibration accuracy have been drawing from literature research， and then the status quo of representative polarimetric SAR systems in the word and the system data calibration accuracy achievements have been systematically presented， including the relative radiometric calibration accuracy， the absolute radiometric calibration accuracy， the polarization channel crosstalk accuracy， the polarization channel amplitude imbalance accuracy， and the polarization channel phase imbalance accuracy， etc. Finally， the key factors affecting the calibration accuracy of polarimetric SAR data have been analyzed， and the future calibration tasks meeting the new polarimetric SAR system design has been briefly discussed. This paper comprehensively describes the calibration accuracy information index of polarimetric SAR systems and their development status， and provides relevant researchers with timely， comprehensive and systematic information on the development requirements of polarimetric SAR systems and the research progress of calibration accuracy achievements.

Taking the SCI and SSCI papers in the Web of Science （WoS） and the journal papers in the CNKI as the data source， we used bibliometrics and the literature analysis software CiteSpace to explore the development and hot spots of urban ecological remote sensing in the past three decades and drew the knowledge map of the field from 1991 to 2021. The results show that： （1） in terms of the number of publications， the research on urban ecological remote sensing has gone through three development stages， i.e.， budding stage， accumulation stage， and rapid growth stage； globally， the center of centroid of the number of published papers shows a trend of migration from the east to the west and then back to the east； （2） in the light of research institutions， Institute of Geographic Sciences and Natural Resources Research， Chinese Academy of Sciences and Beijing Normal University have published most papers in this field in WoS and CNKI； （3） in terms of journals， Remote Sensing and Acta Ecologica Sinica have published most papers in this field in WoS and CNKI， respectively； （4） in terms of core authors， Weiqi Zhou in Chinese Academy of Sciences and Hanqiu Xu in Fuzhou University （China） published the most papers in WoS and CNKI， respectively； and （5） statistics of keywords shows that the keywords of ecological environment， land use， and landscape pattern were the hot spots of urban ecological remote sensing research in recent years. In general， urban ecological remote sensing was playing an increasingly important role in the disciplines of environmental science， physical geography， and geomatics. The application of remote sensing ecological models to quantitatively assess ecological environment quality has become an important trend in this field. Also， urban ecological remote sensing research requires higher and higher spatial resolution of remote sensing images. Collaborative inversion of urban ecological quality with various remote sensing images will also be a development trend in this field.

In the ' post-poverty era ' of 2020， the four prefectures in southern Xinjiang are still underdeveloped areas in the development pattern of the whole country and in Xinjiang.Therefore， it is of great significance to carry out long-term economic measurement and development analysis of the four prefectures in southern Xinjiang. However， the traditional measurement methods using socio-economic data have great limitations.The nightlight remote sensing data is used to objectively invert the economic development characteristics of the poverty-stricken area.This paper selects four prefectures of southern Xinjiang as the study area， and corrects NPP/VIIRS and DMSP/OLS data for noise and supersaturation respectively， based on the integration of two kinds of night light data， the correlation between the total night light amount of 33 counties （cities） and the secondary and tertiary industries was used， the spatial and temporal pattern of economic development in the four regions of southern Xinjiang from 2005 to 2020 was studied by using standard deviation ellipse and Molain index.The results show that ： （1） From 2005 to 2020， the economic center of gravity moved to the northwest. The total economy is dominated by the northeast-southwest direction， the economic development trend is more and more concentrated and contiguous.（2） There has been a high spatial autocorrelation and aggregation during the study period， mainly showing H-H and L-L gathering areas. The higher the economic level of the region， the more prone to aggregation.（3） The probability of occurrence of cold spots and hot spots in regional economic development in the four prefectures of southern Xinjiang has obvious local characteristics. The coordinated economic development among the four regions of southern Xinjiang should be the policy focus.

This study， focused on the area of Beibu Gulf， explores the remote sensing inversion method for chlorophyll concentration based on the Sentinel-3A satellite's OCLI water color sensor. The study partitions the Beibu Gulf by using measured spectral data and then combines the measured chlorophyll-a concentration with Sentinel-3A remote sensing data of which aims to build the remote sensing inversion model for chlorophyll-a concentration. The results show that （1） the remote sensing reflectance curves exhibit distinct partition characteristics， dividing the area into nearshore， transitional， and offshore water types based on the spectral features； （2） Different water types require different inversion factors for model construction， and all of them got relatively good fitted result. Among them， the fitted inversion factor is R_rs（764.375）/R_rs（681.25） that could be used in the nearshore water， for the transitional water， ［1/R_rs（620）-1/R_rs（708.75）］/R_rs（753.75） is the most suitable， and for the offshore water， R_rs（708.75）-R_rs（764.375） achieves the best fitting performance， with corresponding R² values of 0.67， 0.80， and 0.8， respectively； （3） The partitioning method effectively improves the applicability and accuracy of the remote sensing inversion model for chlorophyll concentration in the Beibu Gulf. This study successfully realizes the remote sensing inversion of chlorophyll concentration in the Beibu Gulf by using a partitioning model based on Sentinel-3A satellite's OCLI data. The result provides the important scientific support for the remote sensing monitoring of chlorophyll concentration in the Beibu Gulf， and enhances the management and protection of marine ecological environments.

Object recognition technology based on high-resolution remote sensing images is widely used in the fields of land and resource monitoring and intelligence collection. Accurate and fast object detection methods are the hot spots and difficulties in the current research on remote sensing images. However， the current detection methods overly pursue improving detection accuracy while ignoring detection speed. Therefore， an improved lightweight network is proposed based on YOLOX to balance detection speed and accuracy. Firstly， for the backbone of feature extraction， a Mobilenetv3tiny is proposed to improve the detection speed by reducing the parameters of the network. Secondly， the Ghost is introduced into the feature pyramid networks to reduce the complexity of the network under the premise of ensuring detection accuracy. Finally， Alpha-IoU and VariFocal_Loss are used to optimize the loss function to improve the convergence speed and positioning accuracy of the network. The ablation experiment was carried out on the NWPU VHR-10 dataset. The results show that， compared with the baseline， the improved network has a detection accuracy increase of 0.76%， a speed increase of 19.72%， a weight of 11 M （Mega）， and a parameter reduction of 65.66%. The overall effect of the improved network is better. In addition， comparative experiments on the DIOR dataset show that the detection speed is improved by 26.88% while ensuring high detection accuracy. And that proves the effectiveness of the improved network. Therefore， the improved network can effectively balance detection speed and accuracy and is easy to deploy， which makes it suitable for real-time detection of remote sensing image targets.

In recent years， the abundance of hydro fluorocarbons （HFC） has been increasing， which has huge greenhouse potential value. It has an impact on global warming and also indirectly causes the destruction of the ozone layer. Scholars at home and abroad have carried out a wide range of in-situ ground measurements to obtain global abundance. At the same time， remote sensing technology can monitor the changes of HFC gas in a large range， for a long time and quickly， and has become an important means for the inversion of the gas concentration. The contents of in-situ measurement method， tracer ratio method， satellite inversion sensor development and satellite inversion method are described， and the advantages and disadvantages of different inversion methods are compared in combination with load characteristics analysis. Finally， discusses and prospects the existing problems and future development trend of current inversion.

Using Sentinel-2 data and multiple methods to invert the permanganate index （COD_Mn） of deep-water lakes and reservoirs in the Karst Plateau is of great significance for the regional water environment management and enrichment of water quality inversion theories. Taking Hongfeng Lake and Baihua Lake as the research area， based on the Sentinel-2 MSI image and COD_Mn concentration data， use Random Forest Regression （RFR）， Support Vector Regression Method （SVR）， Gaussian Process Regression （GPR）， Obtaining COD_Mn spatial distribution in different periods of 2018~2020. The results show that： ① The RFR model has the highest accuracy， the verification set is 0.222 mg·L^-1， MAPE is 5.84%， and R² is 0.841； In addition to the upstream of Baihua Lake， the COD_Mn concentration of the overall lake is low and there is not much change over time. Studies have shown that the RFR model and Sentinel-2 data are well applicable to monitoring the COD_Mn concentration monitoring of deep-water lakes in the Karst Plateau.

Phyllostachys edulis is one of the most important and intensively managed forest resources in southern China， Chlorophyll Content Index（CCI） is a crucial indicator of plant health and growth. It is of great significance to realize remote sensing inversion of chlorophyll content in Moso Bamboo forest to monitor the health degree of it. Firstly， three ways of transform including HSV （Hue-Saturn-value） transform、GS （Gram-Schmidt Pan Sharpening spectral Sharpening method） transform and PCA （Principal Component Analysis） were used to make sure that Landsat 8 multispectral image and Unmanned Aerial Vehicle （UAV） high resolution single-band image data were fused well together. Secondly， 8 kinds of vegetation cover indices were then constructed based on multi-source remote sensing data， moreover， three machine learning models including K-nearest Neighbor （KNN） regression， Random Forest （RF） regression as well as CatBoost regression were applied to ensure vegetation index and chlorophyll content could be fitted. Finally， the inversion model of chlorophyll unit content in Moso Bamboo forest was then established. The results indicated that ：（1） In terms of fusion effect， it turned out that GS was the optimal model cause various evaluation parameters derived from it such as mean value、standard deviation、mean gradient joint entropy and spatial frequency were all the highest， which were 73.407 8、80.672 9、29.699 2、9.765 5 and 74.876 9， respectively. （2） In the validation set based on fused multispectral data， Landsat 8 multispectral data and UAV data， RF algorithm turned to be the best algorithm（RF algorithm's corresponding R² is 0.687 6、0.576 1、0.425 4， respectively， while the corresponding RMSE were 2.918 4 μg/cm²、3.559 5 μg/cm²、3.974 5 μg/cm²_， respectively）. （3） The inversion effect of chlorophyll content could be better when based on fusion data than Landsat 8 data and UAV data. This study coupled with multi-source remote sensing data to realize remote sensing retrieval of chlorophyll content in Phyllostachys pubesculus forest， which can provide scientific reference for dynamic monitoring of phyllostachys pubesculus forest health.

In view of the high cost of traditional river water quality monitoring and the sparse ground monitoring stations， based on Sentinel-2 satellite multispectral remote sensing data， combined with MODIS surface temperature， vegetation index， aerosol optical thickness data products， and the surface wind speed data in ERA5 meteorological data products， the monitoring data of the surface water quality monitoring stations with non-optical active parameters Dissolved Oxygen （DO）， Chemical Oxygen Demand （COD） and ammonia nitrogen （NH₃-N） are taken as reference， three machine learning methods， Support Vector Regression （SVR）， Random Forest （RF） and Multilayer Perceptron （MLP）， were used to select the optimal inversion model of each water quality parameter and its corresponding input feature combination through comparative experiments. The experimental results of the model performance test show that the determination coefficients （R²） of DO， COD and NH₃-N are 0.896，0.781 and 0.529， respectively，and the Root Mean Square Error（RMSE） are 0.263 mg/L，0.383 mg/L and 0.061 mg/L， respectively. Compared with the retrieval results using only Sentinel-2 multi-spectral remote sensing data， R² increased by 7.04%， 19.05% and 18.34% respectively， and RMSE decreased by 34.58%， 37.42% and 14.08% respectively. It shows that multi-source remote sensing and meteorological data are of great significance to improve the retrieval accuracy of DO， COD and NH₃-N water quality parameters. Finally， the model robustness evaluation experiment shows that the trained model has better space-time robustness when the representativeness of the model training data is close to the global data.

Unmanned Airborne Vehicle （UAV） based multispectral remote sensing has the advantages of low cost and flexible time in monitoring small water bodies. However， the common multispectral cameras have the problems of few pixels and lack of characteristic bands of inland water bodies， which limit the advantages of UAV based multispectral remote sensing in monitoring the water environment. In order to solve these problems， this study customized the bands for inland water quality monitoring for the Aerospace ShuWei KP-8 multispectral camera with high pixel， including 670 and 700 nm bands for inland water chlorophyll a retrieval； Then， a flight experiment was carried out to obtain the multispectral image of the turbid and eutrophic Luhun Reservoir. And the synchronously obtained water quality parameters from the water surface experiment were used to build the retrieval models of the typical water quality parameters， including the Secchi-disk depth， turbidity， suspended solids and chlorophyll a concentration； The retrieval models were applied to the multispectral image， and the typical water quality parameters in Luhun Reservoir were retrieved and their spatial distribution rules were analyzed. The results show that the UAV based high pixel multispectral camera has important potential in the operational monitoring of inland water environment.

The use of UAV hyperspectral remote sensing data technology to quickly and accurately extract typical grassland vegetation types is of great significance for dynamic monitoring of grassland ecological security.In the typical grassland area of Baiyinxile pasture with severe degradation， hyperspectral images with a spatial resolution of 1.8 cm and a spectral resolution of 4 nm， with a total of 125 bands （450 nm to 950 nm） were collected. The main degradation indicator species， Artemisia cholerae， was selected as the identification target， and after differential transformation， envelope removaland other spectral transformations， the differences in spectral characteristics were analyzed. There are obvious spectral differences at 500 nm、550 nm、670 nm， so the above three bands were selected as characteristic bands， and the degradation indicator species identification model of Support Vector Machine （SVM） and Random Forest （RF） was constructed， and the accuracy was verified. The results show that the recognition accuracy of SVM and RF are 96.92%和97.34%， respectively， and the Kappa coefficients are 0.95 and 0.96， respectively. It can be seen from the results that the identification accuracy of the random forest model is higher， and the pixel spatial distribution of degraded indicator species is closer to the natural state， which can provide technical support for monitoring typical grassland degradation indicator species.

With the rapid development of earth observation technology， high-resolution remote sensing image change detection has become a research hotspot in the remote sensing domain. The increase in spatial resolution brings rich spatial information， but also leads to the problem of "pseudo-change" caused by the change of the spectrum and other performance characteristics， which does not change due to phenological differences. Morphological Attribute Profiles （MAPs）， as an efficient spatial information modeling method， can accurately describe complex change characteristics from different attributes and multiple scales， and have been widely used in the field of change detection tasks. Nevertheless， the existing MAPs methods usually do not consider the properties and scale balance of the differential profile， so they are prone to fall into local optimum； at the same time， the effective fusion of differential features into change detection results is another difficult problem faced by such methods. To this end， this paper proposes a change detection method that combines adaptive MAPs with decision fusion. Firstly， the initial differential feature set is extracted by CVA on the MAPs； On this basis， a Balanced Optimal Objective Function （BOF） is designed to extract the optimal differential feature set； Finally， based on the proposed change intensity evidence index （EVI） and evidence confidence index （IOEC）， a multi-feature decision fusion framework is constructed to obtain change detection results. The experimental results show that the Overall Accuracy （OA） and F1 score （F1） of the proposed method can reach 96.41% and 88.67%， respectively. which are significantly better than the comparison methods in both visual analysis and quantitative evaluation. especially for the "pseudo-variation" proposed in this paper. Compared with the comparison method， the method in this paper can realize more accurate discrimination and effectively alleviate the "pseudo change".

The concentration of chlorophyll-a is a key indicator for evaluating the eutrophication of coastal water， and the research on the influencing factors of coastal chlorophyll-a is significant to marine environmental protection. However， the existing studies mostly focus on the effect of natural factors on the concentration of chlorophyll-a in coastal waters， ignoring the human activities. Therefore， this paper uses nighttime light brightness to characterize the intensity of human activities， and divides cities on the southeastern coast into three types based on its relationship with coastal chlorophyll-a concentration， and also combines sea surface temperature， wind speed， solar radiation， precipitation and human factors， using the Generalized Additive Model （GAM） to analyze the impact of multiple factors on coastal chlorophyll-a concentration in three types of cities in different seasons. The results show that the change of chlorophyll-a concentration was dominated by natural factors in type I cities such as Fuzhou and Shantou， the dominant factor in spring is wind speed， and the sea surface temperature in summer， autumn and winter； while human activities have little effect. Type II cities such as Zhuhai and Dongguan are dominated by natural factors. The dominant factor is wind speed in spring， autumn and winter， and the sea surface temperature in summer； human activities have a greater promoting effect in summer and autumn. Type III cities such as Shenzhen and Hongkong are dominated by human factors. Human activities have the greatest impact on chlorophyll-a concentration in spring， summer and autumn and they are a negatively correlated. In winter， sea surface temperature has the greatest impact.

Compared with the traditional manual field survey method， the use of UAV tilt photogrammetry technology for multiangle photography to extract individual tree crown information has the advantages of high efficiency， accuracy and low cost. In this study， an individual tree crown extraction method combined with a visible vegetation index and watershed algorithm was proposed by taking a larch near-mature forest in the Wangyedian forest farm in southwestern Karaqin Banner， Chifeng city， Inner Mongolia， as the research object and using UAV images obtained by tilt photogrammetry as the data source. First， the Excess Green minus Excess Red （ExGR） in the visible light band was calculated by a digital orthophoto model. The median filter was used to denoise the tree crown area map， and a reasonable threshold was selected to binarize the image to separate the vegetation and non-vegetation areas. Vegetation areas were used to mask the canopy height model. Finally， the accuracy of individual tree crowns was verified by the watershed segmentation algorithm. In the process of extracting the crown area， vegetation and non-vegetation areas are successfully separated based on the ExGR index and threshold method. Through median filtering， speckle and noise caused by uneven brightness， shadow and texture in the non-vegetation area are effectively removed， the integrity of the crown edge and the connectivity of the crown are ensured， and the over segmentation phenomenon of the watershed algorithm is reduced. At the individual tree scale， the accuracy rate of crown parameter information extraction was 88.72% and 79.38%， the recall rate was 93.29% and 88.60%， and the F-score was 90.59% and 83.74%. On the sample plot scale， the relative errors are 15.45% and 22.92% respectively. The results show that the visible vegetation index based on Digital Orthophoto Image can effectively eliminate the influence of bare land and other background factors in the forest， and the watershed segmentation algorithm based on the canopy height model can accurately distinguish individual tree information. The combination of the two data sources based on the UAV tilt photogrammetry technology gives full play to their respective advantages. The method of extracting the single tree crown information based on the UAV tilt photogrammetry technology is feasible and can extract the single tree crown information of the forest with high canopy density efficiently and accurately.

China's coastal areas are not only the most strongly interacting zones extending from land to sea， but also the natural spatial units on the surface of the earth affected by runoff， tides and the effects of human activities and climate change. In this paper， using with the 2003~2018 high quality temporal resolution Land Surface Temperature （LST） and Near Surface Air Temperature （NSAT） products， the spatial and temporal distribution pattern and synergy of urban heat islands and urban heat waves in coastal cities in China are systematically compared. The results show that： （1） Extreme high temperature events in China's coastal regions show a trend of increasing intensity and duration. Specifically， the intensity of urban heat islands in summer is as high as 2.25 ℃， the average heatwave frequency based on ground temperature in the entire coastal area is 24.59 times， and the temperature-based heat wave frequency is 16.33 times， accounting for 90.81% and 96.68% of the annual heat wave frequency， respectively； （2） The frequency of urban heat waves and urban heat island intensity is significantly positively correlated in the North Temperate Zone and North Subtropical Zone along the coast， and is most obvious in the North Temperate Zone. An increase of 1°C in average LST can lead to an average increase in heat wave events twice. Among them， the heat wave frequency in the North Temperate Zone increased the fastest in the three major regions， and the compound growth rate of heat wave frequency based on LST and NSAT exceeded 5%； （3） From 2003 to 2018， the urban population of China's coastal regions increased by 59%， and the number of people affected by heat waves increased by nearly 370%， exceeding 5% of the total urban population （about 40 million people）. Although the urban thermal environment in China's coastal regions and the El Niño and La Niña phenomena in the sea show a more consistent distribution in time and space， how the overall urban change and population growth promote the change of high temperature and thermal environment patterns in different regions still need to be further discussed and analysed using longer time series and high spatial and spatial resolution data.

Monitoring the dynamic changes of water area in Dongting Lake is of great significance for flood control， ecosystem stability and biodiversity. The deep learning algorithm represented by the classic Unet the innovative HRNet has become an efficient way to obtain remote sensing image information. Taking Sentinel-1A SAR image as the main data source， this paper qualitatively and quantitatively analyzes the water extraction results of SDWI index method （Sentinel-1 Dual-Polarized Water Index，SDWI）， object-oriented classification method， UNet network model and HRNet network model. Based on the best water extraction method， the temporal and spatial variation characteristics of water area in the flood season （April to October） of Dongting Lake from 2016 to 2021 are analyzed. The results show that ： ① The deep learning algorithm represented by HRNet and Unet has better water extraction effect than traditional methods. Among them， HRNet has superior performance in noise suppression and shadow resistance， and the F1 score， MRate and MIoU are 0.961 6， 0.007 8 and 0.958 6， respectively. ② During the flood season， the water area of Dongting Lake shows the characteristics of “ increase-full-decrease ” in the monthly variation. The lake surface begins to expand from April to May， and the water area maintained at a high level from June to August. Since then， due to the decrease of inflow， the water area gradually decreases from September to October. The 2 263.90 km² at July 30， 2020 is the largest water area monitored during the study period. ③ The submerged frequency of water body in flood season of Dongting Lake gradually decreases from the center of the lake body and the main stream. The distribution patterns of submerged frequency in different lake areas are different. The submerged frequency of East Dongting Lake is higher than that of South Dongting Lake and West Dongting Lake. In summary， the combination of Sentinel-1A SAR image and deep learning technology can realize the efficient acquisition of water information and high-frequency monitoring of the water surface area at Dongting Lake， providing a new idea for the high dynamic lake water monitoring.

The estimation of rice leaf nitrogen content is important to achieve the goals of high rice yield and efficient fertilization in the field. In this paper， we propose a Particle Swarm Optimization-Deep Forest （PSO-DF） model-based method for estimating the nitrogen content of rice leaves， which determines the number of estimation layers in the optimal cascade and the optimal estimator in the Deep Forest （DF） model parameters by a particle swarm optimization algorithm. The number of trees in the optimal estimator is determined by the particle swarm optimization algorithm to improve the regression accuracy of the DF model on Rice datasets.To verify the effectiveness of PSO-DF， this paper used an unmanned aircraft with a hyperspectral image collector to obtain hyperspectral images of Ningxia japonica rice， and sampled， measured， and analyzed the rice leaves at the same period， and extracted the three feature bands with the highest correlation coefficients with rice leaf nitrogen content， which were used as spectral features for inversion with rice nitrogen content data， and compared PSO-DF， the original model DF， and six other common The rice nitrogen content estimation models constructed by machine learning algorithms were compared. The results show that the model constructed by the PSO-DF algorithm outperforms the other models， and its R² and RMSE indexes are significantly better than those of the other models.

Leaf Area Index（LAI） is an important indicator to reflect the growth state of crops， which is usually estimated by vegetation index. Traditional inversion models are mostly based on multivariate regression models， while the potential of multivariate regression models based on bivariates in LAI inversion has not been fully explored. By extracting the spectral features and texture features of satellite images， the correlation between each remote sensing feature and winter wheat LAI was analyzed based on Pearson correlation coefficient. Using Simple Regression model （SR）， Multiple Linear Regression model （MLR） and Random Forest Regression model （RFR）， the relationship between remote sensing characteristics and LAI of winter wheat was studied. The inversion accuracy of each inversion model was determined by the accuracy index （determination coefficient R²， root mean square error RMSE， relative root mean square error rRMSE）. Based on the above evaluation indicators， the optimal inversion model was proposed. The results showed： （1） All vegetation indexes and some texture indexes have achieved good inversion results in LAI inversion （R²>0.6）. Among them， the Universal Normalized Vegetation Index performed the best among all vegetation indices （R²=0.754，RMSE=0.606，rRMSE=12.99%）. Except for the mean feature inversion accuracy of some bands that is comparable to vegetation index， the accuracy of most texture feature inversion for the winter wheat LAI is poor； （2） The bivariate multivariate linear regression model with the highest LAI inversion accuracy for winter wheat was obtained through bivariate combination （R²=0.780，RMSE=0.573，rRMSE=12.29%）； （3）In the case of multiple input variables （at least 3 feature variables）， RFR performed better than MLR. Compared to texture features， the inversion performance of texture indices was better. The research results can provide a new approach and method for monitoring large-scale crop LAI based on satellite imagery in the future.

Quickly and accurately obtaining information on the area of special plant planting plots is of great significance for drug production estimation and prevention of drug criminal activities. Aiming at the problem that existing special plant planting plot detection algorithms in high-resolution remote sensing images cannot quickly obtain location information and area information at the same time， this paper proposes an improved PSPNet semantic segmentation model suitable for quickly and accurately extracting certain special plant planting plots. . By introducing the channel attention SE module， the problem of holes in the segmentation of a certain special plant planting plot is solved. The Dice Loss loss function is added to improve the problem of imbalance of positive and negative samples. The encoder-decoder structure is introduced to make the extracted special plant planting Lot outline boundaries are more precise. By using the MobileNetv2 backbone network， the model prediction speed is increased by 90%. The improved I-PSPNet model achieved 95% and 84% MPA and 84% MIoU in the extraction of a special plant planting plot， and the detection efficiency reached 84 fps. Comparative experiments between I-PSPNet and UNet， Deeplabv3+， and PSPNet show that the prediction accuracy and speed of the improved model are better than the above three models. Among them， MPA increased by 24%， 7.4%， and 7.7%， and MIoU increased by 24%， 7.4%， and 7.7%. 19%， 4.3% and 4.9%， predicted speed improvements of 57 fps， 56 fps and 40 fps. At the same time， the improved model also has good applicability to RGB band data sets and GF-2 images. The improved model proposed in this article can be used to quickly and accurately obtain the location information and area information of a special plant planting plot， and help the anti-drug department quickly discover the illegal planting of a special plant planting plot， objectively assess the scale of illegal planting， and implement precise crackdowns on illegal drug and criminal activities. Provide technical support.

MODIS time series surface reflectance data is widely used in the dynamic monitoring of land surface， but the influence of factors such as cloud cover causes spatial and temporal gaps in the data， which affects the data availability. In this paper， we propose a time-domain reconstruction method based on L1 regularization， which can effectively repair the gaps in MODIS surface reflectance data and realize the reconstruction of long time-series data with high accuracy. The proposed method firstly identifies the noise generated by natural and systematic factors in the time-series data， and then pre-fills the missing information region inter-annually based on noise detection. On this basis， we introduce a L1 regularization model that is more robust to abrupt noise， and construct a variational model combining the noise masks to restore the time series trend of land surface. The experimental results show that compared with SG filtering， HP filtering， L1 filtering and HANTS， the method in this paper achieves the highest reconstruction accuracy at different percentages of missing pixels of 10%， 25%， 50% and 75%， and also achieves better reconstruction results under different ground surface scenes. Therefore， this method has more advantages in both time series curves reconstruction and spatial details restoration， which shows a high practical value.

Grassland is crucial to the ecosystem stability and agricultural development. In semi-arid or arid areas， the problem of insufficient water for grass growth is mainly solved by irrigation， but grasslands with irrigation conditions are limited. At present， there are few researches on the monitoring of irrigation and rain-fed grassland resources and their spatial-temporal distribution using remote sensing techniques， and the relevant datasets are extremely lacking， which brings inconvenience to the assessment of water resources in areas where irrigation and rain-fed grasslands coexist and land surface process simulation. Therefore， this study used the Google Earth Engine Calculation Cloud Platform （Google Earth Engine， GEE）， called the Sentinel-2 satellite remote sensing image of 2020， selected random forest classification method （Random Forest， RF）， added vegetation moisture index， extracted irrigated and rain fed grassland in the middle and lower reaches of the Bayin River in the northeast of the Qaidam Basin， and formed a dataset. After verification， the overall accuracy of this data set was 99%， and the Kappa coefficient was 0.84. The area of irrigated grass was 74.23 km²， accounting for 4.1% of the total grassland area in the middle and upper reaches of the Bayin River. This data set accurately reflected the interlaced distribution characteristics of irrigated grassland， cultivated land， wetland and other types of land. This study can support the water resources planning and evaluation， land surface process simulation， etc. in the Bayin River Basin.

Rice-fish co-culture， as a model of modern ecological cycle agricultural， with significant social， economic， and ecological benefits on ensuring stable food production， reducing pollution， improving soil fertility， and lowering CH₄ emissions. Therefore， obtaining information on distribution and area of rice-fish fields by using remote sensing technology， is helpful in enhancing the level of agricultural digital management and improving the efficiency of resource utilization efficiency. In this study， we selected the typical rice-crayfish model in the Chengdu Plain for remote sensing identification. First， the time-series data of Sentinel-1 VH polarization backscatter coefficients from 2019~2021 were collected and preprocessed in the Google Earth Engine， to reduce the noise of SAR time-series data. Then the time-series characteristics of typical ground objects were analyzed， including rice-crayfish fields， paddy fields， lotus root fields， orchards， traditional aquaculture， etc， and the characteristic parameters statistical of the backscatter coefficients time-series were statistically analyzed. Finally， the information of rice- crayfish fields， rice fields and lotus root fields were extracted by the classification method of random forest. The results showed that the backscattering coefficients of rice-crayfish fields exhibited typical time-series variation characteristics. Specifically， the annual variation trend of backscattering coefficients began with a smooth transition at low value， then increased rapidly， and finally decreased sharply to low value， due to the state of rice-crayfish fields changed from water body to vegetation and then back to water body. Moreover， the range of coefficient variation and the time of curve peak were significantly different from paddy fields and lotus root fields， respectively. The overall accuracy and Kappa coefficient based on random forest classification were 94.32% and 0.91， respectively. This suggested that time-series data of Sentinel-1 can effectively identify rice-crayfish fields in cloudy regions. The results can provide a reference for remote sensing identification of rice-crayfish fields in cloudy areas.

The high spatial and temporal resolution Sentinel-2 images are increasingly becoming the primary remote sensing data source for surface water extraction.A comparative study of the extraction effects of various water index methods based on this satellite image is a significant reference value for improving surface water’s remote sensing monitoring capability. In this study， the seven water indexes （NDWI， MNDWI， AWEInsh， AWEIsh， WI2015， CDWI and MNDWI_VIs） are used to extract surface water from four sample areas with different combinations of surface water types in North China， Northeast China， the middle and lower reaches of the Yangtze River and Northwest China.The water indexes’ accuracy is quantified using Sentinel-2 MSI images on the GEE （Google Earth Engine） platform. The results show that， all seven water indexes generally can identify surface water well， but there are some differences in performance when extracting different types of surface water bodies； the NDWI index underestimate the distribution of surface water in transient water bodies （e.g.， paddy fields， floodplains， etc.） and have a high miss-score speed； while the AWEInsh， AWEIsh and WI2015 indexes have an overall tendency to overestimate and have a high miss-score rate； the MNDWI_VIs water index maintains the highest extraction accuracy in areas with complex water index； in the field of monitoring water changes in long time series， the performance of the seven water bodies is generally consistent with the conclusions obtained based on single-view imagery. This study provides an essential scientific basis for carrying out surface water monitoring in different water bodies.

The extreme drought disaster in Jiangxi Province in 2022 severely affected the growth and yield of citrus. It is of great significance to use remote sensing technology to assess the degree of drought damage quickly and accurately for reducing losses of citrus planting and subsequent stabilize yield. The citrus planting areas of Jiangxi Province were identified and extracted by using the 2022 Landsat remote sensing image data， the land use type data of Jiangxi Province， and the Google Earth citrus supervised classification sample point data. On this basis， the Normalized Difference Vegetation Index （NDVI） and Land Surface Temperature（LST） of the key growth period （June-October） of citrus in Jiangxi Province were calculated by using the MODIS data products of 2021 and 2022， which used for joint inversion of Temperature-Vegetation Drought Index（TVDI）. Combined with the citrus planting area data of Jiangxi Province Statistical Yearbook and field survey data， the economic losses caused by extreme drought in 2022 in Jiangxi Province were quantitatively assessed under four cases. The results showed as follows： （1） The average TVDI of citrus planting areas in Jiangxi Province from June to October in 2021 and 2022 were 0.83 and 0.62， respectively， and drought stress increased significantly in 2022； （2） Severe drought accounted for 66.1% and moderate drought accounted for 33.7% in the citrus planting area of Jiangxi Province from June to October 2022， and the spatial distribution of drought was more severe in northern Jiangxi Province than in southern Jiangxi Province. （3） From early July to early November 2022， TVDI of citrus planting areas in Jiangxi Province remained above 0.8 for a long time， which was characterized by severe drought. This period coincided with the key growth period of citrus and had a great impact on citrus growth. （4） In 2022， the average reduction rate of citrus yield in Jiangxi Province reached 58.2%， and the economic loss of citrus planting showed an increasing trend from north to south. The direct economic loss of southern， central， and northern Jiangxi Province were 4.964 billion yuan， 4.517 billion yuan， and 1.984 billion yuan， respectively. The research results are helpful for the government to quickly find out the disaster situation of citrus farmers in Jiangxi Province， and provide a certain basis for the decision of citrus planting drought relief， loss reduction， and yield guarantee of citrus planting in the future.

The correlation between the mathematically transformed spectral data including the original spectrum of soil and the heavy metal content was analyzed， and then the VISSA-IRIV algorithm was used for spectral feature extraction， and Partial Least Squares Regression （PLSR）， BP Neural Network（BPNN）， particle swarm optimization BP neural network， genetic algorithm optimization BP neural network models were developed to compare and obtain the optimal inversion models of Cr and Cu contents of soil heavy metals. The results showed that the VISSA-IRIV algorithm achieved efficient dimensionality reduction of the spectral data； the prediction effect of the BPNN model was significantly better than that of the PLSR model； the inversion accuracy and stability of the optimized BP neural network models were greatly improved， and the best inversion model combinations for Cr and Cu elements were FD-GA-BPNN（R²=0.87，RMSE=13.82，RPD=2.95），and SNV-FD-PSO-BPNN（R²=0.92，RMSE=4.25，RPD=3.41）， respectively. This study provides an effective method for the accurate and rapid analysis of soil heavy metal content， which is of great practical significance for the realization of soil heavy metal pollution control.

AW3D30 DEM data is one of the most widely used basic geographic information data， and its accuracy directly affects the reliability and rigor of a series of derivative products. Therefore， the accuracy validation and improvement of AW3D30 DEM data has always been a research hotspot.. However， conventional high-precision verification data are difficult to obtain and expensive to apply in a wide range of research areas. With global coverage and sub-meter elevation accuracy， ICESat-2 data can provide reliable reference data source for AW3D30 DEM data accuracy validation and improvement. Therefore， this paper takes Henan Province as the study area， and uses ICESat-2 data to validate the elevation accuracy of AW3D30 DEM from the perspective of slope， aspect， geomorphic type and land use type and proposes the Random Forest-Long Short Term Memory Network（RF-LSTM） hybrid model to improve AW3D30 DEM. The results show that the elevation accuracy of AW3D30 DEM decreases with the increase of slope， elevation and topographic relief. The slope direction has less influence on AW3D30 DEM’s elevation accuracy， and the error distribution has no obvious regularity. The accuracy is higher in bare land and cultivated land， and worse in woodland land. The RF-LSTM hybrid model can significantly reduce the mean absolute error and root mean square error of AW3D30 DEM， improve the accuracy of AW3D30 DEM， and provide a reference for the establishment of other DEM data improvement models.

Scientific research on the spatio-temporal variation characteristics of surface vegetation GPP and its response to climate change is of great significance for in-depth understanding of terrestrial carbon cycle， evaluation of environmental quality of terrestrial ecosystems， and estimation of the ecological effects of future climate change. At present， most of the time series of GPP spatio-temporal changes on the Qinghai-Tibet Plateau are short and the spatio-temporal changes and characteristics of GPP are not deeply analyzed. Therefore， this study used linear trend detection methods to study the changes of GPP in the Tibetan Plateau between 2001 and 2019， and analyzed the response of this feature to topography and land types， while exploring the relationship between GPP and air temperature， precipitation and snow cover. The results show that the inter-annual trend of GPP on the Tibetan Plateau shows the distribution characteristics of low in northwest and high in southeast， with a maximum value of 0.23\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}} and a minimum value of -0.125\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}}；From October to March， the inter-annual change trend of GPP in most regions is small. The inter-annual change trend and spatial heterogeneity of GPP in July is the largest， from -0.384\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}} to 0.303\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}}，and gradually decrease from August to October. The interannual trend of GPP decreases with elevation increasing.The interannual trend of GPP south of 30° N decreases rapidly and north of 30°N increases slowly with latitude increasing. The increasing inter-annual change trend of cropland/natural vegetation mosaics type is largest， 0.034 8\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}}.The decreasing inter-annual change trend of closed shrublands is largest， 0.007 6\mathrm{g}\mathrm{C}\bullet {\mathrm{m}}^{-\mathrm{2}}\bullet {\mathrm{d}}^{-\mathrm{1}}.In the past 19 years， temperature has the largest contribution to the GPP， followed by precipitation， and by radiation again， the contribution of snow to the GPP is relatively weak.

Remote sensing technology has become an important way to obtain agricultural greenhouse coverage information quickly and effectively. But the spatial resolution size of remote sensing images has a dual influence on the extraction accuracy， and it is important to select suitable resolution images. Taking the southern agricultural plastic greenhouses as the research object， GF-1， GF-2 and Sentinel-2 are used to form six different spatial resolution image datasets between 1 and 16 m. Based on Object-Based Image Analysis （OBIA）， we use the Convolutional Neural Network （CNN） and Random Forest （RF） methods to extract the canopy and analyze the extraction accuracy and the difference between the methods. The results show that： （1） the extraction accuracy of agricultural greenhouses under CNN and RF methods generally decreases as the image resolution decreases， and agricultural sheds can be detected on images from 1m to 16 m； （2） the CNN method requires higher spatial resolution than the RF method， and the CNN method has fewer missed and false extractions at 1~2 m resolution， but at 4 m and lower resolutions， the RF method is more applicable； （3） the 2 m resolution image is the best spatial resolution for shed information extraction， which can realize shed monitoring economically and effectively.

Buildings are integral components of urban areas. Extracting buildings from high-resolution remote sensing data holds significant academic importance in areas such as land use analysis， urban planning， and disaster risk reduction. For the problems of building extraction， an improved Mask R-CNN building instance segmentation model is proposed. Based on the residual neural network fusion convolutional attention model， a residual convolutional attention network is constructed to improve the problem of inadequate feature extraction. The loss function is optimized by adding the Dice Loss method， and then the feature learning process is optimized. And a post-processing strategy combining Douglas-Peucker algorithm and Fine polygon regularization algorithm is introduced to make the building contours more regular and smooth. The experimental results show that the improved model improves the detection mAP value by 7.74% at Iou 0.5 and 7.57% at Iou 0.75 compared with the original model， and the post-processing strategy improves the F1-Score value by 6.01% compared with the original model after selecting the appropriate threshold to optimize the mask. The instance segmentation model coupled with Mask R-CNN and attention mechanism improves the small building misdetection and omission problem， building segmentation boundary adhesion problem， and building segmentation accuracy； building post-processing strategy， improves building regularization.

Object detection in remote sensing images has become a vital aspect of the overall object detection domain. To address the problems of missed detection and false detection of small-scale objects in high-resolution remote sensing images with complex backgrounds， a local adaptive feature weighting algorithm is proposed at the detection stage， combined with the YOLOv5s algorithm. By learning the existing label box information， the foreground containing target features is separated from the background， and the local features of the target with key information in the foreground are obtained. The spatial scale and weight of local features of each layer are calculated adaptively. Meanwhile， a global attention mechanism is proposed to enhance the interaction capability of cross-dimensional feature information between channels and spatial dimensions in the backbone， so as to strengthen the correlation between features and compensate for the loss of global information within local features at the detection stage， thereby reducing the rates of missed detection and false detection of targets. Experimental results show that the improved algorithm achieves certain improvements in precision and recall， with a mean average precision reaching 72.33%， representing an increase of 2.66% compared to the traditional YOLOv5s algorithm.