Please wait a minute...
img

官方微信

遥感技术与应用  2018, Vol. 33 Issue (6): 1073-1083    DOI: 10.11873/j.issn.1004-0323.2018.6.1073
数据与图像处理     
结合LiDAR单木分割和高光谱特征提取的城市森林树种分类
皋厦1,申鑫1,代劲松2,曹林1
(1.南京林业大学 南方现代林业协同创新中心,江苏 南京 210037;
2.浙江省森林资源监测中心,浙江 杭州 310020)
Tree Species Classification in Urban Forests based on LiDAR Point Cloud Segmentation and Hyperspectral Metrics Extraction
Gao Sha1,Shen Xin1,Dai Jinsong2,Cao Lin1
(1.Co-innovation Center for Sustainable Forestry in Southern China,Nanjing Forestry University,Nanjing 210037,China;
2.Center for Forest Resource Monitoring of Zhejiang Province,Hangzhou 310020,China)

 全文: PDF(4847 KB)  
摘要: 基于多源遥感数据的城市森林树种分类对城市森林资源调查、森林健康状况评价及科学化管理具有重要意义。以江苏省常熟市虞山国家森林公园内的典型城市森林树种为研究对象,利用同期获取的机载激光雷达(LiDAR)和高光谱数据,针对5个典型城市森林树种进行了树种分类的研究。首先,基于点云距离判断单木分割方法进行单木位置和冠幅提取,并借助实测数据和目视解译结果进行精度验证;然后,在冠幅内提取4组高光谱特征变量,并借助随机森林模型对特征变量进行重要性分析;最后,筛选出重要性高的特征变量进行2个级别的树种分类并借助混淆矩阵进行验证评价。结果表明:基于点云距离判断分割方法的单木位置提取精度较高(探测率为85.7%,准确率为96%,总体精度为90.9%);利用全部特征变量(n=36)对5个树种进行分类,分类的总体精度达到了84%,Kappa系数为0.80;利用优选特征变量(n=9)进行分类,总体精度达83%,Kappa系数为0.79;利用全部特征变量(n=36)对两种森林类型进行分类,分类的总体精度达91.3%,Kappa系数为0.82,其中阔叶树种分类精度为95.6%,针叶树种分类精度为85%;利用优选特征变量(n=9)进行分类,分类的总体精度达90.7%,Kappa系数为0.80,其中阔叶树种分类精度为93.33%,针叶树种分类精度为86.67%。
关键词: 城市森林树种分类激光雷达高光谱随机森林
    
Abstract: Urban forest tree species classification using multi-source remote sensing data plays a key role in urban forest resources investigation,forest health assessment and scientific management.This study selected typical tree species in Changshu Yushan forest as research objects.The five tree species were classified using combined airborne hyperspectral and LiDAR data which acquired simultaneously.First,the positions and crowns of individual trees were extracted from LiDAR data based on Point Cloud Segmentation method (PCS) and validated using field and visual interpretation data;second,the four sets of hyperspectral metrics were extracted from hyperspectral data and the importance of metrics were assessed using Random Forest algorithm;finally,the tree species were classified in two levels using Random Forest algorithm and accuracies were evaluated by confusion matrix.The results indicated that the PCS approach had high accuracy (Detection Rate =85.7%,Precisio n=96% the Overall Accuracy=90.9%) in the extraction of individual tree positions;the overall accuracy of five tree species classification using all metrics (n=36) was 84%,Kappa coefficient was 0.80;the overall accuracy of five tree species classification using the optimal metrics (n=9) was 83%,Kappa coefficient was 0.79;the overall accuracy of two forest types classification using all metrics (n=36) was 91.3%,Kappa coefficient was 0.82,the overall accuracies of conifer and broadleaved tree species were 85% and 95.6% respectively;the overall accuracy of two forest types classification using the optimal metrics (n=9) was 90.7%,Kappa coefficient was 0.80,the overall accuracies of conifer and broadleaved tree species were 86.67% and 93.33% respectively.
Key words: Urban forest    Tree species classification    LiDAR    Hyperspectral data    Random Forest
收稿日期: 2018-05-17 出版日期: 2019-01-29
基金资助:

国家重点研发计划项目(2017YFD0600904),江苏省自然科学基金项目(BK20151515),国家自然科学基金项目(31770590),江苏省高校优势学科建设工程资助项目(PAPD)。

作者简介: 皋厦(1994-),男,江苏盐城人,硕士研究生,主要从事激光雷达树种分类研究。Email:njfugaosha@qq.com。
服务  
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章  

引用本文:

皋厦, 申鑫, 代劲松, 曹林. 结合LiDAR单木分割和高光谱特征提取的城市森林树种分类[J]. 遥感技术与应用, 2018, 33(6): 1073-1083.

Gao Sha, Shen Xin, Dai Jinsong, Cao Lin. Tree Species Classification in Urban Forests based on LiDAR Point Cloud Segmentation and Hyperspectral Metrics Extraction. Remote Sensing Technology and Application, 2018, 33(6): 1073-1083.

链接本文:

http://www.rsta.ac.cn/CN/10.11873/j.issn.1004-0323.2018.6.1073        http://www.rsta.ac.cn/CN/Y2018/V33/I6/1073

[1] 陈伟民,张凌,宋冬梅,王斌,丁亚雄,许明明,崔建勇. 基于AdaBoost改进随机森林的高光谱图像地物分类方法研究[J]. 遥感技术与应用, 2018, 33(4): 612-620.
[2] 廖凯涛,齐述华,王成,王点. 结合GLAS和TM卫星数据的江西省森林高度和生物量制图[J]. 遥感技术与应用, 2018, 33(4): 713-720.
[3] 虢韬,沈平,时磊. 机载LiDAR快速定位高压电塔方法研究[J]. 遥感技术与应用, 2018, 33(3): 530-535.
[4] 苏阳,祁元,王建华,徐菲楠,张金龙. 基于航空高光谱影像的额济纳绿洲土地覆被提取[J]. 遥感技术与应用, 2018, 33(2): 202-211.
[5] 秦振涛,杨茹,张靖,杨武年. 基于聚类结构自适应稀疏表示的高光谱遥感图像修复研究[J]. 遥感技术与应用, 2018, 33(2): 212-215.
[6] 郭宇柏,卓莉,陶海燕,曹晶晶,王芳. 基于空谱初始化的非负矩阵光谱混合像元盲分解[J]. 遥感技术与应用, 2018, 33(2): 216-226.
[7] 刘爱林,郭宝平,李岩山 . 基于离散粒子群算法的凸多模态高光谱图像端元提取研究[J]. 遥感技术与应用, 2018, 33(2): 227-232.
[8] 吴兴,张霞,孙雪剑,张立福,戚文超. SPARK卫星高光谱数据辐射质量评价[J]. 遥感技术与应用, 2018, 33(2): 233-240.
[9] 孟庆岩,孙云晓,张佳晖,等. 基于多源遥感的建筑区绿化垂直分布特征与空间配置分析——以匈牙利塞克什白堡市为例[J]. 遥感技术与应用, 2018, 33(2): 370-376.
[10] 宋婷婷,付秀丽,陈玉,魏永明,王钦军,程先锋. 云南个旧矿区土壤锌污染遥感反演研究[J]. 遥感技术与应用, 2018, 33(1): 88-95.
[11] 刘慧珺,苏红军,赵-波. 基于改进萤火虫算法的高光谱遥感多特征优化方法[J]. 遥感技术与应用, 2018, 33(1): 110-118.
[12] 梁茜茜,张汉德,孙根云,王鹏. 基于机载激光雷达数据的海岸带水域提取方法[J]. 遥感技术与应用, 2018, 33(1): 136-142.
[13] 李伟娜,韦玮,张怀清,刘华. 基于多角度高光谱数据的高寒沼泽湿地植被生物量估算[J]. 遥感技术与应用, 2017, 32(5): 809-817.
[14] 王濮,邢艳秋,王成,习晓环,骆社周. 机载LiDAR数据提取山区道路方法研究[J]. 遥感技术与应用, 2017, 32(5): 851-857.
[15] 肖昊,王杰. 基于IDL和MATLAB混合编程的两种光谱混合分析方法比较[J]. 遥感技术与应用, 2017, 32(5): 858-865.