Remote Sensing Technology and Application

Different Spatial Resolutions based on Object-oriented CNN and RF Research on Agricultural Greenhouse Extraction from Remote Sensing Images

Xinyi LIN, Xiaoqin WANG, Zixia TANG, Mengmeng LI, Ruijiao WU, Dehua HUANG

Remote Sensing Technology and Application, 2024, 39(2): 315-327. DOI: 10.11873/j.issn.1004-0323.2024.2.0315

特征类别	特征名称	数量
光谱特征	光谱最大差分、各波段均值（Mean B、Mean G、Mean R、Mean NIR）、亮度值、归一化植被指数、归一化差值湿度指数	8
纹理特征	均值、方差、熵、同质度、对比度、非相似性、角二阶矩、相关性(0°、45°、90°、135°)	32
几何特征	面积、长宽比、边界长度、不对称性、形状指数、密度、主要方向、紧凑度、圆、边界指数	10

Table 2 Feature variables for agricultural greenhouse extraction

Other figure/table from this article

Fig.1 Distribution of key research areas and sample points in Fu'an City
Table 1 Remote sensing image acquisition in the study area
Fig.2 Example of local true color band combinations of canopies with different spatial resolution images
Fig.3 Basic structure of the residual unit
Fig.4 Object-oriented CNN method（OCNN_FT） flowchart
Fig.5 Fine-tuning results of CNN models on different resolution images
Fig.6 Greenhouse distribution based on RF method and OCNN_FT method
Fig. 7 Detailed diagram of classification effect of different methods
Fig.8 PA， UA， OA and F-score of agricultural greenhouses under different resolution images
Table 3 McNemar test results based on RF method and OCNN_FT method（significance α=0.05）