准确、快速、低成本的获取高潜水位煤矿区沉陷耕地的面积、分布、受损等信息对耕地保护有重要的意义。色彩空间转换、纹理分析和植被指数等方法能够有效的增强和挖掘影像潜在的信息，对信息提取有很大帮助。利用2018年4月获取的无人机可见光影像对典型高潜水位煤矿区——山东兖州兴隆庄煤矿的沉陷耕地进行了提取研究。首先统计了耕地、积水区等地物在可见光三波段的均值和标准差，比较发现耕地与积水区在红、绿、蓝3个波段均有重合。其次对研究区影像进行了色彩空间转换与二阶矩阵纹理滤波，统计了耕地与积水区共27项色彩与纹理特征指标，利用均值和标准差计算了变异系数和相对差异值，最终选取色度(变异系数26%，相对差异73.33%)和绿色信息熵(变异系数20.59%，相对差异72.79%)作为耕地提取的最优特征，采用最大似然法进行耕地提取。之后计算了备选的6种可见光植被指数，根据结果分布图，选取了过绿指数EXG(excess green index)、可见光差异植被指数VDVI(visible-band difference vegetation index)、红绿蓝植被指数RGBVI(red green blue vegetation index)及归一化绿红差异指数NGRDI(normalized green-red difference index)作为沉陷耕地提取指数，利用双峰阈值法确定了耕地提取阈值。比较提取结果得出，EXG和NGRDI指数无法全面、客观反映研究区实际情况，VDVI指数的耕地总体分类精度为81.05％，高于RGBVI指数的71.38％，是本研究中最适用于高潜水位煤矿区沉陷耕地提取的指数。最后利用验证区影像，以基于样本面向对象提取的沉陷耕地面积作为参考值，通过比较面积及误差得出，基于色彩与纹理特征法提取的面积与参考面积更接近，误差(6.8%)小于可见光植被指数法(16.0%)，更适用于高潜水位煤矿区沉陷耕地的提取。本研究结果客观反映了由于采煤沉陷导致耕地颜色、纹理、疏密等变化特征，为高潜水位煤矿区沉陷耕地的面积测算提供了技术支持。

Accurate,fast and low-cost access to the information on the area,distribution and subsidence damage of cultivated land in high-groundwater-level coal mines is very important for the protection of cultivated land. UAV (Un- manned Aerial Vehicle) remote sensing has great advantages over traditional methods,with lower cost,simpler opera- tion,faster access speed and higher resolution. Color space conversion,texture analysis and vegetation index and other methods can effectively enhance and extract the potential information of the image,which is helpful to the image ex- traction. Taking Xinglongzhuang coalmine area,located in Yanzhou District,Shandong Province as an example,this pa- per focused on the extraction of subsided cultivated land information in high-groundwater-level coal mines using UAV visible-band images obtained in April 2018 as the data source. Firstly,the statistical characteristics ( including mean and standard deviation) of pixels of typical ground objects in red,green,and blue bands were analyzed. It was found that the pixel values of cultivated land and water catchment part were overlapped in all three bands. Secondly,the color space conversion and the gray level co-occurrence texture filtering were performed on the image of the study area,and a total of 27 color and texture features were obtained,which were then calculated by mean and standard deviation to gain respective coefficient of variation and relative difference. The hue (the coefficient of variation is 26% ,the relative difference is 73. 33% ) and the entropy of blue ( the coefficient of variation is 20. 59% ,the relative difference is 72． 79% ) were chosen as the optimal feature for cultivated land extraction,then the maximum likelihood method was used to extract cultivated land information. Thirdly,six alternative visible-band vegetation indexes were calculated,and EXG (excess green index),VDVI (visible-band difference vegetation index),RGBVI ( red green blue vegetation in- dex) and NGRDI (normalized green-red difference index) were finally selected as the subsided land extraction index, besides the dual-peak threshold method was used to determine the extraction threshold of cultivated land. By the com- parison of the extraction results,it is found that the EXG and NGRDI index cannot objectively reflect the characteris- tics of the study area. The total cultivated classification accuracy of VDVI index (81. 05% ) is higher than that of RG- BVI index (71. 38% ),which is the most suitable index for the extraction of cultivated land. Finally,the area of subsi- ded cultivated land extracted from the remote sensing image of the study area using object-oriented extraction based on sample was used as the reference value for verification,and the area extracted by color and texture feature method (2 079. 16 m2 ) is closer to the reference area (2 231. 34 m2 ),which is more suitable for the extraction of cultivated land in high-groundwater-level coal mines. The results objectively reflect the variation features of the color,texture and density of cultivated land caused by coal mining subsidence,which provides basic data and reference method for the calculation of subsided cultivated land in high-groundwater-level coal mines.