基于无人机遥感的高潜水位采煤沉陷湿地植被分类

为了掌握采煤沉陷湿地植被的类别和空间分布,促进矿区土地利用、管理和修复,以山东省济宁市东滩煤矿3304工作面为研究区,以无人机多光谱影像为数据源,分别采用面向对象的分类方法和监督分类方法对研究区湿地植被进行分类。基于优选的面向对象尺度分割参数,确定分类规则后构建面向对象分类模型,对湿地植被进行分类,生成植被分布图。同时,利用野外获取的322个采样点进行精度验证。结果表明:与基于像元的监督分类方法相比,面向对象分类方法显著提高了影像分类精度。监督分类方法总体精度为44. 3%,Kappa系数为0. 4;面向对象分类方法总体精度达到84. 2%,Kappa系数为0. 8。该研究为采煤沉陷区湿地调查与开采沉陷影响下地表植被空间分布规律研究提供了方法与基础数据。

Vegetation Classification by Using UAV Remote Sensing in Coal Mining Subsidence Wetland with High Ground-water Level

After mining in the high ground-water level mining area, the surface subsided and accumulated water. The surface is changed from the farmland ecosystem to the water-land two-phase ecosystem. As the energy fixers and nutrient producers in the wetland ecosystem, wetland vegetation can reflect the changes in the wetland ecological environment. Vegetation classification is the basis for exploring vegetation coverage and monitoring dynamic changes. In order to grasp the type and spatial distribution of vegetation in coal mining subsidence and promote land use, management and restoration in mining area, totally 3304 working face of Dongtan Coal Mine in Jining City, Shandong Province was selected as the study area. The UAV multi-spectral images were taken as data sources, and the object-oriented classification method and supervised classification method were used to classify the wetland in the study area. Based on the optimized object-oriented scale segmentation parameters, the classification rules were determined and then the object-oriented classification model was constructed to classify the wetland vegetation and generate the vegetation distribution map. At the same time, totally 322 sampling points were used to verify the accuracy of the classification results. The results showed that the overall accuracy of the supervised classification method was 44.3%, and the object-oriented classification method was 84.2%. Compared with the supervised classification method which based on pixels, the object-oriented classification method improved the classification results and significantly improved the image classification accuracy. The Kappa coefficient of supervised classification was 0.4, while the Kappa coefficient of object-oriented classification was 0.8. The research result provided a new method and basic data for the investigation of wetlands in coal mining subsidence area and the study of the spatial distribution of vegetation under the influence of mining subsidence.