在中国快速城市化进程中，农村居住人口减少的同时居住用地却呈增加趋势，这一现象在南方山区尤为明显。利用遥感技术进行农村居民地信息的提取与分析具有重要意义。本研究选取江西省泰和县为典型研究区，基于RapidEye卫星影像，利用面向对象的分类方法进行居民地的分类提取。构建的分类规则集包括归一化植被指数NDVI、归一化水体指数NDWI、亮度以及长/宽比。通过多次实验，图像的最佳分割尺度设置为200。通过设置不同的分类规则参数，提取了研究区的居民地及其他土地利用类型，并进行分类结果验证和精度评价。研究区的居民地分类总体精度为78.40%，生产者精度和用户精度分别为68.75%和77.33%。结果表明通过利用面向对象的分类方法，RapidEye 卫星影像为农村居民地信息提取提供了适宜数据源。与第二次全国土地调查结果相比，存在93.67公顷的分类差异，主要分布在居民地与其他土地利用类型的毗邻地区。%The process of rapid urbanization in China features two opposing trends: declining rural population and increasing rural residential land, especially in southern hilly areas. The extraction and analysis of residential land in rural China represents an important application for remote sensing technology. The study aimed to discover rural residential land information using RapidEye satelite imagery, taking Taihe County as the research area in the hily region of southern China. Based on multiple experiments, classification was conducted with an optimal image segmentation scale set to 200. The object-oriented classification rule set was constructed using the customized parameters NDVI, NDWI, brightness, and length/width. The areas of residential land and other land use types were interpreted by varying the parameter values for classification rule sets. Finaly, validation and accuracy evaluations were carried out. The overal accuracy of residential land interpretation is 78.40%, and producer’s accuracy and user’s accuracy are 68.75% and 77.33%, respectively. The results indicate that RapidEye provides a suitable data source for extraction of rural residential land using an object-oriented approach. Compared with the second national land survey of China, the classification gave an absolute difference of 93.67 ha residential land within the study area. Recognition errors occurred mainly in regions adjacent to the boundaries between residential land and other types of land.
展开▼
