景观破碎化区域土地覆被面向对象分类研究——以皖南山区为例

Study on land-cover classification using object-oriented method in landscape fragmentation zone: in the mountainous region of southern Anhui

以景观破碎化的皖南山区为试验靶区,基于2009年4期的Landsat 5 TM与HJ-1/CCD遥感影像,结合野外调查数据,分析土地覆被类型在影像上的特征,挖掘出相应的光谱、纹理、空间及时相等信息,获取相关特征波段.基于See 5.0软件,根据训练样本筛选特征波段及组合并获取最佳阈值.基于面向对象分类软件,对TM影像进行尺度10、30的2级分割,参照最佳特征波段及阈值建立规则集并获取分类数据.结果表明:基于面向对象的多时相分类法能精确地获取土地覆被数据,精度均高于92%,其在以皖南山区为代表的景观破碎化区域适用性较高.

Landsat 5 TM and H J-1 CCD images in 2009 were selected for land-cover classification in mountainous region of southern Anhui. Combining with field survey data, we analyzed land-cover characteristics to dig out spectrum, texture, spatial and temporal information, and obtained the feature bands of the corresponding land-cover types. Using See 5.0 software, we oltained the optimal feature bands and thresholds of land-cover types based on training samples. The rule sets of land-cover classification were established using eCognition Developer 8.64 software. The results showed that, by using the object-oriented classificationmethod based on multi-temporal remote sensing images, the land-cover classification with accuracies higher than 92% was acquired. This method is suitable for land-cover classification in the mountainous region of southern Anhui.