ASTER数据对雀儿山区土地覆盖分类的有效性评价

Effectiveness Evaluation of Land Cover Classification in the Queer Mountains by ASTER Satellite Data

用ASTER卫星的可见光-近红外(VNIR)的3个波段、短波红外(SWIR)6个波段和热红外(TIR)的5个波段对四川西部雀儿山区土地利用覆盖进行分类,并结合地面实地调查数据对不同波段分类效果进行了数量精度分析和空间分布评价。研究结果表明,与地面实地调查的统计数据比较,用经过几何纠正后全部14个波段组合(方法1)的分类精度达83.9%以上;用波段1、2、3、4、6、9、10和13(方法2)的分类精度介于76.6%～92.6%;用波段1、3、6、8、12和14(方法3)的分类精度介于85.6%～97.5%;仅用可见光-近红外3个波段和波段4、6和9的短波红外波段组合(方法4)的分类效果极差。因此,在雀儿山这种相对高差极大的山区土地利用覆盖分类必须采用能灵敏探测地表温度差异的热红外波段数据。

Three bands of the Visible and Near-infrared （VNIR）, 6 bands of the Shortwave infrared （SWIR） and 5 ones of the Thermal infrared （TIR） from ASTER satellite data were used to study the spectral features and classification of the land cover in the Queer mountains of the west of Sichuan province, China. Based on the data of surface survey, we also evaluated their quantitative precision and accuracy of spatial distribution. The results indicated that some band combinations of ASTER data are effective on the land cover classification. The classification precisions of the 14 band combination （case 1） after geometric correction was more than 83.9%. In comparison with the result, the precisions of the combination of band no. 1, 2, 3, 4, 6, 9, 10 and 13 （case 2）, and the one ofl, 3, 6, 8, 12and 14 （case3）, were 76. 6%-92. 6% and 85. 6%-97. 5% respectively. The one of band no. 1, 2, 3,4,6,9 （case 4） was not effective on the land cover classification in the mountains. Therefore, the bands of TIR are necessary for the mountains with more than 2000 m relative altitude to classify land cover. From the accuracy of spatial distribution of the land cover types, road, residential land, cropland and rock were easy to be misinterpreted for each other, and rock was also misunderstood as the residential or cropland.