多源遥感影像提取火烧迹地面积不确定性研究

Uncertainty study of fire scar area extraction with different image sources

利用遥感影像计算火烧迹地面积是火灾评估的常用方法,为了论证不同影像提取火烧迹地面积的差异,以4种国产卫星影像作为数据源,提出区分指数和确定度指数分析火烧迹地与林地在不同影像中的区分程度,并利用NDVI多阈值和混合像元分解方法分别提取火烧迹地面积。研究结果表明：高分辨率遥感影像对火烧迹地和林地的区分性较好,并且采用NDVI指数的区分程度高于近红外波段,NDVI多阈值提取随着阈值变化在不同影像中结果差异较大,而混合像元分解提取结果一致性较好。

Calculating the fire scar area with remote sensing images is a commonly used method of fire risk assessment.In order to demonstrate the differences of fire scar area extraction with difference images,we use the discrimination index and uncertainty index of fire scar and unburned forest to analyze distinguishing degree in different images,multi thresholds of NDVI and mixed pixel decomposition method to extract fire scar area respectively.The results show that the distinction between fire scar and unburned forest is better in high resolution remote sensing image,and the distinction using NDVI index was higher than that using the near infrared band.The extraction results with NDVI vary with threshold changes while mixed pixel decomposition extraction results perform in good agreement.