基于中巴卫星影像的土地利用分类精度评价

Accuracy assessment methods of classification results of land use monitoring based on CBERS-02 satellite image

本文对某地区的中巴地球资源一号02星CCD数据进行监督分类,并将影像分类结果进行精度评价,得到了中巴影像分类传统的精度报告。在此基础上,为了使精度可信性更高,提出了一种新的精度评价方法;即在中巴影像分类后结果图上随机抽取其图斑总数的5%,将其与研究地区土地利用数据库结合ArcGIS,进行叠加比较,对抽取出的各图斑类别进行面积对比,在一定程度上评价了中巴影像的分类精度,并对分类结果进行了量化的精度评价,从而丰富了精度评价的内容。从研究结果表明,经过面积对比,可以较精确地看出中巴影像分出的沙地、城市用地及冬小麦的精度比较高,但村庄和有些耕地光谱特征相似难以区分。

In this article,taking some region as test field, the CBERS-2 CCD image is supervised classified.After precision evaluation,a classification report of the CBERS-2 image is obtained.Then,in order to improve precision,a new evaluation method is proposed.With this method,the authors extract the 5% composition spots of the total classification picture of the CBERS-2 image,and overlay the classification plot on the land use database of 1: 10,000 in ArcGIS.Based on area compare,the classification precision of CBERS-2 image is evaluated.Test result shows that CBERS-2 image can exactly classify the sand,the city and wheat from area compare.But village and farming land are difficult to be distinguished as they are in the similar spectrum.