中文摘要：

提出了以CBERS-02B星CCD图像为数据源的林带提取方法。以MSAVI作为特征指数,经过边缘检测,PC波谱融合生成了较为连续完整的林带边缘线。利用灰度共生矩阵的均值去除面状的果园、苗圃和部分芦苇,从而构建了基于中巴影像的干旱区林带信息提取模型。对仍然没有检测出的极少数不清晰边缘进行了膨胀处理,使其与已检测出的边缘相连接,最后对提取结果矢量化。结果表明：整体提取准确度为91.3%,矢量化结果可靠。文中方法较好地解决了不清晰边缘的林带提取,可以用于CBERS-02B星CCD图像林带的自动提取,对其它线状体的提取也具有借鉴意义。

英文摘要：

In this research we proposed forest belt extraction method with CBERS02B as its data source. Using MSAVI as the characteristic index,by edge detection algorithm,fusing of PC spectrum to enhance the forest belt information. By using the mean value of grey scale coexistence matrix,the artifacts caused by orchards,nurseries and reeds could be removed from the image. We developed a forest belt extraction model based on images of CBERS02B,for few of those undetected and unclear edges,they were expanded to be connected with the edges that had been detected. In the end,the detection results were vectorized. Experimental results show that the overall extraction accuracy is 91. 3%,indicating the factorization results are quite reliable. This proposed method can effectively solve the forest belt extraction with unclear edges,and thus it can be used for automatic extraction of forest belt in arid region,making it widely applicable.