利用ERS-2 SAR图像纹理分析方法揭示长白山天池火山近代喷发物空间分布特征

简单介绍了SAR图像的纹理特征以及正交小波变换纹理提取方法。论述了SAR图像的纹理特征参与分类的重要性。以长白山天池火山为例,通过对ERS2SAR图像进行纹理分析,提取了SAR图像两个层次的尺度变化、时频局部化和方向性纹理特征。并将SAR纹理特征与TM图像及DEM进行复合,利用多源信息各自的优势,进行了BP神经元网络分类,从较大范围对长白山天池火山735±15aB.P.大喷发的喷发物空间分布进行评价。获取了长白山天池火山近代喷发物的空间分布及规模。这对长白山天池火山未来喷发危险性初步评价、火山地质制图及火山灾害预测有重要意义。

REVEALING DISTRIBUTION OF MODERN ERUPTION OF CHANGBAISHAN MOUNTAIN TIANCHI VOLCANO BY ERS-2 SAR IMAGE

The characteristics SAR image and algorithm of orthogonal cross course wavelet translation were introduced. On the basis of the previous work, the continuous wavelet transform, discrete wavelet transform and Mallat fast algorithm of orthogonal cross course wavelet disassemble of numerical images were studied; the BP neural network classification of remote sensing image, including network configuration, BP algorithm, network training and classification were analyzed. The importance of SAR image in classification was discussed. This article compared the outcomes of BP neural network classification with and without vector layer, and elicited that importing vector information such as geological data could supplement the spatial information of objects and reduce effectively the cases such as 'diverse objects with same spectrum' and 'multi spectrum for same objects'. In the Tianchi volcano, Changbaishan Mountain, the data of crater, lithology, chronology, contour, fault, earthquake, gravitation are collected, by digitalization, GIS vector databases were established. TM and ERS 2 SAR images and pant aerial photos were collected also, through the previous handling, the images databases under the Mapinfo platform were established. By TM image the linear structures, water system, the outer edge of Tianchi volcano trachite were extracted. By the sensitivity of SAR images to geomorphology, the distribution of parasitical volcano cone and lava fornix around Tianchi volcano were discussed. With the aerial photos the parasitical lava of Qixiangzhan, Baiyun and Bingchang stage were filtered. At the same time by orthogonal cross course wavelet transform, texture analysis was carried on to ERS 2 SAR images, and the texture characters were filtered. On the basis of the spectrum character of multi band of TM images, multi band algebraic grouping were made, and the factor of texture character factor and band grouping of TM images were input to BP neural network classification model. The good classified results were received. Finally, with the geological data, the eruptible stages of cone forming period of Tianchi volcano, part parasitical eruptible stages and the distribution of recent eruptible materials were obtained. The results will be helpful for the volcanic hazard assessment, volcanic geological mapping and hazard prediction.