陆地卫星MSS图像地表水域信息的机助识别提取

本文介绍了一种利用陆地卫星MSS图像进行水域信息机助识别提取的快速、经济而有效的方法。该方法以对水与非水信息的光谱特征,比值图像的数据结构特点和大气校正的研究与分析为基础。 研究表明,水与非水的主要差异集中于MSS4与MSS7波段,这两个波段的比值图像可在压抑阴影的同时将两图像的信息集中于一。根据对比值图像数据特点的分析,发现比值图像在作等比例拉伸并取整时所不可避免的信息损失主要集中在低值区,高值区则相对得到扩展增强。由于在MSS4/MSS7图像中,水的信息位于高值区,因此其水域识别能力优于MSS7/MSS4。但是,即使MSS4/MSS7图像也无法完全排除山区深阴影对水域识别的干扰。在进行MSS4/MSS7比值运算之前,首先对MSS4图像作粗略的大气校正,则可圆满地解决这一难题。在经大气校正后的MSS4/MSS7图像上,水体像元的值大于1,而深阴影及其它所有非水信息像元的值则小于1,因此,只要以此为门限值将图像二值化,即可获得精度很高的水域识别图像。完全消除阴影和其它因素的干扰。 本方法运算简单,处理速度为最小距离分类的3倍以上,且精度比分类方法高。它不仅适用于MSS图像,也适用于TM图像。采用本方法进行湖北省地表水域机助识别和面积测量的实际应用表明,在各种地貌类型区内本方法均能以较高的精度完成

A TECHNIQUE FOR COMPUTER AIDED RECOGNITION AND EXTRACTION OF SURFACE WATER AREA INFORMATION USING MSS IMAGES

In order to measure the area of surface warer bodies in Hubei Province, a highly accurate and efficient technique for computer aided recognition of water information using Landsat MSS images was developed. The technique was designed based on the study of spectral signatures of water and non-water information, the data character of ratio images and the enhancement of atmospheric correction.According to the study of spectral signatures, the main information which differentiates water bodies from non-water objects concentrates on band MSS4 and band MSS7. The ratio image of two bands can both merge the information contained -in each band into one image and supperss shadow which is harmul to waiter recognition. Analysis of data character of latio images indicates that the information loss of a ratio image concentrates on the low value section of its value range, when the image is scaled and changed from real type to integer type. Whereas the information in high value section is relatively enhanced by the processing. As the pixels of water bodies possess high values in a MSS4/MSS7 images and low values in MSS7/MSS4 image, the MSS4/MSS7 is superior in water recognition to MSS7/MSS4. But even MSS4/MSS7 images can not completely eliminate dark shadow, the major obstacle to raise the accuracy of water recognition is in mountain areas. This problem is solved by introducing the atmospheric correction to MSS4 before MSS4/MSS7 ratio processing. In a MSS4/MSS7 image with atmospheric correction, the pixel values of water bodies are larger than 1, while the pixel values of dark shadow and other non-water information are less than or equal to 1. Therefore, the water recognition image with high accuracy can be obtained simply by thresholding.The technique is very simple and more precies than common used classification techniques. Its processing speed is more than three times faster than that of mini -distance classification. Besides, the technique can be used not only to MSS images but also to TM images. The successful water area recognition and measurement of whole Hubei Province has shown that the technique can produce perfect results in various areas of different geomorphic types.