论文部分内容阅读
提高TM图像的分类精度,是图像处理及应用领域中一个很重要的研究课题。本文在总结已有成果基础上,首先利用现有的统计分类技术,对待分类图像进行预分类,并检测出“不确定”像元。然后综合光谱、地理、土壤类型、早期判别结果、目视判读经验等各种知识和信息,充分发挥专家系统的推理判断能力,对“不确定”像元的类别作进一步判别,使得整幅图像的分类精度得到改善。并据此初步建立了一个土地利用的分类系统。试验证明,这种分类方法的精度比仅用单一多光谱信息的统计分类法(最大似然法)提高约8％。
To improve the classification accuracy of TM images is a very important research topic in the field of image processing and application. On the basis of summarizing the existing achievements, this paper first classifies the classified images by using the existing statistical classification techniques and detects “uncertain” pixels. Then, based on various knowledge and information such as spectrum, geography, soil type, early discriminant result and visual interpretation experience, this dissertation makes full use of the expert system’s reasoning ability to judge the category of “uncertain” pixel, making the whole image The classification accuracy is improved. Based on this, a land use classification system was initially established. Experiments show that the accuracy of this classification method is about 8% higher than the statistical classification (maximum likelihood method) using only single multispectral information.