Detection of Detail Land Cover Change by Use of Digital Data from Air Survey

土地利用及其变化趋势是各类区域规划的基础数据。这意味着，根据规划或项目的目的，土地覆被数据的必要特征应该不同，或者适当的材料因项目而异。至于日本，我们已经有了土地覆被和土地利用图以及其他基本调查数据。然而，由于无法多次核对，调查数据的陈旧性是一个瓶颈。改善这种情况的一种方法是尽快检测变化的位置。在发现变化的地方后，不难对该地方的详细土地进行核查。基于这些考虑，研究了利用CCD线阵传感器获取的数据进行位置变化检测的方法。在一个完整的系统来检测变化的地方，主要研究的主题如下。(1)最重要的课题是几何精度。为了实现图像的预定位，在考虑地面测量误差特点的基础上，确定了定位参数。(2)利用图像相关性对数字立体匹配技术进行了改进，实现了自动化处理，但仍有少量误匹配现象发生。(3)研究了利用不同日期的两幅图像密度差产生的几种类型图像的变化检测方法。还包括主成分图像。在这项研究中，考虑简单的决策树。

Land coner land use and a trend of them are fundamental data for every type of regional planning. This means that according to the purpose of a planning or project, the necessary characteristics of land cover data should be different, or suitable materials differ from project by project. As to Japan, we already had land cover and land use maps and other fundamental survey data. However, it is a bottleneck that the survey data are outdated because we can not check up so many times. The one way to improve this situation is to detect changed places as soon as possible. After the changed places were found, It is not difficult to check up detail land converges in that places. Based on these considerations, Detection methods of changed places were investigated by use of the data obtained from CCD line sensors. In a total system to detect changed places, the main subjects investigated are as follows. (1) The most important subject is geometrical accuracy. In order to perform presize orientation of images, the orientation parameters are determined under consideration of the characteristics of errors in ground surveys. (2) Techniques of digital stereo-matching by use of image correlations are improved in order to realize an automatic processing, but miss-matchings still take place a little. (3) Change detection methods by use of several types of the images which are produced by dencity differnces of two images obtained on different dates are investigated. Principal component images are also included. In this study, simple decision trees are considered.