【作者】 徐丽华；

【导师】 江万寿；

【作者基本信息】 武汉大学 ， 地图学与地理信息系统， 2005， 硕士

【摘要】 近年来,随着生产力的发展和社会的进步,城市建设也得到了快速的发展。为了及时地了解城市布局、土地使用、违章建筑区域、非法占地、地面绿化状况以及道路小局等情况,需要对城市的变化进行适时监测。随着遥感技术的迅速发展,利用遥感影像进行城市变化检测科学研究已经越来越普遍。 卫星遥感影像的低分辨率不能满足对城市内部结构分析的要求,它不适合其中较详细的变化检测(如房屋、道路的变化检测)工作,而高分辨率遥感影像更能反映较详细的变化,比较适合城市变化检测的需求,所以高分辨率遥感在城市变化检测有着巨大的应用潜力。 一般地,影像的配准精度直接影响变化检测的结果,这种影响在利用高分辨率遥感影像进行变化检测的过程中表现更为明显。一方面,由于影像还不能实现完全配准,由于投影差引起的配准误差就会被当作变化检测出来。另一方面引起投影差的建筑物等物体,一般也会引起阴影,导致灰度的局部变化。因此,在利用高分辨率遥感影像进行变化检测过程中,必须考虑投影差及阴影对变化检测结果的影响。 目前,利用高分辨率遥感影像进行城市地区的变化检测的方法十分有限,而对于建筑物引起的投影差和阴影对变化检测结果的影响,目前国内还没有相关研究。 本文针对不同时期高分辨率遥感影像进行城市变化检测,在利用传统变化检测方法得到的检测结果的基础上,考虑消除投影差和阴影对变化检测结果的影响。其主要内容包括: (1) 对遥感影像的变化规律进行了分析,并对传统的变化检测方法进行了归纳和总结。 (2) 实现了归一化影像差值变化检测方法,并针对全局辐射校正的不足,提出了迭代辐射校正方法来纠正不同时期影像之间的灰度变形。该方法并在对不同时期的机场影像的变化检测中取得了较好的检测效果。 (3) 将纹理特征差值变化检测应用于高分辨率遥感影像,并对该方法进行了实验和分析。 (4) 提出了消除或者减少阴影对变化检测结果的影响的方法。首先对不同时期的影像进行传统方法的变化检测;然后对不同时期的影像用简单的二值化提取阴影,比较检测结果中由于阴影引起的变化区域与阴影影像中阴影区域形状的差异,选择形状较为相近的阴影影像,来去除阴影对变化检测结果的影响。 (5) 提出了消除或者减少投影差对变化检测结果的影响的两种方法。一种是基于移动变化率的变化检测方法。在传统变化检测的基础上,对检测出变化的像素进行整体平移,得到不同的平移后的影像。通过平移后的影像与原始影像进行变化检测得到不同的检测结果。统计平移对检测结果影响的规律,找到投影差偏移的方向。在偏移方向上,最大限度的消除投影差对变化检测结果的影响。另一种方法是基于区域变化率的变化检更多还原

【Abstract】 With the increase of production and development of society in this years, especially the construction of city is more highlighted .In order to learn from the layout of city , the utility of land ,illegal buildings, the conditions of vegetation covers and layout of road so on, it is necessary to monitor the change of city. With the help of RS, the method of detecting the urban change by RS images is more and more popular.But the traditional RS images have a shortcoming: low resolution, so they can not be used for precise change detection (such as the change detection of road and house etc.).In fact, the high resolution satellite images can reflect the detail of change, which is fit for detection of city change. Apparently, high resolution satellite images will make a important pole in detection of city change.Generally, the precision of registering makes direct effect on the result of detection, especially for high resolution satellite images. On the one hand, because of misregistration in local regions, the error resulted from projection may be regarded as change. On the other hand, the buildings which cause the error of projection, always give rise to shade, then lead to the change of local gray. So during the process of detection by high resolution satellite images, the effect of projection and shade must be taken into accounted.In this paper, by use of high resolution satellite images at different points of time, based on the traditional method, take projection and shade into the effect of change detection. The main contents as follows:1. analyze the law of RS image change, and summary the traditional detection method.2. Implement the detection of method by image regression, and the experiments which covers various timely images of airplane prove its superiority.3. Exploiting the texture difference method on the detection of high resolution RS images, and make some experiments and analysis.4. Put forward to the method to decrease the effect resulted from shade. Firstly, apply the traditional detection method, then extract the shade by producing binary images, furthermore compare the changed regions resulted from shade with the shape difference in shade images, and select the similar shade images in shape for eliminate the effect of shade.5. Put forward to two methods for reducing the effect of projection difference. The first method is based on shift change frequency. On the base of traditional detection effect, translate all the pixels changed, and get the various images. By the detecting the shifted images and origin images, get the various change images. Then by deducing the effect resulted from translation, find the direction of projection difference offset. In the direction we can reduce effect of projection difference .The second method is based on region change frequency. Firstly segment the images ,then according to the preliminary result ,make a statistics on the change frequency of various regions for determining whether there are some change.更多还原