倾斜影像整体变分模型阴影检测算法改进  

Improved Shadow Detection Algorithm Based on Total Variation Model Using Oblique Images

作　　者：闫利[1] 莫楠[1] 费亮[1] 朱睿希 

机构地区：[1]武汉大学测绘学院,武汉430079

出　　处：《遥感信息》2017年第2期54-59,共6页Remote Sensing Information

基　　金：测绘地理信息行业科研专项项目(201512008)

摘　　要：利用整体变分模型进行倾斜影像的阴影检测时,绿色植被的误检测会降低阴影检测的正确率,对此,提出一种适用于包含复杂地物的倾斜影像阴影检测算法。算法引入可见光波段差异植被指数(visible difference vegetation index,VDVI)去除植被对阴影像素的干扰,并对检测结果进行有效补偿。首先,利用VDVI提取影像中植被区域,剔除阴影中错误的植被像素,提高算法阴影检测的正确率;然后,利用Wallis滤波算法对检测出的阴影区域进行补偿,使其像素灰度与周围非阴影像素具有一致性;最后,对Wallis补偿后的结果采用渐入渐出插值算法对影像边缘效应进行处理。以倾斜影像作为实验数据,对改进后的算法进行阴影检测精度评定与补偿测试。实验证明,改进的阴影检测算法性能得到提升,整体检测率达到97.24%,并且经过边缘处理的阴影区域具有较好的目视补偿效果。While we use the total variation model for shadow detection of oblique images, error detection of green vegetation may decrease the probability of shadow detection, Therefore, we propose a shadow detection algorithm for oblique images containing complex ground objects. The algorithm introduces the visible difference vegetation index （VDVI）to remove the disturbance of vegetation mistaken as shadow pixels, First of all, we improve the total variation model shadow detection algorithm by using VDVI to extract vegetation areas of the images and eliminating shadow vegetation error pixels, to improve the accuracy of the algorithm shadow detection, Then we apply the Waliis filter to the detected shadow area to compensate for the texture of shadow pixels,stretching gray value of shadow areas so that the mean gray value and variance of shadow areas keep consistent with those of non-shaded pixels. Finally,we implement a gradually fading out interpolation algorithm on results after compensation to handle edge effects of images. Using oblique image as the experimental data, we perform a test for improved algorithm to get the detection accuracy and compensation results. Experimental resutls show that the improved shadow detection algorithm performs better with the overall shadow detection rate of 97.24 % and good visual compensation effect after edge processing.

关 键 词：倾斜影像 阴影检测 阴影补偿 边缘处理 整体变分模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]