光学学报, 2019, 39 (3): 0315003, 网络出版: 2019-05-10
基于区域辐射一致性的移动阴影检测 下载: 860次
Moving Shadow Detection Based on Regional Radiation Consistency
机器视觉 运动目标检测 移动阴影检测 区域生长 machine vision moving object detection moving shadow detection region growth
摘要
为了同时检测本影与半影区域,提出并证明了阴影区域辐射的一致性属性。获取超像素区域轮廓内的点集合,将像素点集合(PCC)分为目标前景区域(目标PCC)和阴影区域(阴影PCC),利用所提出的基于区域生长的完整阴影检测与目标掩码增长算法,通过融合完整的阴影区域、完整的目标前景区域和ViBe掩码这三个部分,实现了前景目标掩码反向增长。在公开数据集中的实验结果表明,所提方法的阴影检测平均精度达到了82.5%,性能显著优于传统方法。目标掩码的平均增长率达到了8.84%,准确率达到了95%以上。
Abstract
In order to simultaneously detect the shadow and penumbra regions, this study proposes and proves the consistent properties of shadow region radiation. Point collection in contour (PCC) within the outline of the super-pixel region corresponding to moving object is obtained, and the PCC is divided into the object foreground area (foreground PCC) and the shadow area (shadow PCC). With the proposed region-based full shadow detection and object mask growth algorithm, the foreground object mask is inversely grown by combing the complete shadow region, the complete foreground region and the ViBe mask of the moving object. The experimental results in the public dataset show that the average accuracy of the shadow detection of the proposed method reaches 82.5%, and the performance is significantly better reaches the traditional method. The average growth rate of the object mask reaches 8.84%, and the accuracy rate reaches over 95%.
陈海永, 郄丽忠, 刘坤. 基于区域辐射一致性的移动阴影检测[J]. 光学学报, 2019, 39(3): 0315003. Haiyong Chen, Lizhong Qie, Kun Liu. Moving Shadow Detection Based on Regional Radiation Consistency[J]. Acta Optica Sinica, 2019, 39(3): 0315003.