融合色彩比和梯度不变性的运动阴影检测

发布时间：2018-05-03 03:29

本文选题：运动阴影检测 + 色彩比不变性　；参考：《计算机工程与应用》2017年22期

【摘要】：为解决运动前景的准确分割受运动阴影影响的问题,提出了一种融合色彩比和梯度不变性的运动阴影检测算法。该算法分析了阴影像素的色彩比和区域纹理梯度的光照不变性,利用亮度变化特性和色彩比不变性初步确定候选运动前景中的阴影像素,然后在候选阴影区域利用纹理梯度不变性进行去错处理,两者的结合弥补了单一特征或单一类型特征的阴影检测性能差的缺陷,提高了阴影检测率和阴影分辨率,能够准确地将阴影和前景区别开来。
[Abstract]:In order to solve the problem that the accurate segmentation of motion foreground is affected by motion shadow, a new algorithm of motion shadow detection, which combines color ratio and gradient invariance, is proposed. The algorithm analyzes the color ratio of shadow pixels and the illumination invariance of regional texture gradient, and preliminarily determines the shadow pixels in the foreground of candidate motion by using the characteristics of brightness variation and the invariance of color ratio. Then the texture gradient invariance is used to remove errors in candidate shadow regions. The combination of the two methods makes up for the poor performance of single feature or single type of shadow detection, and improves the shadow detection rate and shadow resolution. The ability to distinguish shadows from prospects accurately.
【作者单位】：安徽农业大学工学院;南京大学电子科学与工程学院;
【基金】：国家自然科学基金(No.61502226) 安徽省自然科学基金(No.1708085QF148) 安徽农业大学稳定与引进人才项目(No.yj2017-38) 校青年基金项目(No.2016ZR008)
【分类号】：TN948.6;TP391.41

【相似文献】