融合颜色和CS-LBP纹理的运动阴影检测

针对现存阴影检测方法存在的实时性和精确性兼顾不周的问题, 提出加权融合颜色和纹理特征的阴影检测方法: 首先利用HSV颜色信息提取疑似阴影点; 其次, 通过阴影的亮度比计算阴影亮度隶属度, 对于高亮度隶属度的疑似阴影点, 直接判定为阴影点, 减少了纹理检测的计算量; 然后对低亮度隶属度的疑似阴影点提取高效的CS-LBP纹理, 并进行纹理匹配, 根据纹理的相似程度及阴影空间分布特点, 计算出纹理隶属度; 最后, 根据实际中纹理随亮度变化的特点, 提出了依据亮度比自适应调整纹理隶属度权重的特征融合方法, 进行有效的阴影检测. 实验表明, 本文方法实时性良好, 可去除自阴影, 分割精度较佳, 隶属度方法的使用, 使本方法对光照变化及噪声更具有鲁棒性.

Moving Target Shadow Detection Based on Color and CS-LBP Texture

Considering the contradiction of real-time and accuracy in existing shadow detection method, this paper presents a new shadow detection method, the method combines color feature and texture feature by weight fusion. Firstly, we use HSV color information to extract the suspected shadow points. Secondly, we calculate the shadow brightness membership according to the brightness ratio, then judge the points with high brightness membership as real shadow points, so we can reduce the calculation of texture detection. For those suspected shadow points with low brightness membership, we extract the CS-LBP texture of these points, because CS-LBP is highly efficient. Via matching texture, we calculate the texture membership according to the level of similarity of textures and distribution of shadow. At last, considering the fact that texture change with the brightness, we put forward the method of feature texture membership by weight fusion, and this weight self-adapts to the brightness ratio. Experiment results show that, the proposed method has a good real-time performance, it can remove the self-shadows, and performs better accuracy in segmentation.With using membership, the proposed method is more robust to noise and illumination changes.