一种基于Gabor小波和颜色参数模型的阴影检测算法

本文提出了一种基于Gabor小波纹理特征分析和阴影参数模型颜色分析的阴影检测算法.首先,建立背景的混合高斯分布模型,通过对特定帧的颜色分析建立阴影参数模型.对图像序列中的每一帧,首先通过差分法提取出前景区域,然后通过Gabor小波滤波的方法,对背景和前景图中的相同小区域进行纹理特征分析,比较特征向量的距离,以判断潜在的阴影点.然后,通过阴影参数模型对这些潜在的阴影点进行颜色分析,并通过连通区域标记,形态学运算,FPR(去除假阴影)分析等方法,找出真正的阴影区域.实验结果表明,本算法具有较高的阴影检测率和物体检测率,可以满足实时检测的需要.

A Shadow Detection and Suppression Algorithm Based on Gabor Wavelet and Color Parameter Model

In this paper, a new shadow detection algorithm based on Gabor wavelet texture analysis and shadow parameter model color analysis is proposed. First, the background’s Gaussian mixture distribution model is created, and the shadow parameter model is created based on color analysis on pixels in a special frame. For each frame, foreground figures are extracted by some background subtraction method. The texture coefficient vectors of small areas in background and foreground figures are created by Gabor wavelet method, and the distances between those vectors are used to judge whether foreground pixels are potential shadow pixels. Those potential shadow pixels are analyzed by the shadow parameter model to search real shadow pixels. The labeling connected component method, morphology operation and false positive reduction analysis is also used in this algorithm. Experiment results show that the proposed algorithm has a high shadow detection rate and object detection rate and it can be used in real time object detection.