基于颜色和纹理的运动车辆阴影检测算法

Shadow Detection Algorithm for Moving Vehicles Based on Color Space and Texture

根据运动车辆的阴影特性,对基于RGB颜色空间和基于纹理特征的阴影检测算法分别进行改进,提出一种综合颜色和纹理的运动车辆阴影检测算法.在该算法中,利用阴影覆盖区域像素点亮度值比背景亮度值低、颜色和纹理保持不变的特性,通过计算各像素点颜色畸变、亮度畸变和一阶梯度值,并与背景各像素点进行比较,进而区分阴影点和非阴影点.采用形态学方法对去除阴影后的前景进行目标重建,进一步提高阴影检测的效果.在视频图像的车道近端设置ROI区域,提高车辆检测的效率.实验结果表明,提出的方法能较好地对运动车辆的阴影进行检测.

According to the shadow characteristics of moving vehicles, the two shadow detection algorithms respectively based on the RGB color space and texture are improved, and a shadow detection method integrat- ing color and texture is presented. In such an algorithm, the fact is used that the brightness of pixels in shad- ow region is lower than that in background, while the colors and textures remain the same. By computing and comparing the color distortion, brightness distortion and first-order gradient, the shadow pixels are recog- nized, and then the shadow pixels and non-shadow pixels are distinguished. Additionally, the effect of shadow detection can he further improved, which the foreground objective after removing the shadow is reconstructed by using the morphological method. Finally, the efficiency of vehicle detection can be increased by setting the ROI area near the driveway in video. The experimental results show that the proposed approach has good per- formance in shadow detection for moving vehicles.