摘要
在基于视频图像的实时车型识别系统中 ,车窗的大小和位置为车型识别提供重要的特征信息。当汽车为低亮度、低饱和度的黑色系列车和深颜色车时 ,它们的颜色特征直方图常常是单峰的 ,传统的分割方法难以正确定位车窗。分析了HSV颜色空间在低饱和度时的色度奇异性 ,结合纹理检测技术 ,提出了基于色差均值的快速车窗定位算法。实验结果表明 ,该算法定位准确 ,速度快 。
In the application field of Vehicle Recognition System(VRS)based on video image,the size and location of window is an important feature for vehicle recognition.Traditional image segmentation methods can hardly localize window when the vehicle is with lower intensity and lower saturation color such as the black and the dark green car because of the feature histogram usually has one peak only.In this paper,we analyzed the oddity of HSV color space in case of low saturation.A fast algorithm for localization vehicles window by way of the mean of the neighbor pixels chromatism is proposed.The results of experiment show our localization method is fast and exact,which can be used in the real time Vehicle Recognition System field.
出处
《计算机应用与软件》
CSCD
北大核心
2004年第1期78-79,117,共3页
Computer Applications and Software
