Abstract: A vehicle detection algorithm based on low contrast images is presented. The Region of Interest(ROI) is obtained by using the segmentation algorithm in the image. It performes preprocessing, extracts the edges and texture features of the ROI with the algorithm of Haar wavelet feature extraction, and uses Support Vector Machine(SVM) to detect vehicle. Experimental results demonstrate the superiority of the proposed approach, whose detection rate is 90.6% and false alarm rate is 3.8%. The performance of the algorithm can be improved by cumulate learning.

Key words: vehicle detection, feature extraction, wavelet transform, Support Vector Machine(SVM)

摘要： 提出一种基于低对比度图像的车辆检测算法。对图像分割算法得到的感兴趣区域(ROI)进行预处理，利用Haar小波特征提取算法提取ROI的图像边缘及纹理特征，利用支持向量机对ROI进行车辆检测。实验结果表明，该方法对车辆检测率达到90.6%，误报率为3.8%。通过再学习还可以进一步提高算法的识别性能。

关键词: 车辆检测, 特征提取, 小波变换, 支持向量机

CLC Number: 

• TP391.41