基于图像的车道线内车辆目标识别

Vehicle target recognition in lane line based on image

为了获取本车道内前方车辆信息,文章提出一种基于HOG特征与SVM模型的车辆识别方法。根据检测出的车道线确定搜索区域,结合车底阴影特征实施车辆的检测并确定车辆可能存在的假想区域,针对假想区域进行HOG特征提取,构建车辆正、负样本库,将假想区域进行HOG特征输入到训练好的SVM识别器,实现车辆目标的识别。通过大量测试图像进行验证,结果表明文章采用的方法可识别出本车道内前方车辆目标。

In order to obtain the vehicle information in the lane,a vehicle recognition method based on HOG feature and SVM model is proposed.According to detect the lane search area,combined with the feature of car shadow implementation vehicles detection and determine possible imaginary area,in view of the hypothetical areas to HOG feature extraction,the positive and negative samples library build vehicles,imaginary area will HOG feature input into the trained SVM recognizer,achieve vehicle identification.Through the verification of a large number of test images,the results show that the method in this paper can identify the vehicle target in front of the lane.