基于Adaboost算法的车辆轮胎检测研究与实现

Research and Realization on Vehicle Tire Detecting Based on Adaboost Algorithm

为了更有效地对路段车流的车型进行分类统计,可通过测量车辆轴距的方法实现归类。提出了一种基于Adaboost算法,利用OpenCV视觉库对摄像机采集的车辆图片进行轮胎识别的检测方法。其基本思想是建立轮胎样本和级联分类器,利用被训练成的强分类器对摄像机采集的视频截图进行目标车辆轮胎的检测,然后通过检测结果计算出车辆两个轮胎之间的距离参数,从而求出轴距以确定其车型。通过分析检测出来的轮胎图案,发现存在较高的漏检率和错检率。最后,通过调整样本结构,发现大幅提高了检测准确率。

In order to undertake classification statistic to the traffic stream of space interval efficiently,it can be realized by measuring the wheelbase. A detection method of the tire identification based on Adaboost algorithm was proposed, using the OpenCV visual library on the vehicle images taken by the camera. The basic idea is to establish the vehicle tire samples and the cascade classifier,then to detect the vehicle tire from the video screenshots which taken by the camera with the strong classifier after training. Calculating the distance parame- ter by the detection results and obtaining the wheelbase to determine its models are the final steps. By analyzing the fire patterns detected, it is found that there is a high missing rate and false rate. Finally, the detection accuracy rating can be improved significantly by adjusting the sample structure.