CAJ | 학술논문

为了提高基于视频的车辆检测技术在应用中的实时性和准确率,提出了一种应用单目视觉进行车辆检测的方法;首先,提取车道边缘,由车道边缘得到道路区域,根据经验知识在车道区域内确定感兴趣区域,减少车辆检测算法搜索范围;接着基于车辆的对称性特征,阴影和边缘特征对兴趣区域进行过滤,进一步缩小感兴趣区域;最后用离线训练好的AdaBoost分类器对过滤后的图像进行分类识别,检测出动态的车辆;实验结果表明,利用该算法能满足实时性和准确性的要求.
위료제고기우시빈적차량검측기술재응용중적실시성화준학솔,제출료일충응용단목시각진행차량검측적방법;수선,제취차도변연,유차도변연득도도로구역,근거경험지식재차도구역내학정감흥취구역,감소차량검측산법수색범위;접착기우차량적대칭성특정,음영화변연특정대흥취구역진행과려,진일보축소감흥취구역;최후용리선훈련호적AdaBoost분류기대과려후적도상진행분류식별,검측출동태적차량;실험결과표명,이용해산법능만족실시성화준학성적요구.
In order to improve the real-time and accuracy of the vehicle detection technology in the application, it shows a vehicle detec-tion method based on monocular vision. First of all, extract the edge of lane, obtain the driveway area from the edge of the road, confirm the region of interest with the experiences in the driveway area, it can reduce the scope for searching. Then filter the region of interest based on the symmetry, the shadow and the edge of the car and narrow the region of interest even more. Finally, classify the filtered images with the AdaBoost classifier which is trained off-line, detect those motorial vehicles. The experimental results show that the algorithm can adapt to the real-time and accuracy requirements.