【摘要】 高级驾驶辅助系统(advanced driver assistance systems,ADAS)已经变成近几年汽车领域的研究热点,交通场景下的车辆检测是ADAS系统的重要组成部分。为实现车辆检测性能和计算效率之间的平衡,针对Fast RCNN使用选择性搜索算法进行特征提取耗时较长且检测准确率较低的问题,提出了基于Faster RCNN和YOLO的一种新的网络架构。该方法非常简单,可以在具有挑战性的KITTI数据集中进行端到端的训练并表现的不错,超越了Faster RCNN的检测效果。同时,该方法也非常有效,运行速度超过每秒24帧,可以达到实时检测。更多还原

【Abstract】 Advanced driver assistance systems(ADAS) has become a research hotspot in the automotive field in recent years,and vehicle detection under traffic scenarios is an important part of the ADAS system. In order to achieve a balance between performance of vehicle detection and computational efficiency,a new network architecture based on Faster RCNN and YOLO is proposed for Fast RCNN using selective search algorithm for feature extraction taking a long time and the detection accuracy is low. This method is very simple,can be trained end-to-end and performs extremely well in the challenging KITTI dataset,outperforming the Faster RCNN. At the same time,the method is also very efficient,allowing perform at more than 24 frames per second,and real-time detection is possible.更多还原