Abstract:

The maximum range of scenes that can be captured by road traffic surveillance cameras is known. By analyzing the traffic flow in the surveillance video, a scene based compression method is proposed. Based on the distribution of motion in space, the motion energy map is established and the mask is obtained by negation. The HEVC coding method is improved：According to the characteristics of strong directivity and violent motion of the traffic flow in the motion distribution area, an asymmetric template is used to improve the TZSearch motion search algorithm; and for the characteristics of weak motion in the mask area, a motion search termination threshold is set in advance. The experimental results show that compared with AVS2 and MPEG-4 encoding, the improved method can ensure both traffic saliency and compression performance; and compared with HEVC encoding, the average bit rate saving, the average PSNR enhancement and the average time saving of the improved method are11.80%, 3.90 dB and 5.55%, respectively. The accuracy of vehicle identification can be improved by 7.41%, which enhances the video encoding efficiency and the analysis performance of the video itself.

Key words: intelligent transportation, traffic surveillance, video compression, High Efficiency Video Coding（HEVC）, motion energy map

摘要：

道路交通监控摄像机所能拍摄的最大场景范围已知，分析监控视频中交通流流动的规律，提出基于场景的压缩方法。通过运动在空间上的分布，建立运动能量图，取反获取掩模。改进HEVC编码方法，针对运动分布区域交通流的方向性强且运动剧烈特征，使用非对称模板改进TZSearch运动搜索算法；针对掩模区域运动性弱特征，提前设置运动搜索终止阈值。实验结果表明，与AVS2和MPEG-4编码相比，改进方法能够同时保证交通显著性以及压缩性能；与HEVC编码比较，改进方法的平均比特率节约、平均PSNR增效和平均时间节约分别为11.80%、3.90 dB和5.55%，车辆识别的准确率能够提高7.41%，增强了视频编码效率的同时强化了视频本身的分析性能。

关键词: 智慧交通, 交通监控, 视频压缩, HEVC标准, 运动能量图