基于目标检测的码率优化算法

Semantic-Based Bitrate Control Algorithm

随着智能多媒体技术和人工智能技术的融合发展,目标检测已广泛应用于移动监控网络远程传输场景。为了提升视频的目标检测精度,提出一种针对目标检测的码率优化算法,在有限的码率范围内使得物体的检测准确率达到最大化。将目标检测作为视频语义任务,分析视频码率比特与目标检测之间的关系,建立目标比特分配模型,从而使目标检测准确率达到最大化。实验结果表明,所提算法不仅能节省视频码率,还能提升目标检测精度。

With the integrated development of intelligent multimedia technology and artificial intelligence technology,target detection has been widely used in the remote transmission scenario of mobile monitoring network.In order to improve the accuracy of video object detection,this paper proposes a bit rate optimization algorithm for object detection,which maximizes the accuracy of object detection in a limited bit rate range.Taking object detection as a video semantic task,the relationship between video bit rate and object detection is analyzed,and the object bit allocation model is established to maximize the accuracy of object detection.Experimental results show that the proposed algorithm can not only save video bit rate,but also improve target detection accuracy.