面向船舶监控的目标基高效视频编码关键技术

Due to increasing video resolution, intelligent shipping surveillance system becomes the emerging and urgent demands of the shipping transportation safety and the precaution of major disasters. The object detection under no fixed scene as well as real-time transmission of the high definition video are the challenge of shipping surveillance system, and the moving object detection as well as high efficiency coding are the two key techniques of the video surveillance system.. This work takes sufficient consideration of the salient object detection efficiency and the vibrate picture of the moving camera, firstly, the video salient object detection model based on the super-pixel is proposed; secondly, based on the perceptual coding, the salient map is used to optimize the coding efficiency and coding complexity; At last, the present algorithm performance is improved by parallel processing.. Based on the new generation of video coding standard HEVC to establish the encoder with high coding efficiency, low coding complexity and high parallelism. Different from traditional encoders, the coding function of the proposed system is more attractive based on the content, meeting the demands of intelligent shipping surveillance system to precaution the abnormal phenomenon and the real-time transmission of the high resolution video.

面向船舶的高清晰度智能监控系统已成为航运安全和海上重大灾难事故防范的迫切需求，摄像机非固定场景下的目标检测和视频图像高质量实时传输是船舶监控系统面临的重要挑战，视频运动目标检测和高效编码是智能监控应用的两大关键技术。. 本课题充分考虑摄像机非固定场景下的目标检测效率和图像抖动因素，首先，建立基于超像素的视频显著性目标计算模型；其次，基于感知编码理论将图像的显著图和视频编码相结合优化编码效率和计算量；最后，对编码框架进行并行化设计，提升算法整体性能。. 在新一代视频编码标准 HEVC 基础上构建高编码效率、低计算量、高并行度的编码器，不同于传统的编码器，该系统创新地引入了基于内容的编码功能，能够满足智能船舶监控系统对异常行为预警，以及高分辨率图像高质量实时传输的要求。

随着多媒体技术发展以及智能海事监控服务的普及，如何通过新的编码技术进一步提高视频编码效率、降低计算复杂度受到了极大的关注。本项目的目标是：(1) 建立高效的视频显著性计算模型，能快速、准确地检测出运动目标，为研制支持异常行为和灾难预警的海事智能监控系统提供科学依据; (2) 构建高编码效率、低计算量、高速并行的编码器, 编码效率平均提高20%-30%, 为高分辨率视频的高质量实时传输奠定理论基础。在项目中，主要研究内容包括：(1) 视频显著性目标检测算法研究; (2) 编码效率和计算量优化算法研究; (3) 整体算法高速并行实现。本项目取得的主要成果包括：(1) 针对视觉显著性建模分析、高效码率优化、车联网视频应用优化等关键技术问题，在理论、算法等方面取得一定成绩，代表性工作获得国际同行的认可; (2) 在相关领域国际期刊、会议上发表论文20篇，其中以第一作者发表SCI检索论文8篇， 包括计算机领顶级期刊IEEE Internet of Things Journal、IEEE Communications Surveys and Tutorials、IEEE Transactions on Intelligent Transportation Systems等; (3) 项目共培养博士研究生2名，硕士研究生5名, 另外还有2名本科生参与了本课题的研究。