【作者】 苑春苗；

【导师】 金志刚；

【作者基本信息】 天津大学 ， 计算机应用技术， 2012， 博士

【摘要】 网络视频监控系统广泛应用于安防领域，是协助公共安全部门打击犯罪、维持社会安定的重要手段，目前正在向高性能、低耗能和智能化发展。视频相关技术是视频监控系统的最关键部分，直接影响整个视频监控系统的使用效果。监控系统中视频处理的一般流程为监控端进行视频采集、编码；网络传输；接收端解码和视频分析。流程中的每个部分都将直接影响系统的最终性能，因此，文中对流程中涉及到的前端监控设备的硬件设计和码率控制技术、传输时的控制技术和接收端的智能化视频分析技术进行了研究和改进。给出了一种低耗能的支持无线移动监控的网络监控设备设计方案和一种基于H.264技术的网络监控设备设计方案，两种方案均采用嵌入式设计，具有高性能、低耗能和低成本等优点。提出一种基于复杂度变化改进的H.264速率控制机制。针对视频监控系统希望在场景变化时尽可能分配更多比特的需求，提出了一种改进的H.264码率控制算法，该算法利用图像的复杂度变化信息，使得监控端可以根据监控场景里的变化信息调节码率。通过实验结果可以看出，在场景变化时，改进算法的PSNR较标准算法提高，说明在场景变化时可以获得更好的视觉效果。提出了一种基于自适应帧率策略的QoS控制机制。在实时视频数据传输时采用一种新的基于帧率自适应的策略，在该策略中，首先需要进行网络带宽的动态探测和评估，然后采用合适的调整策略来动态调整码流，从实验结果可知，该机制使平均传输数据量下降了15%。提出了改进的α-blending算法。α-blending算法使用灰度函数进行背景建模，解决了当一个单一的检测失败或光线剧烈的变化会导致前景检测失败或出现鬼影，甚至扩大到整个检测区域的问题，最终得到较好的目标检测效果，改进算法中，针对该算法中背景建模方法的背景更新选择进行优化，经实验测试可以使算法运算时间平均降低5%，尤其是对大分辨率视频文件进行分析时，改进效果十分明显。上述硬件设计方案和改进算法目前已经被应到体育赛事监控等实际项目中，并取得了很好的应用效果。更多还原

【Abstract】 Network video surveillance system is one of the most important part of safe andsecurity industry which assists the public security forces to fight against crime andmaintain social stability and now is developing toward high performance, lowconsumption and intelligent video analysis. Video technologies directly affect therunning effects of the whole video surveillance system. Video process of Videosurveillance system include monitor devices’ video data collection, networktransmission and end user’s decoding and video analysis. Each part of the Videoprocess will impact the system’s performance. Therefore, in the paper, the videorelated technologies are studied and improved including hardware design，rate controlalgorithm, video transmission control technology and intelligent video analysis.A design scheme of wireless mobile monitor network video camera (NVS) and adesign scheme of H.264based NVS are given. Two schemes all are implemented byembedded design and have advantages of high performance, low energy consumptionand low cost.An improved H.264rate control based on the image complexity is proposed. InJVT, all frames are encoded in the equal way, but in video serveillance system, theinformation of scene change need to be paid more attention. More bits should beallocated when there’s t strong picture change fluctuation, and less bits should beallocated in the relatively still scenes will have a better encoding perceptual effect.The changes of image are described by complexity, so a modified H.264rate controlalgorithm is proposed which has obtained good results.An adaptive frame rate control strategy(AFRS) QoS mechanism is put forwarded.During real-time video data transmission, a new method AFRS is used. In AFRS,network bandwidth need to be detected and evaluated dynamically firstly, then datastream will be adjusted according to above evaluation result. AFRS can usereasonably network bandwidth and reduce data conflictions on the network, so thatthe monitor has a better video quality in the video serveillance system. A doublecaching mechanism(DACM) and AFRS can insure constant video quality for the enduser when wireless network sector switches or bandwidth changes greatly.An improved fast alpha blending algorithm is given. Alpha blending algorithm solves the problem that when moving object detection algorithm running, a single testfailure or strong illumination changes will lead to the failure of foreground detectionor appearance of ghost, and even extends to the whole detection region. By optimizingthe method of background modeling, an improved method is put forwarded whichreducing the running time of alpha blending algorithm. Experiment shows improvedmethod is useful especially for videos with larger resolution.The hardware design and the improved algorithms above mentioned have now beapplied to the practical projects such as Sports Event Monitoring Projects and have agood running effect.更多还原