The Design And Implementation Of The Intelligent Unattended Video Surveillance Platform

With the rapid development of the socio-economic, accumulation of people’s wealth,the protection of right and interest received more and more attention. From the coal minemonitoring to traffic surveillance, from national defense and security checks to communitysafety maintenance, video surveillance has been extended to every corner of people’s lives.Driven by the development of network and computer, ordinary video surveillance system ismoving in the direction of the network, intelligent and high-definition.In the realization of the hardware platform, this paper selected the TI’s DaVinci DSPTMS320DM6437 which has powerful data processing capability and the VPSS subsystemthat particularly suitable for processing video information. The chosen of video standardfor video compression we selected the newest H.264/MPEG-4 Part-10.H.264 contains a number of new features that allow it to compress video much moreeffectively than older standards and to provide more flexibility for application to a widevariety of network environments.First of all, through the analysis of complexity of H.264 encoder, we determine thecorresponding optimization. In the intra prediction, we propose a fast intra-frame predictionalgorithm. More specifically, the macro block based on MAD values can pre-determine theselection mode. If the intra 4×4 prediction mode is chosen, it considered of the correlationbetween neighboring blocks and prediction direction to reduce the number of predictiondirection.In H.264’s block-matching motion estimation, a new fast block matching algorithm based on the asymmetric diamond search pattern is proposed in this paper. In using the twoasymmetrical search patterns, this fast algorithm achieves 31.38% speed-up with only0.008dB of degradation in the quality, especially suitable for the video sequences havedynamic information.Data transfer is another optimal direction because of the 32% consumed in the encodingcycles. From the experimental results we can see that, though re-configuration EDMA,proposed re-allocation Cache and optimal the framework includes three loops, the timeneeded for date transfer reducing more than 50%, speed of original encoder is 2-3 fps, fivetimes as quickly as the former when optimized.