Video Vehicle Detection Device Based On DM642

The vehicle is an important part of the urban traffic,so there is a lot of meanings for urban intellient traffic system when researching for video vehicle device.Now intelligent traffic system urgently need the product which are such of smaller device,lower power consumption and less cost.Therefore,it is becoming more and more essential for developing video vehicle detection in practice.The article is based on a digital signal processor (DSP) TMS320DM642platform produced by TI company. Reasearched embedded vehicle detect algorithm and implemented video video detect device. The main research contents and results are as follows:1.Algorithm design:Based on virtual loops algorithm, bring speed,simple variance and average of the virtual loops,the article proposes a fast video vehicle detection algorithm based on virtual loop’s statistics. Firstly,calculate the speed of virtual loops according to block matching theory.Secondly.Higher-order statistics block method are used to established the background model in order to get the average and variance statistic of foregroud.At last,logistic operations are used to tell the statet of the vehicle.The algorithm is not only satisfy the need of real-time but also have a good resist against disturbance in practice.lt is tested to be fit for embedd system.and t combine the algorithm which run in the PC can achieved the rate of50frames per second,98percents accuracy in the ideal condition,93percents accuracy in the night, and more than94percents in the shadow and crowd situation.2Algorithm realization and optimization:the algorithm proposed by this article can be operate in DM642to detect vehicles.By the use of inline functions, loops unrolling.rearrangement of software pipelining, using a ping-pong mechanism to transmit data and so on to optimize the algorithm.the actual test results shows that the algorithm can achieve the20frames per second and reach the accuracy of96percents in the day and90percents in the night, and detection error rate in not higher than2percents.probability of miss is4percents in day. And detection error rate in not higher than4percents,probability of miss is10percents in night.Experimental results show that the designed video vehicle detection device based on DM642platform can real-time detect in the premise of accuracy.It reflects the excellet balance between real-time ability and accuracy.And lay a certain foundation for further research and application of the embedded video vehicle detect device.