Traffic Flow Video Detection Technology Based On Dsp And Its Implementation

Intelligent Transportation System is the future development direction of transportation systems, Video Detection Technology is more and more important in Intelligent Transportation Systems for its unique advantage such as real-time,fast and accurate。Most of the tradition video detection is based on the personal computer platform and use general purpose processor, but the method brings a lot of inconvenience when used in traffic scene. With the rapid development of information technology, embedded systems are applied in many fields, it not only make the dedicated processor can be full play, but also greatly reduces application costs, so the embedded video vehicle detection system gradual become a hot research in intelligent transportation.In this paper, the moved vehicles of the video surveillance system is chosen as study object, First, several common background extraction algorithms are introduced, and then put forward a Background Modeling Algorithm Based on Pixel Frequency, which greatly shorten the time of background subtraction. It also introduce the moving target detection algorithm and the main method of traffic detection technology, and after compares with several frequently-used method of moving target detection algorithm, we choose the background subtraction method for moving target detection which is used in practical application,then use the track targets method which is a method of the video vehicle detection technology design and implement a centroid-based algorithm for traffic statistics, experiments show that the algorithm can shorten time and count accurately.Finally, according the above-mentioned algorithms, we make a hardware and software design in the DSP, in hardware design,in order to provide a better performance of the hardware platform, we choose the BF533 processor which is designed for audio and video applications by ADI company, and system requires some components; in software design, we choose the high performance VDK operating system platform, LWIP protocol stack for developing the system, it greatly accelerate the development of the network applications, it also design a ping-pong buffer data structure to input and access the video data, this improve efficiency of accessing data, and apply saddlebag which is designed for processing image by ADI company, the transplant of the OpenCV algorithm in DSP is greatly accelerated.