Research On Moving Vehicle Detection And Tracking Technology Based On BWDSP1042

As the core of Intelligent Transportation System, moving vehicle detection and tracking system is becoming more and more important in vehicle monitoring, traffic management and so on. Its development will help alleviate traffic congestion, traffic accidents and other issues. That it uses machine vision technology is an important research direction in Intelligent Transportation System applications. With the rapid development of IC design and production capacity, digital signal processors are widely used in the occasion of processing large amounts of data including video and audio. It reflects the incomparable superiority in high-intensity, high-speed data processing.At the beginning, this thesis describes background and significance of this subject, research status of moving vehicle detection and tracking algorithms, different developments of DSPs at home and abroad; Then this thesis introduces BW Series DSP--BWDSP1042, development environment and instruction set in detail which developed independently by China Electronics Technology Group Corporation 38th Research Institute (CETC No.38 for short). This chip has strong parallel execution capability and is equipped with a complete set of development and debugging tools; Next, It list and analyzed pros and cons of optical flow method, frame difference and background subtraction, we decided to use the vehicle detection algorithm based on frame difference method, because this algorithm is small, easy programming compared with optical flow method and it does not require update background in real-time compared with background subtraction. Meanwhile, this thesis used a histogram matching algorithm as vehicle tracking algorithm, the shifter look-up table on BWDSP1042 can be used to achieve histogram statistics. This thesis’s main work is to design and optimize program based on BWDSP1042 and synergy with Matlab to simulate vehicle detection and tracking system. Firstly, we analyzed key processes of frame difference and histogram matching when designing, the key process of frame difference is gray-scale pixel matrix subtraction and the key process of histogram matching is counting histogram of gray scale image pixel values, then we imported gray-scale image data to test program and optimized program by inserting pipeline on the basis of ensuring correct operation results, we reached a relatively good optimization effect from the experimental data. After designing and optimizing program, Matlab programming synergies with BWDSP1042 to simulate vehicle detection and tracking system. Firstly we extracted video frames by Matlab programming and converted frames to gray-scale, then imported gray-scale data to BWDSP1042 and obtained vehicle position after frame difference, count histogram of expanded box which include vehicle and prediction boxes. Finally, histogram matching was used to track vehicle.