Research On Road Front Vehicle Recognition Algorithm Based On DSP

With the development of economy, vehicle has become a necessity in people’sdaily life. Along with the continuous increase in the number of cars, traffic accidentshappen frequently. Therefore, how to effectively reduce the number of trafficaccidents has become a serious problem. Due to the development of science andtechnology and people’s increasing emphasis on transportation issues, research onvehicle safety technology emerge as the times require, and it reflects the trafficengineering, information engineering, communication technology, control engineering,computer technology and many other fields of science.As one of the research on vehicle safety technology, front vehicle detectionbased on vision has become the hot spot. By processing the information gathered fromthe sensor on the car to judge whether there is a vehicle in front of the current vehicle,which provide more security to the driver, effectively reduce the rear-end collisionaccidents. With the continuous development of DSP technology, the idea of puttingdriver driving safety assistance systems on the vehicle becomes possible. In this paper,the road front vehicle recognition algorithms is researched based on DSP, and hascarried on the following work.(1)Collection and display of the road front vehicle image. In this paper, The CCDcamera installed above the vehicle collects the road front vehicle image. The collectedvideo image is decoded by decoder chip TVP5150, and then the decoded data isstored into SDRAM through EDMA, in order that DM642uses the pingpong cachingtechnology to complete the data reading and subsequent processing.(2)Design of real-time operating system program based on DSP/BIOS.DSP/BIOS is a simple embedded real-time operating system integrated in in theintegration development system, and it greatly facilitates the users to write multi taskapplication. In this paper, the design of real time operating system program based onDSP/BIOS is realized.(3)Research on the road front vehicle recognition algorithm. After the road imagecollection, the first step of the algorithm in this paper is the road image preprocessing, which includes gray scale and edge detection, and then using improved Houghtransform to extract the lane from the image after edge detection. By determining thecarriageway and removal of other interference background, and then the region ofinterest of road front vehicle is extracted. After that the vehicles in the region ofinterest are marked by using the contour characteristic of vehicles, the information ofvehicles in the region of interest is extracted and the candidate regions are marked.Using the HOG features extraction that has achieved good results in the pedestriandetection to extract the vehicle characteristics and SVM classifier to classify them,which can recognize the vehicle and mark it.(4)Road front vehicle recognition algorithm based on DSP. The hardware andsoftware of this system are debugged，the internal resources of the DSP are in full use,compiler configuration is optimized, and part of the road front vehicle identificationcodes are optimized, which improve the program utilization and achieve real-timeperformance. Experiments have proved that the road front vehicle recognitionalgorithm is reliable.