Design Of Traffic Flow Video Detection System Based On DSP

As the holdings of vehicle and the covers of parking lots increase, the conventional parking space prompt method does not meet the peoples’ increasing need of high quality traveling. This paper focuses on traffic flow detection system based on DSP. The video of traffic flow is measured by DSP platform in every parking area in the underground parking lot, after which the counting signal is transmitted to external LED display system. The introduction of DSP ensured the miniaturization of the system as well as the improved video detection algorithm ensured the real time capability and accuracy.The hardware platform of system control and detection algorithm was based on TI TMS320DM642 digital signal processor, involving 2 CCD cameras, 2 video encoder chip, 2video decoder chip, LED display, and video interface on chip to complete the video capture/display module, performing the vehicle video capturing and assistance device. The external storage module consisted of SDRAM and FLASH chips, realizing the storage of video data and algorithm. The GPIO ports on chip were utilized to output the counting signal.The driver software of hardware module and decoder/encoder chips were designed in CCS3.1programming environment, using the CSL lab functions and I2 C protocol. Aimed at designated circumstance in underground parking lot, the detection algorithm was improved over the generalized algorithm. Background frame differential assistance and single frame differential results were used in extraction of vehicle features. The combination of dual virtual line detection and auxiliary background line was used in counting the vehicle.The system design was proved to be feasible in the in-field test. The complexity of improved algorithm was reduced by 2 orders while the accuracy was as high as 91.6%, which made the presented approach of the underground parking lot traffic flow video detection system a high efficiency and functional design.