Implementation Of Bus Pedestrian Flow Calculation System Based On DSP

The bus pedestrian calculation system uses camera equipment and high performance digital image processor (DSP) in the bus to determine the passengers whether they get on or get off, and records the data in real time. This paper analyzes video image detection algorithm, constructs hardware system based on DM642 DSP processor, designs a real-time detection algorithm to count the passengers who get on and get off, and gives the algorithm analysis and test results.Firstly, the relevant theory of pedestrian flow calculation is studied in the paper. From the five aspects, video image preprocessing, moving target detection, segmentation, recognition and tracking, the paper summarizes advantages and disadvantages, selects the appropriate algorithm. For the video image preprocessing, the color image gray processing, image denoise and grey image binary method are introduced, the mean filtering, median filtering, averaging multiple images, wavelet thresholding and spatial low-pass filtering method in image denoise are described, and the 3x3 mean filter method is selected. For the moving target detection, the optical flow method, adjacent frame difference method, edge detection method and background subtraction method are introduced, and the background subtraction method is selected. For the moving segmentation, the method based on region, pixel clustering and watershed transform are detailed recommended, and the method based on region is selected. For the moving target recognition, the ellipse fitting method and Hough transform method used for target recognition are described, and the Hough transform method is selected. For the moving target tracking, the Unscented Kalman filter and the Mean Shift using two order information are introduced. Because of these two algorithms are too complex, taking into account the special bus scenarios, a simple target tracking method using the position to detect a passenger’s behavior is designed.Secondly, the hardware platform is build. The image processing chip is DM642, the image acquisition uses the Sony CCD camera and TVP5150 decoder chip, the display uses a 7 inches LCD screen and SAA7121H encoder chip, and corresponding circuit and program codes are given. Then, based on the front algorithm research, hardware performance and the requirements, a suitable pedestrian flow calculation algorithm is designed, by using the color image gray processing method and image denoise to preprocess the video image, using background subtraction and image binary for target detection and segmentation, using ellipse fitting methods to identify the head, and using the position to judge the behavior of passengers. Finally, the paper designs a set of real-time detection algorithm, and gives its analysis and verification. The system can effectively detect and determine the behavior of passengers.