Real-time visual vehicle detection method based on DSP platform
1. School of Automotive and Traffic Engineering, Jiangsu University, Zhenjiang, Jiangsu 212013, China; 2. Institute of Automotive Engineering, Jiangsu University, Zhenjiang, Jiangsu 212013, China
Abstract：To improve the real-time performance of vehicle detection algorithm for the present embedded platform, a vehicle detection algorithm was proposed based on multi-range region of interest（ROI）according to the algorithm computational costs. The traditional Adaboost algorithm was improved by traversing multiple layers of interested regions in multi-scale window to reduce the traversal times of entire image and single ROI. The visual studio （VS）software was used to simulate and compare the results before and after the algorithm optimization. The static allocation memory was used to instead the dynamic allocation memory in hardware transplantation, and the embedded transplantation of algorithm was completed by direct memory access （DMA）insteading memory library function. The results show that the proposed algorithm has good detection accuracy and real-time performance. The method can effectively improve detection speed with good vehicle detection accuracy. The processing speed reaches 13-15 frames per second at the input image size of 720 × 480 pixels, and the real-time vehicle detection is achieved.
2019, Vol. 40 
