The improved algorithm of traffic flow detection based on image processing
At present lots of approaches to traffic flow measurement based on image processing have been brought forwards, but these approaches are complex, poor real-time and low accuracy. In the view of this question, this paper presents an algorithm of vehicle flow measuring for the purpose of improving the accuracy of detection. At first, the background model is formed using Median method, and updating the model based on Kalman filter to enhance its robustness. Then, the moving vehicle is got from the difference between current frame and the background that is formed by the former step, and then the OTSU method has been chosen for the threshold segmentation to get the binary image of the moving vehicle. To avoid the inaccurate counts caused by vehicle shadows the shadow detection method based on HIS color space has been used to suppress the vehicle shadow, meanwhile, improved conversion formula between RGB and HIS is used to overcome the disadvantage of time consumption on computer. At last, acquiring vehicles count is based on virtual loop instead of virtual line detection technology. According to processing the actual traffic video, the experimental result indicates that the approach presented in this paper can meet real-time requirement, and the accuracy rises from 82.2％ to 97.8％ by using properly improved algorithm.
Intelligent Transportation System Traffic Flow Moving Object Detecting HSI Shadow Detecting
Li Tiankun Chen Wanzhong Lin Yong
College of Communication Engineering, Jilin University of China, Changchun, Jilin, China, 130025 Jilin Electric Power Company Limited, Changchun, Jilin, China, 130000
国际会议
第八届国际测试技术研讨会(8th International Symposium on Test and Measurement)
重庆
英文
762-765
2009-08-01（万方平台首次上网日期，不代表论文的发表时间）