The Research Of Traffic Flow Detection Algorithm Based On Machine Vision

The study of Traffic Flow on-line detection is based on machine vision, which is great significant to improve development of our country Intelligent Transportation .At present, the detection of traffic flow method exists in the process of detection of low efficiency, high rate of false detections, slow speed and so on doesn't meet requirements of real-time detection. Based on this, the algorithm study of traffic flow detection based on machine vision is focused on automotive image processing and auto image feature extraction, auto image target detection, traffic statistics and algorithms to determine the traffic status.In this paper, first of all, the vehicle's images that are collected will be changed for 24-bit GDI bitmap by A/D, the vehicle's images converted are processed by smoothing, segmentation, gray, sharpening and so on. Secondly, after the vehicle image extraction processing size of the area, perimeter and centroid of the region and other parameters, based on the extracted characteristic parameters to achieve image target detection, and statistical vehicular traffic, conducting a traffic state judge according to certain time and a certain section of the average flow. Auto image target detection based on image feature of Automotive extraction, using differential-based method to detect car image target, and calculate the traffic flow, its statistical accuracy rate of 98%.Traffic condition judgement use BP neural network of optimal genetic algorithm as a judging device, taking traffic flow parameter which is presented as input parameter of BP neural network, taking two traffic state that is traffick blocking and flowing as output of BP neural network, training BP neural network which optimized by GA, and taking it as a traffic judging device, completing traffick flow's detecting, the correcting rate of judgement is more than 96%. The final design is traffic flow inspection system which based on machine vision, and through pilot testing to achieve the traffic flow and traffic state judge.As a result, the Traffic Flow Detection system is designed in this paper, and the experiments verify that this system meets requirements of real-time and the development of Intelligent Transportation ,at the same time, feasibility and correctness of this paper's the theory study and design.