In recent years the intelligent transportation system (ITS) made a spurt of progress of development. Vehicle queue length is an important parameter in intelligent traffic system, The technology of vehicle queue length detection based on video is an important research trend in intelligent transportation system as it has extensive application prospects in the realization of intelligent traffic system in the city. This paper research the key technology used in the detection of vehicle queue length. The main research work of this paper is presented as follows:(1) Video image pre-processing and jitter elimination technique. By comparing several different filtering methods treatment effect in this image, choose the median filter; through several images of different operator enhancement effect comparison,choose airspace enhancement template as image enhancement; according to the characteristics of the video image,choose image stabilization technique based on image registration using.(2) According to the different light intensity we use different treatment methods to realize the detection of vehicle queue length. In day light the edge information of vehicle is sufficient and obvious, the traditional method can get good performance. However, as the night illumination condition is complicated, the traditional detection method shows many shortcomings, such as the night light is dark the edge information is reduced, the existence of the vehicle detection robustness is greatly reduced by using the edge information of vehicles. Through analysis on the video, in the night illumination lamp is a prominent feature of the vehicle, especially the front lights, so we use the front lights to detect vehicles to realize vehicle queue detection.(3) In view of vehicle queue when driving slowly or motionless, use statistical methods the image horizontal integral projection computed from the two-value image to to implement detection of queue length.(4) In order to realize the automatic switch between the day and night vehicle queue detection algorithm, an algorithm based on the features of the difference of color component image is proposed to release the switch. |