A Red-Rush Vehicles Record System On Video Sensor

Intelligent Traffic System ( ITS) apply new technology of Computer,Telecom, Automation to realize traffic management automation.Compared with traditional vehicles detector, the video sensor has lots of advantages such as fast response, easy installation and maintenance, the ability to monitor wide areas and obtain more kinds of traffic parameters, and as a result, it has been used in ITS recently. Up to now, a number of video processing and analyzing methods basing on video sensor have been proposed for vehicle detection and tracking.In this paper, vehicles are detected by background subtraction. Background picture can be auto-updated by iterativing the previous background picture and the current. The system detect the detecting-area difference of the scene in successive frames , if the difference beyond the threshold , system make sure a car is running in the detecting-area. If the traffic light is red , the system create a picture of the car and upload to the server .In this system , scene of the crossroad is divided into detecting-area and non-detecting-area, the area of detecting-area is only 1/6 of the whole scene, this can greatly reduce computation of system. System compute the difference by luminance and hue of pixels.The system divide the detected vehicles area D_i into blocks of16*16 pixels, and get the feature of the area such as∑F_i~I(x,y),∑F_i~H(x,y),(x,y)∈D_i. Block numbers smaller than 4 is excluded. If thesame car detected in the first and second detecting-area in succession when the traffic light is red, It is identified as illegal. This is tested by NCCF (Normalised Cross-Correlation Function) .