Video-based Public Transport Detection Technology Research

The traditional public transport detection technology includes loop detector, GPS detect technology, RFID detect technology and image processing technology based on video. The loop detecting method will caused some affects on traffic flow in installation, GPS detector has some strict restrict on signal strength, RFID is the commonly used technology in bus detection, but the installation of signal receivers is limited for some particular affects, so it restrict its application scenarios. The public transport detecting technology is novelly developed in these years, its calculation speed and application environments as well as detection accuracy need to be enhanced.At present, the traditional methods of vehicle detection are not suitable for public transportation detections required in bus priority. A bus detection method is provided in this paper according to high-definition videos of1296*968. In this method, the process of detection is consist of three bus identification processes, first computing the pixel area of moving objects in ROI(Region of Interest) regions, then the width of vehicles in ROI are counted, finally the vehicle windows in ROI are indentified, if all the three criterions are fitted, the vehicles in ROI is a bus. The traffic videos on Chang An Avenue is used to identify the different kinds of vehicles. The main work of this paper includes:(1)Analysis the traditional vehicle detection technologies, providing this bus detection methods based on Computing pixel area, body width and its windows. Analysis the ROI image and count the bus numbers and other vehicle numbers in the video.(2)Extracting background using Single Gaussion background model methods and detecting moving objects by background subtraction. Filling the empty areas in binary images and computing the body width and body windows for bus identification.(3)Using LSD and a contrario approach (Line Segment Detection) for body width detection, and providing a line continuity identification criterion to guarantee the accuracy of the test results.The test results on Chang An Avenue bus lane traffic video shows that this algorithm has a good real-time properties, the detection error is small and the stability is good.