Vehicle Detection System Design And Implementation Based On OpenCV

Intelligent Transportation System (ITS) is a commonly-used application system inthe area of modern traffic management and control. At the bottom of ITS is theVideoVehicle Detection System, which acquisitions, processes and manages the originalimage data and therefore greatly affects the accuracy and realtimeness of a series ofdetection process of the upper subsystems such as Vehicle Target Recognition, VehicleTarget Tracking, Vehicle License Plate Localization, Vehicle License Plate Recognitionand so on. However, the core detection algorithms of two technically mature VideoVehicle Detection Systems each have different major flaws.(1) The road vehicle detection algorithm based on frame difference. Thought fastand simple, this algorithm requires a great amount of experiment to determine a propertrigger threshold and often produces a lot of spurious triggering if a high-level of targettrigger rate is to be ensured.(2) The road vehicle detection algorithm based on background subtraction. As fastand simple as the frame-difference algorithm, this detection algorithm cannot solve theproblem of local background mutation and it also has a relatively high-level of spurioustriggering rate.Targeted at the above-mentioned existing problems in the Video Vehicle DetectionSystem, a research team from the Traffic and Municipal Engineering Department ofSichuan College of Architectural Technology proposes a research subject InformationCharacteristics and Data Mining of Intelligent Transportation System. As one of theparticipants, the author has mainly done the following work:(1) This paper presents an algorithm to extract sub-window coordinates and size ofthe vehicle from videos making use of the morphologically symmetrical characteristicof the vehicle target’s external contour. This algorithm detects edge by adopting properedge detection operators and then calculates and keeps track of the highly symmetricalregion of the edge images so as to detect the section of the original image in which avehicle can possibly be spotted. It is proved by experiments that this algorithm canrapidly detect vehicle targets in an image. (2) This paper completes the requirement analysis of the information managementsubsystem in the Video Vehicle Detection System with a view to both the users’requirement and the system function, on the basis of which an UML Analysis is carriedout and a model is built. Finally, a road traffic information system is created withfunctions such as logging-in of users, playing back of original videos, detecting dataupload, management of system files and etc.(3) On the basis of the vehicle detection algorithm presented in this paper, a newtype of Video Vehicle Detection System is built under a VC2005platform usingOpenCV1.0Library and SQL Server2005Database.