Video-based Vehicle Detection In Intelligent Transportation System

Today, the traffic has become one of the important factors that affect people’s life. With thedevelopment of video surveillance equipment, computer image processing and computervision technology, vision-based intelligent traffic monitoring system has become an importantpart of modern intelligent transportation research field. As traffic is the most important part,how to quickly detect the vehicle in the video surveillance has become a core technology inintelligent traffic monitoring system.Although there have been many researchers focus on how to detect vehicles in visionapproach, and lots of academic research achievements have been proposed, there are stillmany challenges in application. There are three major reasons as follows:First, video data acquisition devices are usually installed outdoor, so the outdoor light willdirectly affect the quality of imaging. Shadow is easy to be extracted as a part of objects.Shadow is easily extracted in moving object detection stage, as part of the object itself, whichaffects the subsequent feature extraction results.Second, although there are lots of moving object detection methods, but many are stillsubject to the requirement of time and space. So the moving object detection methods whichcould be applied in industry projects are still limited.Third, simple and fast vehicle feature extraction methods are still desperately needed inengineering application.In view of the above reasons, this paper studies some critical technologies of video-basedvehicle detection in intelligent traffic monitoring system, including moving target detectiontechnology, background modeling techniques, shadow detection and removal techniques,and vehicle feature extraction and detection technology. Through research andimplementation of these technologies together to build a vehicle detection system based onvision. The system implementing is through a modular way, and it is with today’s mostpopular machine vision library OpenCV, thus it is not only easy to maintain and upgrade thesystem, but also has a high performance optimization. After testing on surveillance video, thesystem can detect the video vehicle well. The frame detection rate is87.21%, and the vehicledetection rate is89.23%, which has a certain industry prospect.