Theoretical Study Of Vehicle Flow Detection Based On Video Image Processing

Traffic information real-time acquisition plays an increasingly important role in the intelligent transportation systems,and vehicle detection is the main means to obtain the traffic information.Vehicle detection is an important research field of intelligent transportation systems,which can be used to detect the traffic flow,vehicle speed,the share of traffic data and so on,providing the numerical information for the decision-makers.The detection of traffic flow is key to information collecting,not only conductive to the transportation systems dispatch,but also in favor of the control strategy for single traffic signals.By analyzing the traffic flow,decision-makers can dispatch the time properly,control the traffic flow intelligently,maximize the passage rate,and reduce the waiting time of vehicles.There are many approaches for traffic flow detection.Nevertheless,in virture of flexible system settings and being able to detecting large areas,video vehicle detection technology based on image processing has become a research focus in the area of intelligent transportation systems.With the inter-frame difference method for the background establishment,this thesis proposes an improved algorithm for fast detection of moving objects with the mixed Gaussian model by analyzing the pros and cons of hybrid Gaussian modes.The algorithm solves two problems.Firstly,this thesis uses the inter-frame difference method to initialize the background parameters to improve the modeling speed and ghost problem.Based on the comparison between the adjacent background and thresholds,this approach selects the background update program to improve the background update calculation and addresses the problem of involving large calculation and being time-consuming and sensitive to light.For a system with high real-time requirements,speed is one of the most important factors.The method proposed in this thesis can eliminate the ghost image and improve the efficiency of background updates.Secondly,the shadow problem has been a challenge to extract the video image object,which has great influence on the vehicle extraction and is an indispensable part in the traffic flow detection.With histogram method shadow detection algorithm,the methods proposed in this thesis can detect and further remove the shadow.Finally,the final step in the vehicle flow detection is to count the vehicles.Based on the previous steps,by utilizing the virtual coil detection method,this thesis can count the traffic flows accurately,even with image shadows.