Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

The Research of Level Set Methods and Their Applications in Video Vehicle Detection

Author QiuCui
Tutor QiaoShuang
School Northeast Normal University
Course Circuits and Systems
Keywords Vehicles Detection Level Set Methods Statistical Energy Symmetrical Difference Shadow Removal Matlab
CLC TP391.41
Type Master's thesis
Year 2011
Downloads 34
Quotes 0
Download Dissertation

Social economy’s fast development brings us more convenience,meanwhile it also makes some negative influences,the road traffic problem is one of them. Serious traffic jam,contrary driving,frequently happened traffic accidents etc,these questions have brought pressure to the traffic control,so the research on ITS received more and more treasures.In the ITS,The vehicles detection is a basal part.This part’s aim is getting the regions of intrest, gathering traffic parameters such as traffic volume,vehicular speed etc.And, the detection’s quality will afect the following operations like vehicles tracking,the calculation of traffic parameters,even affects overall system’s practicability and performance.So finding one effective method for vehicles detection is important.The article’s value not only in theory but also in application is very high.At first,we studied the traditional algorithms of vehicles detection,like background difference,interframe difference,optical flow method.Then contrastively analyse their good and bad points from the theory and practical application,The background difference is the most commonly used in object detection,the utility is better,but when the grayscale of background and objects is similar,this method will make mistakes,and will induce failure. When above several algorithms were used for segmentation,it will present holes,goal break,goal adhesion etc,that is serious goal distortion,synthesized the above situations,we chose the level set methods for vehicles detection.The LS methods can automatically realize topotactic transformation like fission,merge etc,during the evolution process,effectively get goals’contour.But in the traditional Level Set Methods,the speed function is based on the gradient information.When the boundary’s gradient hasn’t obvious changes or the boundaries are obscure,objects omission will happen.In order to overcome the above shortcoming,we made further improvement to the speed function in the LSM.We used a LSM based on statistical energy.To improve the processing speed,we took the mask picture received by background subtraction as the initial curve.In practical applications,due to the shadow influence,the accuracy of vehicle detection is decreased,therefore,we put forward a shadow removing method,then fused the sports information without shadow to speed function,finally, realized vehicles detection without shadow influence and increase the algorithm’s precision.We have tested the algorithm on Matlab platform.From the test we found that it had good effect,fast processing speed and high precision.It can meet the requirement of real-time.

Related Dissertations
More Dissertations