| Image/video processing is the effective technique in intelligent transportation systems(ITS),and also becomes the promising trend in future traffic object detection and understanding.The main issue existed is how to accurately detect the movable object,restrict the object shadow,robustly recognize and track the specific instances.This dissertation begins from expounding several typical applications in ITS,and focuses on the movable object detection,shadow detection,object recognition and tracking.These tasks are classic but still unsolved effectively.In view of this,the contributing works of this dissertation are as follows.The research status of moving target detection methods are firstly discussed,and three methods,namely optical flow method,frame difference method,background subtraction method are re-visited from their disadvantages operation theory,comparative performance.Then,Gaussian Mixture Model(GMM)is concentrated on in the target detection,which is followed by the shadow detection.To detect the movement of targets is how to justify the true movement of them.this is very important for following recognition and tracking.In this paper,the basic concept of shadow is firstly expounded,and the advantages of several methods are compared for shadow suppression.This dissertation takes the morphological and motor characteristics of moving objects into account and replenishes them as the basis for object recognition,so as to establish a fuzzy recognition system for trucks,cars,motorcycles and pedestrians.Vector machine combined with fuzzy integral technology,the operation.The commonly used tracking algorithm is analyzed,and then the most useful Camshift algorithm and Kalman filter algorithm are summed up as the basic idea of its gist and advantages and disadvantages,as well as their operation.From the experiments,we can conclude than Camshift can robustly tackle the scale variation,and Kalman filter can restrict the influence of the occlusion effectively. |
PDF Full Text Request