Study Of Multiple Object Tracking Algorithm Based On Fuzzy Logic

Multiple Object Tracking(MOT)is an important research topic in the field of computer vision,image processing and pattern recognition.It is extensively applied in many areas,such as intelligent surveillance,human-computer interaction,and unmanned vehicle.Numerous factors affect the performance of a tracking algorithm,including frequent interactions,illumination variation,occlusions,background clutters.Therefore,how to realize a real-time,robust tracking system which can work in various complex scenarios is still a hot topic.This paper mainly studied several key issues such as data association and trajectory association in multi-object tracking of static camera scenario.The main research contents are as follows:To improve the performance of multi-object tracking for static single camera in the complex scenario with frequent occlusions and cluttered backgrounds,a novel online multiobject tracking algorithm based on fuzzy logic is proposed.In the proposed algorithm,firstly,the affinity of multiple features including appearance feature,shape feature and spatial feature,between the objects and the detections are calculated.Secondly,the fuzzy rule base is built by introducing the expert knowledge.It adaptively allocates the weight values of each feature by using fuzzy logic,and the association cost matrix between the objects and the detections is obtained.Thus the greedy algorithm in optimization solution problem is used to associate each detection to its object by using the association cost matrix above.Finally,the Kalman filter is used to estimate the object's track to obtain smooth tracking trajectories.Experimental results demonstrate that the proposed algorithm,which has strong robustness and accuracy,can effectively track multiple object in the case of appearance similarity,frequent interaction,occlusion and background clutters.To solve the identity switch problem of multi-object tracking with long-term occlusion and missing detection,a multi-object tracking algorithm based on fuzzy track-to-track association is proposed.In the proposed algorithm,firstly,the fuzzy membership degrees are built by introducing the appearance feature,shape feature and motion feature of the object trajectory,and the fuzzy synthetic function is applied to calculate the synthetic affinity degree between the terminated tracks and new tracks.Then,both maximum comprehensive degree and threshold discrimination principle are utilized to associate the fragmented tracks.Finally,a bidirectional predictive method is used to fill the vacancies between the fragmented tracks to obtain complete and continuous tracking trajectories.Experimental results demonstrate that the proposed algorithm can greatly reduce the number of the identity switch in the multi-object tracking and improve the tracking performance.