Moving Object Detection And Tracking

Moving object detection and tracking is of great significance to the construction of intelligent city.Also,the application scenarios are developing from fixed field of view to large field of view and from static scene to dynamic scene,which bring great challenge to moving object detection and tracking.In this thesis,new methods for moving object detection and tracking based on large field of view and dynamic scene are studied separately with a fully understanding of the existing moving object detection and tracking algorithms.The main content of the thesis includes the following aspects:Firstly,moving object detection based on the static scene is studied.The image registration is used to compensate the motion of background by analyzing the difference between dynamic scene and static scene.To reduce the error of moving object detection result which is obtained with inter-frame differential,the motion analysis method with block optical flow is proposed.At the same time,the temporal and spatial enhancement method is used to make the moving object detection result more integrated.Besides,MSER algorithm is applied to extract the moving object area.The experimental result shows that an ideal performance of moving object detection can be achieved.Then,the problem of moving object tracking in cross-field is to be considered.According to a good knowledge of common feature extraction methods of moving object tracking,the major color spectrum histogram which is validated with feature matching distinction and reliability is exploited to represent moving object.With the moving object detection result based on static scene,an algorithm for moving object tracking in cross-filed is realized.The experimental results indicate that a stable and continuous moving object tracking result based on large field of view can be obtained with the proposed algorithm.Finally,a two-dimensional association table of blob matching is established by making use of moving object detection result based on dynamic scene.On the basis of association table,an algorithm for moving object tracking based on dynamic scene which focuses on the criteria and solution for the object splitting and merging is realized.According to the experimental result,the proposed algorithm is proved to be robust for the moving object tracking.