Research On Moving Object Detection And Tracking Algorithms In Complex Condition

To detect and track moving target, which is the foundation to fulfill intelligentsurveillance, target recognition, is the core part of the video surveillance, therefore alarge number of domestic and international scholars concern about them, and theyhave already obtained some achievements. However, the technology of detection andtracking of moving target in real outdoors with complex background conditions isimmature, and need to overcome the interference caused by the light mutation，theshaking leaves, the obstructions and the obscures between multiple targets. All thesefactors are great challenges to achieve the accurate detection and stable tracking ofmoving target. This article focused on the detection and tracking of moving target inthe complex and changeable outdoor environments. The contents of the paper are asfollows:1. A hardware platform of the target detection and tracking system was set up,and the improvements of the algorithms of target detection and tracking in thecomplex and changeable environments were proposed.2. For target detection in dynamic backgrounds, a modeling approach based onGaussian mixture background model was proposed. Background subtractionalgorithm was used to detect moving target. The target segmentation algorithm basedon dynamic threshold could reduce the impact of illumination changes, whichimproved the accuracy of the detection of the target.3. A target detection algorithm based on the replacement of the background of thecentral area was presented so as to enable rotating camera to track special one.According to certain searching strategy, accurate detection of targets under movingbackgrounds was achieved.4. The MSPF tracking algorithm put forward by our laboratory was improved.First, with the use of the original algorithm, it was easy to lose the target in a regionwhere illumination changed. Now when the illumination changed strongly, a newtarget matching algorithm based on HS chromaticity histogram was adopted to reducethe impact on the accuracy of the target matching caused by the luminance component.By doing so, the problem was solved. Second, when original algorithm was applied in environments with obstructions, matching the moving target accurately wasimpossible. The new matching algorithm using multiple sub-modules instead ofsingle-template made it possible. Now when multiple moving targets are under mutualobstructions or partially blocked, even fully, by some static objects, stable tracking isavailable.5. A large number of experiments in the outdoor environments were carried out,providing some facts for the study of the algorithms for target detection and trackingunder complex environments. Finally, through a series of experiments, theeffectiveness of the improved MSPF was verified.