Research On Multi-camera Moving Target Matching In Overlapping Fields Of View

With the developmental process of social informatization, the targettracking system based on multi-camera has attracted more attentionamong the researchers. The multi-camera target tracking system has theadvantages of expanding the filed of vision scope, solving the objectocclusion and analysing the abnormal behavior etc, so it has broadapplication prospect. However how to match the same target betweendifferent camera is the key problem of Multi-camera target tracking. Thisthesis focus on the key technology of muli-camera moving targetmatching in overlapping fields of view, and main contributions and workare described as follows:(1) Research on moving target detection, moving targetclassification and the target’s feature extraction, according tocharacteristics of camera fixed and the camera sence exits brightnesschanges, this thesis uses gaussian mixture model dynamically update thebackground, and uses8-adjacent connection area method label and tracktarget, finally get the target’s feature information to be prepared for thefollow-up study.(2)For the problem of target matching in single camera, this thesiscomes up with a method of moving target matching based on colorcharacteristics.Firstly, define main color region of image by colorhistogram,and then establishs RGB colorspace fuzzy subset of the maincolor region of Image information,and finally compare fuzzymembership function and fuzzy nearness function of the matching and tomatch target by using fuzzy similar matching principle.The experimentsprove that this method can reliably solve the problem of target matchingin video monitoring.(3) According to the problem that matches target in muti-camera inoverlapping view fields.This thesis introduces an algorithm namedSHFSM based on the information of hue feature. By analyzing theinvariance of the target color chrominance components of the lightintensity and according to the histogram of chrominance components, we defined Hue Features Section named HFS and proved its relatedcharacteristic. This algorithm, first of all, obtained the interval set oftarget chromaticity characteristics, and then sortd the set based on thenumber of characteristic pixels distributed in HFS. Finally, matched theset of target sorted by using Fuzzy similar matching principle.Resultsobtained from experiments demonstrate that the algorithm has highaccuracy and good reliability.