Detection Of Human Movement And The Gait Recognition

Moving human detection and gait recognition is an important part in thefield of computer vision, the main task of it is moving human targetsegmentation from video sequences or monitor screens, then, extract the gaitfeatures of moving human, at last to classify the gait based on the features of it.The whole process contain several stages like moving human targetsegmentation, shadow elimination of human, extraction of the gait features,recognition of the gait follow the former work and so on.To reduce the computation of Gauss mixture model effectively and improvethe accuracy of shadow elimination in moving object detection, an algorithmwhich updated the model selectively and eliminated the shadow by the changeof brightness was proposed. Firstly, the weight of the Gauss distribution and therate of that not belong to the background were compared before updating theGauss distribution, if the former was larger, then not to update it, otherwise,update it; Secondly, the ranges of brightness’s change were chose to be athreshold factor of shadow detection, so that the threshold could be adjustadaptively according to the change of brightness. Finally, compared thisalgorithm with the traditional ones through experiments on indoor and outdoorvideos, the experimental results show that the time consumption of the algorithmis about one-third of the traditional ones, the accuracy of shadow eliminating isimproved and the efficiency of the algorithm is confirmed.Based on the accurate moving object extraction, this paper selected amethod for extracting effective gait characteristics, first using the human bodymovement and the horizontal width of the features as the basis of gait cycledetection, so that the number of frames a gait cycle contain can be selectedadaptively; then acquire five key frames of each cycle through the minimum andmaximum distance method and extract five joint points of the gait in each frameamong the cycle; at last, on the base of doing much researching work on therecognition method based on HMM (Hidden Markov Model), common gait havebeen recognized through template matching method by the nearest neighbor method.