Abnormal Behavior Based on Improved Hu Moments
Liu Danping · Computer Technology and Development · 2009
Proposed an abnormal behavior recognition algorithm based on improved Hu moments.There are six kinds of abnormal behavior,including jump, accelerate run,trip,squat,wave,stick in hand.Firstly,the contours of motion human are extracted.Then,the features of the shape are extracted by improved Hu moments.Finally,Hausdorff distance is used to calculate the similarity between currently shape features vector which is recognized and normal shape template one.Calculating corresponding threshold value judge whether the behavior is abnormal or not.Experiments demonstrate the proposed recognition method can identify the six abnormal behaviors with the recognition rate of 100% and have practical value.