| Summary: | 碩士 === 國立屏東教育大學 === 資訊科學系碩士班 === 101 === Traditional Moment although could be tracking objects, not only could not directly tracking multiple objects, but also it’s susceptible to noise interference. In this study, we use Orthogonal Variant Moments features to solve this problem. The first step we use the Background Subtraction method to find the some edges of motion objects, and then we use the Morphology method to solve some edges are disconnected of objects. The second step, compute the OVM features in the tth frame in while OVM features can be use to five the different image features of property (e.g., average value of pixel, vector of horizontal orientation, vector of vertical orientation, position of horizontal, position of vertical ). Then we use OVM features with some edges of multiple objects to do logic AND operator. The third step, we label each objects. The fourth step, we use K-means Clustering algorithm to partition the n OVM features into k sets from each objects, and save it. The fifth step, we repeat from the first step to the fourth step in t+1th frame. The last step, since we have gotten information of each object of two continuous frames, we can match these objects.
In the experimental results showed that OVM features have high discriminatory power for multiple of motion objects, and we effective use OVM features of property combine with K-means Clustering algorithm achieved the multiple objects tracking technique.
|
|---|
