MODIFIED BOOSTING CLASSIFICATION SYSTEM FOR HUMAN ACTION CLASSIFICATION USING 3D MODIFIED HARRIS CORNER DETECTOR

Abstract

In real world most of the applications are using the data mining techniques for mining the movable or stable images.  The data mining technique is mainly used in disease diagnosis, action classification, object identification, military application and etc. This paper gives the technique for action classification of real time video images. Most of the research papers are using the support vector machine (SVM) for action classification. Commonly, the SVM and Adaboost techniques are having the good performance compare to other classification techniques. But, when compared to SVM the adaboost technique having less tuning parameters to increase the performance of classification algorithm and the main disadvantage of this technique is noise sensitivity. This paper concentrates the modified adaboost technique for action classification. This technique avoids the drawbacks of existing adaboost classification technique. This technique is applied into an image for action classification of moving objects. The result shows the preprocessed image, identification of moving objects, corners detection and action classification of an image. Here, the evaluation measure is used to evaluate the performance of classifier. Here the performance analysis shows that the performance of proposed algorithm is increased when compared to existing classification algorithm.