EFFICIENT HIERARCHICAL MOTION FILTERING FOR ENHANCED ACTION RECOGNITION IN DENSE VIDEO ENVIRONMENTS

This study suggests a powerful hierarchical motion filtering technique for enhancing the precision of action detection in busy video situations. In order to extract the most important motion information and minimise distracting background motions, it requires segmenting the person from the video, detecting 2-D Harris corners, computing the Motion History Image (MHI), while applying a hierarchical filtering technique. We utilise a Gaussian Mixture Model (GMM) to categorise the retrieved characteristics in order to categorise the activities. Our method produces cutting-edge outcomes on the two difficult datasets, UCF50 and HMDB51. The suggested method differs from existing approaches in a number of ways, including its ability to handle the distracting background motions brought on by moving objects in busy video environments, the accuracy of its action recognition results due to the hierarchical filtering technique and the use of GMM for classification, and its computational efficiency and suitability for real-time applications.