Multiple Action Recognition for Human Object with Motion Video Sequence using the Properties of HSV Color Space Applying with Region of Interest
N. Kumaran1, U. Srinivasulu Reddy2, S. Saravana Kumar3
1N. Kumaran, Research Scholar, Department of Applications, NIT, Tiruchirappalli (Tamil Nadu), India.
2Dr. U. Srinivasulu Reddy, Assistant Professor, Department of Computer Applications, NIT, Tiruchirappalli (Tamil Nadu), India.
3Dr. S. Saravana Kumar, Professor, Department of CSE, Shanmuganathan College of Engineering, Pudukkoattai. Tamil Nadu, India.
Manuscript received on 07 April 2019 | Revised Manuscript received on 20 April 2019 | Manuscript published on 30 April 2019 | PP: 1118-1128 | Volume-8 Issue-6, April 2019 | Retrieval Number: F3483048619/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Recognizing the activities of human beings from sequence of images is an on-going area of research in computer vision. Deep Convolutional Neural Network (CNN) models, recently started studies in recognising action of human together with categorization of images. However, these models are excessively encouraged by the background of the image, as well as understanding the existence of clues in standard computer vision datasets. For different robotics practice, the amount of innovation and variation activity backgrounds is too larger than PC vision datasets. To tackle this issue, we present the approach called “Action Region Proposal (ARP)”. In this approach image regions and optical flow are acting as the informing agents which are probably to contain movements for giving contribution to the system at the training just as testing phase. The proposed sub-activity descriptor comprises of three stages: the posture, the locomotion, and the gesture. The proposed activity detection classic localizes and perceives the activities of various persons at the same time in video surveillance by utilizing appearance-based temporal features with multiple CNN. Through a series of examinations, we have demonstrated that the manual background segmentation is insufficient; However the active ARP at the time of training and testing enhances the execution on specific spatial and temporal video constituents. Finally, the study indicates that by focusing attention through ARP, the performance of the present up-to-date spatio-temporal action recognition can be improved.
Keyword: Sub-action Descriptor, Action Detection, Video Surveillance, Convolutional Neural Network, Multi-CNN.
Scope of the Article: Pattern Recognition