Detection and Recognition of Multitask Human Action Identification from Preloaded Videos Using CCTV Stationary Cameras
Year: 2025
Type: article
Abstract: Human activities include group activities, individual actions, and interactions between objects and people. In computer vision technology, recognizing and categorizing these activities is a vital process. This system will aim at developing a model that can recognize and detect such behavior and apply it in surveillance, health care, military operations, and patient monitoring. In the first place, videos were gathered in order to get a better understanding of the various human activities and interactions. Subsequently, we have converted Video frames into images and pre-processed each image. Characteristic features are extracted from video images by capturing spatial and temporal details. Spatio-temporal interest points using three descriptors Harris STIP, Gabor STIP, and HoG3D STIP are extracted as features. Extracted features are passed to a Heatmap generation process which gives confident key features related to human action. Support Vector Machine (SVM) Classifier is used to analyze these confident key features to label and classify human actions. Various classifier performance metrics, such as accuracy, sensitivity, and specificity, were used to evaluate the performance of the system. Classifiers exhibiting accuracy of around 98.60% served as an indicator of the overall reliability of the proposed system in effectively recognizing human actions.
Source: Journal of Machine and Computing
Authors: Pavankumar Naik, Srinivasa Rao Kunte R
Institution: Srinivas University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4411891817