Artificial Neural Network for Human Object Interaction System Over Aerial Images
Mahwish Pervaiz, Ahmad Jalal
Abstract
Recognition of human and object interaction is an important milestone in image understanding and event analysis. Recognizing interactions between humans and objects is the most promising way in visual classification to analyze activities or events happening at any place. Many researchers have invested their efforts in the field of activity recognition between humans and objects. However, some challenges are still open due to incorrect interaction inferences, occlusion between a person and target objects, unrelated target objects, or unclear activities. The major goal of this research project is to provide a useful system for categorising event classification and human-object interaction. Preprocessing, feature extraction, feature optimization, and classification using an artificial neural network are the four main processes of the proposed method. The Games Action dataset, which contains aerial photos, has been used to test the proposed technique. Results demonstrate the effectiveness of the suggested system.