Spatiotemporal Particle Swarm Optimization with Incremental Deep Learning-Based Salient Multiple Object Detection

Abstract

AbstractThe recent developments in the computer vision application will detect the salient object in the videos, which plays a vital role in our day-to-day lives. Difficulty in integrating spatial cues with motion cues makes the process of a salient object detection more difficult. Spatiotemporal constrained optimization model (SCOM) is provided in the previous system. Since the better performance is exhibited in the detection of single salient object, the variation of salient features between different persons is not considered in this method and more general agreement related to their significance is met by some objects. To solve this problem, the proposed system designed a spatiotemporal particle swarm optimization with incremental deep learning-based salient multiple object detection. In this proposed work, incremental deep convolutional neural network (IDCNN) classifier is introduced for a suitable measurement of success in a relative object saliency landscape. Spatiotemporal particle swarm optimization model (SPSOM) is used for performing the ranking method and detection of multiple salient objects. In this system to achieve global saliency optimization, local constraint temporal as well as spatial cues is exploited. Prior video frame saliency map and change detection motion history are done using SPSOM. Moving salient objects are distinguished from diverse changing background regions. When compared with existing methods, better performance is exhibited using proposed method as shown in results of experimentation concerning recall, precision, average run time, accuracy and mean absolute error (MAE).KeywordsSalient objectSpatiotemporal particle swarm optimization model (SPSOM)Incremental deep convolutional neural network (IDCNN) classifierGlobal saliency optimization