An efficient multiple salient object Detection in video using swarm Intelligence with ensemble learning Techniques

Abstract

The Salient Object Detection (SOD) has attracted an increasing newlineamount of research attention over the years. There has been a rising focus on newlineVideo Salient Object Detection (VSOD) in the last few decades. Therefore, newlineVSOD has prominent importance and it is essential for an extensive array of newlinereal-world applications, e.g., video segmentation. VSOD faces big hurdles newlineowing to the problems introduced due to video data and the intrinsic newlinecomplexity of human visual attention behavior during scenes in dynamic newlinemovement. To overcome the above mentioned issues, in this work, swarm newlinebased optimization algorithms and ensemble learning algorithms are proposed newlinefor improving the SOD and VSOD performance considerably. The main aim newlineof this research is to provide optimal multiple salient objects with improved newlineperformance and developing a methodology for reducing the noise in the newlinemovement video.In the first work, spatiotemporal particle swarm optimization with newlineincremental deep learning based salient multiple object detection is proposed. newlineInitially, for a given video sequence, visual and temporal detection of salient newlineobjects in every frame of a video sequence is a major goal. Assumption that, newlinefor given video sequence, by analyzing spatial and temporal cues, background newlineor salient objects of some reliable regions can be found is used in this newlineproposed saliency model and from these detected reliable regions, saliency newlineseeds can be derived for achieving global optimization of salient detection of newlineobject. newline newline