Design of Algorithm for Robust Detection of Multi object

Abstract

Digital cameras are heavily engaged in our daily life and used for various applications such as medical diagnosis, investigations, surveillance, driver assistance systems, robotics, parking assistance system, packing industries etc. About thousands of photographs are generated each day and in order to understand it effectively, a robust multiple object identification and detection of an image appeared as an essential requirement. The present research work is carried out focusing the issue of an automatic object(s) localization in an image. The process is also known as object detection. newlineThe multiple challenges that have been witnessed from the reported literature work such as lack in accuracy, high error rate, least accurate object identification and localization, and also are the main reasons to cause limitations in optimal object detection from an image. The aim of the study is to emphasize the limitations by exploring the major approaches of object detection in order to find an optimal solution. newlineEdge detection serves as a most reliable tool for feature detection and object detection in computer vision, digital image processing and machine vision. The study presents an optimal multiple threshold approach with an effective and simple method for edge detection (entitled as B-Edge method). The proposed edge detection method endeavours an improvement in connectivity, thinness of the edges, higher similarity to the ground truth and reduces the error rate from the reported fallouts. Salient object detection has always been a valuable tool in image processing. A new robust object detection algorithm through edges using saliency approach is proposed (entitled as BE-SAL) that reconsidered some design varieties such as background probability, background weighted contrast, foreground regions and element distributions of the reported prior methods. It covers both boundary connectivity from background cues and foreground connectivity. Experiment results also reveals that the proposed BE-SAL method has rendered a promising