Moving object detection in dense fog environment using YOLOv7 framework

Abstract

Moving object detection in thick haze climate is a difficult issue in computer vision, which has significant applications in smart transportation frameworks, reconnaissance, and independent driving. In such conditions, perceivability is fundamentally decreased because of the dispersing and assimilation of light by the haze particles, making it challenging for conventional computer vision computations to recognize moving objects precisely. To address this test, analysts have proposed learning-based approaches that influence the power of profound neural network to gain proficiency with the hidden highlights of the hazy scene and identify moving objects. In this paper, we propose a learning-based approach for moving object detection in thick haze conditions using the YOLOv7 framework. The proposed approach comprises of four principal stages: feature extraction, feature combination, object detection, and non-maximum suppression. The results obtained are very much encouraging as compared to the state-of-the-art schemes.