A Novel Approach for Object Recognition in Hazy Scenes: Integrating YOLOv7 Architecture with Boundary-Constrained Dehazing

To improve the detection of objects in the haze environment, we have integrated boundary delimitation techniques into the YOLOv7 architecture in this paper. Currently, existing methods have encountered difficulties balancing the need to improve visibility for low visibilities images with an efficient detection of objects. We are proposing a novel methodology to overcome this problem. Using hazy pictures from the RESIDE SOTS dataset, we assess the effectiveness of dehazing methods using the PSNR and SSIM measures. We contrast several dehazing techniques using hazy pictures taken from the RESIDE SOTS dataset and used PSNR and SSIM measures. By reducing air haze using boundary constraints and then utilizing the cutting-edge YOLOv7 architecture for accurate item identification, our method improves the precision of object detection in hazy settings. The empirical results, derived by contrasting SSIM, PSNR, and mAP measures, show that our suggested strategy outperforms competing dehazing techniques, leading to improved object identification accuracy. To demonstrate the precision of our proposed approach, we also conducted a comparison with the current state-of-the-art method. This method is applicable to many different real-world situations, including autonomous driving, video surveillance, and environmental monitoring, where it is crucial to accurately identify objects in hazy conditions.