Improved Vehicle Detection Using Weather Classification and Faster R-CNN with Dark Channel Prior

Abstract

Recent advancements in artificial intelligence have led to significant improvements in object detection. Researchers have focused on enhancing the performance of object detection in challenging environments, as this has the potential to enhance practical applications. Deep learning has been successful in image classification and target detection and has a wide range of applications, including vehicle detection. However, object detection models trained on high-quality images often struggle to perform well under adverse weather conditions, such as fog and rain. In this paper, we propose an improved vehicle detection method using weather classification and a Faster R-CNN with a dark channel prior (DCP). The proposed method first classifies the weather within the image, preprocesses the image using the dark channel prior (DCP) based on the classification result, and then performs vehicle detection on the preprocessed image using a Faster R-CNN. The effectiveness of the proposed method is shown through experiments with images in various weather conditions.