Year: 2025
Type: conference-paper
Abstract: Object detection in low-light conditions remains a critical challenge for applications in surveillance, autonomous navigation, security, etc. were poor image visibility results in poor detection accuracy especially in very low-light images.This paper presents an analysis of two popular object detection frameworks YOLOv9 and Faster R-CNN on low-light images after enhancing them using optimization techniques.We implemented a Proposed Method technique consisting of a set of image enhancement steps.Our results indicate that YOLOv9 and Faster R-CNN achieve superior performance with higher average recall, precision, and mAP.These results confirm the impact of integrating low-light enhancement methods on the detection accuracy in dealing with the complexities of low-light image environments.This paper contributes to the development of an image enhancement system for low-light conditions and the pursuit of better results in object detection in images.
Source: Computer Science Research Notes
Institution: University of Kerbala
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4412762415