Work

Low-light object detection via adaptive enhancement and dynamic feature fusion

Year: 2025

Type: article

Abstract: Under low-light conditions, object detection tasks face challenges such as low brightness, low contrast, and noise, which can lead to missed or incorrect detections. To address this issue, this paper proposes a low-light enhancement algorithm, called DAMFCN, and an improved DarkYOLOv8 method, aimed at enhancing low-light image quality and object detection performance. DAMFCN significantly improves the quality of low-light images by integrating the Low-Light Adaptive Module and the Multi-Scale Feature Compensation Block, where LLAM effectively extracts fine details and suppresses noise, and MSFCB compensates for lost details by integrating multi-scale information. The DarkYOLOv8 framework, built on the EfficientNet backbone, combines a multi-scale attention mechanism and the Dynamic Feature Fusion Attention Module, demonstrating superior object detection performance under low-light conditions. Experimental results show that the proposed methods outperform existing state-of-the-art techniques in terms of accuracy, robustness, and efficiency, offering broad application potential.

Source: Alexandria Engineering Journal

DOI: 10.1016/j.aej.2025.04.047

Authors: Siyong Fu, Qinghua Zhao, Hesheng Liu, Qiuxiang Tao, Danjuan Liu

Institution: Xinyu University

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $1,900

Funder: Education Department of Jiangxi Province