Infrared and Visible Light Fusion for Object Detection with Low-light Enhancement

In low-light environments, conventional object detection methods often suffer from poor performance and low accuracy. To address these challenges, this paper proposes a novel object detection method that combines low-light enhancement with visible-infrared fusion. By innovatively integrating infrared spectrum images into the object detection process and incorporating a dedicated low-light enhancement module, our approach significantly improves the detection capabilities of traditional methods. We conducted extensive qualitative and quantitative experiments to validate the effectiveness of the proposed method. The results demonstrate that our approach not only enhances detection accuracy but also reliably identifies objects in challenging low-light conditions, thus providing a robust solution for applications requiring high precision in adverse lighting environments.

I. Introduction

The fusion of infrared and visible images is a hot topic in the field of dual-band object detection [1]. Visible light cameras can provide rich visual details in well-lit environments, capturing abundant color, texture, and detail information on the surface of objects. In contrast, infrared cameras are highly sensitive to temperature variations, making them suitable for special conditions such as nighttime and occlusions. They can detect object contours based on thermal radiation, unaffected by ambient light, which makes them significant for object detection. Even under low light or other harsh conditions like occlusion, fog, or smoke, they can still achieve good detection accuracy for vehicles, pedestrians, and other objects. In contrast, visible light images are more susceptible to the influence of ambient lighting. To address the problem of all-day operations, visible light images and infrared images are often fused or switched to address lighting issues [2], [3].