Low-Light Object Detection Based on Reflection Map Guidance and Cross Attention

The performance of existing object detection algorithms is significantly affected by image quality degradation caused by low-light conditions, making detection in such scenarios challenging. To enhance the accuracy and robustness of object detection under low-light conditions, this paper proposes a low-light object detection method guided by reflection maps and a cross-attention mechanism. First, a decomposition module based on Retinex theory is designed to extract reflection map information, effectively preserving the detailed features of target regions. Secondly, a feature adjustment module is proposed to optimize extracted features from the reflection maps. Finally, a cross-attention-based feature fusion module is developed, dynamically integrating high-level semantic features with low-level detailed features, thereby improving the model's detection capabilities under complex illumination conditions. Experimental results on the ExDark dataset demonstrate that the proposed method achieves superior detection accuracy compared to current mainstream methods, validating its effectiveness in low-light environments.