A method for detecting foreign objects in railway perimeters under low-light conditions based on image enhancement

Foreign object detection along railway perimeters is critical to ensuring the safe operation of trains. However, lighting conditions in nighttime and low-illumination environments significantly constrain existing methods, resulting in insufficient accuracy in detecting small objects. Additionally, the limited computational capacity of edge devices hampers the deployment of high-performance detection algorithms. To address these challenges, this paper proposes a novel approach that integrates image enhancement with an object detection network. To improve image quality under low-light conditions, the enhancement algorithm is designed by combining Adaptive Median Filtering (AMF) with an improved version of Contrast Limited Adaptive Histogram Equalization (CLAHE). This method significantly enhances image clarity while effectively suppressing noise. For the subsequent foreign object detection task, a lightweight detection network is developed based on the YOLO architecture, to adapt to edge devices and enhance small-object feature extraction. It includes two important parts: adaptive boundary compensation (ABC) and lightweight pyramid fusion (LPF), which allow for effective and precise detection while using very few resources. Experimental results show that the proposed AMF-CLAHE algorithm improves the Peak Signal-to-Noise Ratio (PSNR) by 36.22% over traditional CLAHE. The AF-YOLO network achieves a detection accuracy of 93.0% on AMF-CLAHE-enhanced images. This method effectively addresses the challenges of low image quality and weak small-object features, while its lightweight design ensures compatibility with edge deployment scenarios.