Abstract: To address the issues of low detection accuracy, excessive model complexity, and high deployment and maintenance difficulty in existing scraper conveyor chain detection methods based on deep learning under low illumination conditions in coal mines, a lightweight chain detection algorithm based on improved YOLOv11n—YOLO-Chain—was proposed. First, an image edge information enhancement module was constructed to optimize the C3k2 module of YOLOv11n, effectively extracting edge features from chain images. Then, a weighted Bidirectional Feature Pyramid Network (BiFPN) was used to replace the neck network of YOLOv11n, thereby effectively reducing the number of model parameters and lowering model complexity. Finally, a lightweight detection head was introduced to capture subtle features of chain scale variations in complex underground scenarios, further reducing redundant parameters and model complexity, improving the detection performance of the lightweight model, and providing support for subsequent chain fault detection. Experimental results on a single-scenario scraper chain image dataset from a coal mine in Shanxi showed that, compared with the original YOLOv11n model, YOLO-Chain improved the mAP@0.5:0.95 accuracy by 3.7%, while reducing the number of parameters and computational load by 35% and 10%, respectively, and decreasing the model size by 31%. Compared with current mainstream models such as the YOLO series, SSD, Faster RCNN, and RT-DETR-R18, YOLO-Chain also demonstrated advantages in multiple indicators. Experimental results on a multi-scenario chain image dataset collected from multiple coal mines under complex working conditions such as low illumination, smoke interference, dust occlusion, and partial obstruction showed that the F₁-score and mAP@0.5 of YOLO-Chain increased by 0.2% and 0.5%, respectively, compared with YOLOv11n, with arithmetic speed increased by 8, demonstrating good applicability and generalization ability.