Abstract: Nighttime operations at mine weighbridges are characterized by complex working conditions such as insufficient lighting, pervasive coal dust, and frequent truck traffic. Manual inspection of coal products not only poses safety risks to personnel but also leads to unstable recognition accuracy due to subjective bias and high workload. To address these challenges, this paper proposes SDCD−YOLO11, an automatic coal product recognition network designed for such complex operational environments. The model incorporates four key improvements: first, SCINet is integrated at the front end to enhance image quality under low−light and coal−dust interference; second, dynamic snake convolution is embedded to adaptively adjust the receptive field for better discrimination of coal product texture features; third, the CBAM attention module is employed to optimize weight distribution in the feature fusion layers; and finally, a Dual Dynamic Task−Aligned Head (DDTAH) is introduced to address recognition challenges caused by unclear coal pile boundaries, irregular shapes, and dust occlusion. Experimental validation on a self−constructed multi−disturbance dataset shows that the proposed method outperforms the baseline YOLO11 by 6.5 percentage points in mAP@0.5 and 4.3 percentage points in mAP@0.5:0.95, while maintaining an inference speed of 120 FPS. This method can effectively replace manual inspection during nighttime weighbridge operations, achieve automatic and accurate coal product recognition under complex working conditions, and provide technical support for the development of unmanned intelligent monitoring systems for mine weighbridges.