Based on the YOLO-DIS Open-Pit Mine Vehicle Target Detection Algorithm

To address issues such as complex background interference, obstacle occlusion, and missed detection of small targets caused by dust in the open-pit mine operation area environment, an improved YOLO11-based target detection algorithm for mine vehicles, denoted as YOLO-DIS, is proposed. Firstly, the iterative attention mechanism (iAFF) is introduced to improve the traditional C3k2 module, and a C3k2_iAFF structure is designed to enhance the feature extraction capability under complex conditions. Secondly, the lightweight dynamic upsampling method (Dysample) is adopted to replace the traditional upsampling method, which effectively compensates for the defect of inaccurate edge feature restoration of occluded targets caused by fixed rules. Finally, the SlideLoss function is used to realize dynamic adjustment to tackle the problem of uneven sample distribution. Experimental results show that the detection accuracy of the proposed algorithm on the self-made dataset reaches 81.5%, which is 4% higher than the mAP50% of the original YOLO11 model. To verify the generalization ability of the improved algorithm, validation experiments are conducted on the KITTI dataset, where the mAP50% increases from 0.911 to 0.926. The algorithm demonstrates better detection accuracy, indicating its practicality and scientificity.