面向无人机图像矿区车辆细粒度识别与检测方法研究OA

In response to the increasingly severe problems of irregular mining and illegal extraction in open-pit coal mines,as well as the difficulty of vehicle supervision,this study conducts research on the recognition,detection,and management of various engineering vehicles based on UAV aerial monitoring.Due to the wide variety of vehicles and their high similarity from the UAV perspective,fine-grained recognition accuracy of engineering vehicles faces high requirements.To address this,a lightweight LE-DETR fine-grained rec-ognition and detection model is proposed,with four structural improvements made on the RT-DETR framework:introducing the LSKNet backbone to enhance detail modeling capability;employing TransformerEncoderLayer_HiLo for multi-scale semantic fusion;designing the CSP-OmniKernel module to improve cross-scale feature perception;and incorporating the Shape-IoU loss function to optimize bounding box regression.Meanwhile,a high-resolution self-constructed dataset,MINEV,was built,on which ablation and comparative experiments were conducted.The results show that the proposed algorithm achieves an overall accuracy improvement of 5.8%compared with the baseline algorithm on the dataset.Therefore,the improved model enhances the detection accuracy of mining vehicles and demon-strates good detection performance.