基于改进YOLOv8n的煤矿井下受限场景目标检测算法

打开文本图片集

中图分类号：TD67 文献标志码：A

Abstract：In constrained scenariosofunderground coal mines，object detectionaccuracy is low due to complexobject scale variation，partial occlusion ofobjects，and dificulty inextracting effective features.To address these problems，an improved YOLOv8n-based object detection algorithm forconstrained scenarios in underground coal mines was proposed.In the backbone feature extraction network，Receptive Field Atntion Convolution （RFAConv）wasadopted to better process the spatial location information of objects inconstrained environmentsand was used to dynamically adjust weightsaccording to the importanceof features，thereby focusing more on the key features of objects.In the neck，the Efcient Multiscale Atention （EMA） module was introduced to fuse feature informationat diferent scales，which improved the detectionaccuracy of objects with scale variation. The new Deformable Convolutional Networks v3 （DCNv3） and Dynamic Head were combined， integrating scale-aware attention，spatial-aware atention，and task-aware atention， which helped the model focus on spatial scale information and adapt to different detection tasks，thus enhancing the detection ability for multiscale and partially occluded objects.A Unified-IoU （U-IoU）loss function thatconsiders the weight distribution of prediction boxes was introduced. By dynamically adjusting the atention on prediction boxes of different qualities， the model focused more on high-quality prediction boxes， improving the convergence speed and accuracy. Experimental results showed that the improved YOLOv8n achieved a 5.6% improvement in mAP @0.5 compared with YOLOv8n in underground coal mine conveyor belt foreign object detection for the CUMT-BelT dataset; in different fully mechanized mining face operation scenarios， the overall mAP @0.5 increased by 4.8% compared with YOLOv8n for the DsLMF dataset， eectively reducing false detections and duplicate detections.

Key words： object detection; constrained scenarios in underground coal mines; YOLOv8n; Receptive Field Atention Convolution; Eficient Multiscale Attention; Deformable Convolutional Networks v3; Dynamic Head

0引言

煤矿复杂且受限的环境使决策的不确定性和风险上升[1]，需要高性能的目标检测模型为操作人员提供关键的决策支持。（剩余12870字）