计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (6A): 250500016-8.doi: 10.11896/jsjkx.250500016

• 图像处理&多媒体技术 • 上一篇    下一篇

融合多尺度特征筛选的轻量露天矿爆堆矿石图像分割算法

顾清华1,2,4, 马祥1,4, 李学现3,4   

  1. 1 西安建筑科技大学信息与控制工程学院 西安 710055
    2 西安建筑科技大学资源工程学院 西安 710055
    3 西安建筑科技大学管理学院 西安 710055
    4 西安建筑科技大学西安市智慧工业感知计算与决策重点实验室 西安 710055
  • 出版日期:2026-06-16 发布日期:2026-06-12
  • 通讯作者: 顾清华(qinghuagu@126.com)
  • 基金资助:
    国家自然科学基金面上项目(52374135,52074205);陕西省金属矿智能开采理论及技术创新团队(2023-CX-TD-12)

Multi-scale Feature Screening-integrated Lightweight Algorithm for Blast Heap Ore Image Segmentation in Open-pit Mines

GU Qinghua1,2,4, MA Xiang1,4, LI Xuexian3,4   

  1. 1 College of Information and Control Engineering School of Recourse Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China
    2 School of Recourse Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China
    3 School of Management,Xi'an University of Architecture and Technology,Xi'an 710055,China
    4 Xi'an Key Laboratory of Smart Industry Perception Computing and Decision Making,Xi'an University of Architecture and Technology,Xi'an 710055,China
  • Published:2026-06-16 Online:2026-06-12
  • About author:GU Qinghua,born in 1981,Ph.D,professor,is a member of CCF(No.F4272M).His main research interests include mutil-objective optimization and intelligent mining.
  • Supported by:
    National Natural Science Foundation of China(52374135,52074205) and Shaanxi Province Metal Mine Intelligent Mining Theory and Technology Innovation Team(2023-CX-TD-12).

PDF (PC)

42

摘要： 随着智慧矿山的快速发展,实时精准识别爆堆矿石铲装过程中的大块矿石已成为保障运输安全与效率的关键需求。针对露天矿爆堆矿石图像存在形状高度不规则、颗粒间严重重叠、图像分辨率低及特征稀疏等挑战,文中提出一种轻量化爆堆矿石图像分割算法,通过多维度模型优化实现精度与效率的平衡。首先,使用DynamicHGNetv22(Dynamic High Performance GPU Network version2)层级图网络的拓扑特性重构主干网络,通过动态路由机制压缩冗余特征,将模型体积缩减42.4%;其次,设计HSFPN(High-level Screening-feature Fusion Pyramid)高阶筛选特征融合金字塔作为颈部网络,采用通道注意力引导的多尺度特征筛选机制,在降低27.4%计算量的同时提升跨尺度特征融合能力;然后,构建轻量化分割头(Light Head),通过深度可分离卷积与特征蒸馏技术进一步优化计算效率;最后,引入EMASlideLoss(Exponential Moving Average SlideLoss)损失函数,基于指数移动平均策略动态调节难易样本权重,显著提升模型对低质量矿石目标的边缘分割精度。实验结果表明,相较于 YOLO11n-seg基准模型,所提方法的参数量和计算量分别缩减42.4%和27.4%,mAP50和mAP50:95分别提高了0.1%和2.1%,不仅满足矿山场景高精度实时分割需求,其轻量化特性更是可直接部署于边缘计算设备,为智能铲装系统的大块矿石预警提供可靠技术支撑。

关键词: 露天矿, 爆堆, 图像分割, YOLO11, 轻量化网络

Abstract: With the rapid development of smart mines,the real-time and precise identification of large rocks in the blasting operation for ore loading has become a key requirement for ensuring transportation safety and efficiency.To address the challenges posed by highly irregular shapes,significant overlap among particles,low image resolution,and sparse features in the images of the blasted heap ore,this paper proposes a lightweight image segmentation algorithm for blasted heap ore that achieves a balance between accuracy and efficiency through multi-dimensional model optimization.Firstly,the topological characteristics of the DynamicHGNetv2(Dynamic High Performance GPU Network version2) hierarchical graph network are utilized to reconstruct the backbone network,compressing redundant features through a dynamic routing mechanism,which reduces the model size by 42.4%.Secondly,the HSFPN(High-level Screening-feature Fusion Pyramid) is designed as the neck network,employing a multi-scale feature screening mechanism guided by channel attention,which reduces the computational load by 27.4% while enhancing the ability for cross-scale feature fusion.Subsequently,a lightweight segmentation head is constructed,optimizing computational efficiency further through depthwise separable convolutions and feature distillation techniques.Finally,the EMASlideLoss(Exponential Moving Average SlideLoss) loss function is introduced,dynamically adjusting the weights of difficult samples based on an exponential moving average strategy,significantly improving the model's edge segmentation accuracy for low-quality ore targets.Experimental results indicate that,compared to the YOLO11n-seg benchmark model,the proposed method reduces the number of parameters and computational costs by 42.4% and 27.4%,respectively,while mAP50 and mAP50:95 improve by 0.1% and 2.1%,respectively.This not only meets the needs for high-precision real-time segmentation in mining scenarios but also can be directly deployed on edge computing devices,providing reliable technical support for early warning systems for large ore in smart shoveling systems.

Key words: Open-pit mine, Bast heap, Image segmentation, YOLO11, Lightweight network

中图分类号: 

  • TP391
[1] SADEGHI F,MONJEZI M,JAHED A G D.Evaluation and Optimization of Prediction of Toe that Arises from Mine Blasting Operation Using Various Soft Computing Techniques [J].Natural Resources Research,2020,29(2):887-903.
[2] WANG W,LI Q,ZHANG D Z,et al.A Review of Ore Image Processing Research Based on Deep Learning [J].Journal of Engineering Science,2023,45(4):621-631.
[3] ZHAN Y,ZHANG G.An Improved OTSU Algorithm UsingHistogram Accumulation Moment for Ore Segmentation[J].Symmetry,2019,11(3):431.
[4] MUKHERJEE D P,POTAPOVICH Y,LEVNER I.Ore image segmentation by learning image and shape features[J].Pattern Recognition Letters,2009,30(6):615-622.
[5] WANG R,ZHANG W,SHAO L.Research of Ore Particle Size Detection Based on Image Processing[C]//Proceedings of 2017 Chinese Intelligent Systems Conference.Singapore,2018:505-514.
[6] LIU X,ZHANG Y,JING H,et al.Ore image segmentationmethod using U-Net and Res_Unet convolutional networks[J].RSC Advances, 2020,10(16):9396-9406.
[7] SUN G,HUANG D,CHENG L,et al.Efficient and Lightweight Framework for Real-Time Ore Image Segmentation Based on Deep Learning[J].Minerals,2022,12(5):526.
[8] JIAO A B,HE M,LUO H B.Research on Significant Edge Detection of Infrared Image Based on Deep Learning[J].Infrared Technology,2019,41(1):72-77.
[9] GU Q H,WEI F W,GUO M L,et al.A Fragmented Ore Image Segmentation Method Based on an Improved HED Network Model [J].Laser & Optoelectronics Progress,2022,59(2):262-270.
[10] PENG D G,CHEN C,WANG D H,et al.Leakage Segmentation and Detection of Pipelines and Valves in Thermal Power Plants Based on Improved YOLOv7 [J].Control and Decision,2024,39(9):2977-2986.
[11] LIU X B,ZHONG B R.An Efficient Rock Cuttings Instance Segmentation Algorithm Based on Improved YOLOv8 [J].Modern Electronics Technique,2025,48(9):1-7.
[12] LIU J J,SUN F Q,ZHANG H X,et al.A Lightweight Instance Segmentation Method for Coal Dust Particle Groups Based on YOLOv8 [J].Journal of China Coal Society,2024:1-14.
[13] LIU X,RAO Z,ZHANG Y,et al.UAVs Images Based Real-Time Insulator Defect Detection with Transformer Deep Learning[C]//2023 IEEE International Conference on Robotics and Biomimetics(ROBIO).Koh Samui,Thailand:IEEE,2023:1-6.
[14] CHEN Y,DAI X,LIU M,et al.Dynamic Convolution:Attention over Convolution Kernels[J].arXiv:1912.03458,2020.
[15] DING X,ZHANG X,MA N,et al.RepVGG:Making VGG-style ConvNets Great Again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).Nashville,TN,USA:IEEE,2021:13728-13737.
[16] CHEN Y,ZHANG C,CHEN B,et al.Accurate Leukocyte Detection Based on Deformable-DETR and Multi-Level Feature Fusion for Aiding Diagnosis of Blood Diseases[J].Computer in Biology and Medicine,2024,170:107917.
[17] FAN Q,LI Y,DEVECI M,et al.LUD-YOLO:A novel light-weight object detection network for unmanned aerial vehicle[J].Information Sciences,2025,686:121366.
[18] HE K,GKIOXARI G,DOLLÁR P,et al.Mask R-CNN[J].arXiv:1703.06870,2018.
[19] BOLYA D,ZHOU C,XIAO F,et al.YOLACT:Real-time Instance Segmentation[J].arXiv:1904.02689,2019.
[20] WANG X,ZHANG R,KONG T,et al.SOLOv2:Dynamic and Fast Instance Segmentation[J].arXiv:2003,10152,2020.
[21] YU W H,WANG X H.MambaOut:Do We Really Need Mamba for Vision? [J].arXiv:2405.07992,2024.
[22] HE H Y,ZHANG J N,CAI Y X,et al.MobileMamba:Light-weight Multi-Receptive Visual Mamba Network [J].arXiv:2411.15941,2024.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发