基于MBI-YOLOv8的煤矸石目标检测算法研究

doi:10.11996/JG.j.2095-302X.2024061301

图学学报 ›› 2024, Vol. 45 ›› Issue (6): 1301-1312.DOI: 10.11996/JG.j.2095-302X.2024061301

• 图像处理与计算机视觉 • 上一篇下一篇

基于MBI-YOLOv8的煤矸石目标检测算法研究

李珍峰¹(), 符世琛¹(), 徐乐², 孟博¹, 张昕¹, 秦建军¹

1.北京建筑大学机电与车辆工程学院，北京 102616
2.麦克马斯特大学工程实践与技术学院，汉密尔顿 L8S4L8

收稿日期:2024-07-26 接受日期:2024-08-29 出版日期:2024-12-31 发布日期:2024-12-24
通讯作者:符世琛(1991-)，男，讲师，博士。主要研究方向为机器人智能导控。E-mail：fushichen@bucea.edu.cn
第一作者:李珍峰(2000-)，男，硕士研究生。主要研究方向为计算机视觉。E-mail：bluekite00@163.com
基金资助:
中国博士后科学基金(2019M660860);北京市属高校基础科研业务项目(青年科研创新专项X21051);住房和城乡建设部科技项目计划(2020-K-150)

Research on gangue target detection algorithm based on MBI-YOLOv8

LI Zhenfeng¹(), FU Shichen¹(), XU Le², MENG Bo¹, ZHANG Xin¹, QING Jianjun¹

1. School of Electromechanical and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 102616, China
2. W Booth School of Engineering Practice and Technology, McMaster University, Hamilton L8S4L8, Canada

Received:2024-07-26 Accepted:2024-08-29 Published:2024-12-31 Online:2024-12-24
Contact: FU Shichen (1991-), lecturer, Ph.D. His main research interests cover intelligent robot guidance and control. E-mail：fushichen@bucea.edu.cn
First author：LI Zhenfeng (2000-), master student. His main research interest covers computer vision. E-mail：bluekite00@163.com
Supported by:
China Postdoctoral Science Foundation Project(2019M660860);Basic Scientific Research Business Projects of Beijing Municipal Universities (Youth Scientific Research and Innovation Special Project X21051);Ministry of Housing and Urban Rural Development Science and Technology Project Plan(2020-K-150)

摘要/Abstract

摘要：

为在煤矸石分拣领域实现检测性能与资源消耗的平衡，提出一种基于改进YOLOv8的适用于低性能检测平台的高效实时轻量化目标检测算法。首先以YOLOv8n为基础网络架构，引入MobileNetv3替换原有的主干网络，利用其轻量级结构特性降低模型参数量及运算量，提高模型检测速度；其次引入特征增强网络BIFPN模块，通过多尺度特征融合来弥补引入轻量级网络带来的检测精度损失，实现在保证检测精度的情况下完成模型轻量化；最后引入Inner-CIoU边界框回归损失函数平衡不同质量图像的训练结果，提高模型的定位能力，进一步提高检测精度及速度。为验证改进算法的有效性，进行了实验对比分析，将其与YOLOv3-tiny，YOLOv5n，YOLOv7以及YOLOv8n等算法在自建数据集上进行对比。实验结果表明，该算法展现出了最优的综合检测性能，在保证检测精度的前提下，其参数量降低到1 188 725，相较于YOLOv8n减少了60.46%，运算量由原模型的8.1 GFLOPs降低到2.8 GFLOPs，FPS由YOLOv8n的86.02 Hz提升到216.58 Hz。研究表明，该算法是一种高效实时轻量化煤矸石检测算法，综合检测性能有效提高，实现了模型检测性能与计算资源消耗的平衡，在煤矸石检测领域有较大的潜力和优越性。

关键词: 煤矸石分拣, 目标检测, 实时性, YOLOv8n

Abstract:

To achieve a balance between detection performance and resource consumption in the gangue sorting domain, an efficient, real-time, lightweight object detection algorithm based on an improved YOLOv8 was proposed, suitable for low-performance detection platforms. This algorithm built on the YOLOv8n architecture and incorporated MobileNetv3 to replace the original backbone network, leveraging its lightweight structure to reduce model parameters and computational load, thereby enhancing detection speed. Additionally, the algorithm integrated the BIFPN module for feature enhancement, which employed multi-scale feature fusion to compensate for the loss of detection accuracy associated with the lightweight network, thus achieving model lightweighting while maintaining detection accuracy. Furthermore, the Inner-CIoU bounding box regression loss function was introduced to balance the training results of images with varying qualities, improving the model’s localization capability and further enhancing detection accuracy and speed. To validate the effectiveness of the proposed algorithm, experiments were conducted to compare it with YOLOv3-tiny, YOLOv5n, YOLOv7, and YOLOv8n on a custom dataset. Experimental results demonstrated that the proposed algorithm exhibited optimal overall detection performance. While maintaining detection accuracy, the model’s parameter count was reduced to 1,188,725, representing a 60.46% decrease compared to YOLOv8n. The computational load was reduced from 8.1 GFLOPs to 2.8 GFLOPs, and the FPS increased from 86.02 Hz to 216.58 Hz. This research indicated that the proposed algorithm is a highly efficient, real-time, lightweight gangue detection method with significant potential and advantages in balancing detection performance and computational resource consumption.

Key words: gangue sorting, object detection, real-time performance, YOLOv8n

中图分类号:

李珍峰, 符世琛, 徐乐, 孟博, 张昕, 秦建军. 基于MBI-YOLOv8的煤矸石目标检测算法研究[J]. 图学学报, 2024, 45(6): 1301-1312.

LI Zhenfeng, FU Shichen, XU Le, MENG Bo, ZHANG Xin, QING Jianjun. Research on gangue target detection algorithm based on MBI-YOLOv8[J]. Journal of Graphics, 2024, 45(6): 1301-1312.

图/表 22

图1 C3结构

Fig. 1 C3 structure

图2 C2f结构

Fig. 2 C2f structure

图3 MBI-YOLOv8模型整体结构图

Fig. 3 Overall structure of the MBI-YOLOv8 model

图4 MobileNetv3模型结构图

Fig. 4 Structure diagram of the MobileNetv3 model

图5 SE注意力模块

Fig. 5 SE attention module

图6 各特征融合网络结构

Fig. 6 Structures of various feature fusion networks ((a) FPN; (b) PANet; (c) BiFPN)

图7 Inner-IoU数学模型

Fig. 7 Inner-IoU mathematical model

图8 数据采集平台 1. 工业灯条；2. 绿色橡胶垫；3. 样本；4. 工业面阵相机；5. 镜头；6. 控制台

Fig. 8 Data collection platform

图9 原始样本((a)煤矸石；(b)煤炭)

Fig. 9 Original samples ((a) Gangue; (b) Coal)

表1 煤矸石数据集明细表

Table 1 Details of the coal gangue datasets

种类	数量			总计
种类	单目标图像	双目标图像	多目标图像	总计
煤	200	260	1 844	2 400
煤矸石	96	260	1 844	2 400

图10 软件采集图像界面

Fig. 10 Software image acquisition interface

图11 Roboflow平台标注后的图像

Fig. 11 Images after annotation on the Roboflow platform

图12 数据增强前后对比图((a)旋转；(b)翻转；(c)亮度调节；(d)添加噪声)

Fig. 12 Comparison of images before and after data augmentation ((a) Rotation; (b) Flipping; (c) Brightness adjustment; (d) Adding noise)

表2 实验环境配置

Table 2 Experimental environment configuration

实验环境		版本型号
硬件配置	CPU	Intel Core i5-13400F
	内存	16 GB
	GPU	NVIDIA GeForce RTX 4060 Ti
	显存	8 GB
	操作环境	Windows 11
软件配置	Python	3.8.19
	Pytorch	1.13.1
	CUDA	11.7
	CUDNN	11.7

表3 损失函数实验结果

Table 3 Experimental results of loss functions

损失函数	mAP0.5/%	mAP0.5:0.95/%	Postprocess per image/ms	FPS/Hz
CIoU	99.5	95.7	0.9	86.02
DIoU	99.5	96.2	0.5	236.27
GIoU	99.5	96.3	0.6	231.62
EIoU	99.5	96.4	1.8	73.17
SIoU	99.5	96.4	2.7	101.91
WIoU v1	99.5	96.5	4.7	85.86
WIoU v2	99.5	96.4	1.2	133.80
WIoU v3	99.5	96.0	1.9	165.41
Inner-CIoU (ratio=0.7)	99.5	96.5	0.7	141.47
Inner-CIoU (ratio=0.75)	99.5	96.3	0.8	128.15
Inner-CIoU (ratio=0.8)	99.5	96.4	0.7	138.73
Inner-DIoU	99.5	96.7	1.0	169.88

表4 消融实验结果

Table 4 Results of ablation experiments

YOLOv8n	Inner-CIoU	MobileNetv3	BiFPN	Precision/%	Recall/%	mAP0.5/%	mAP0.5:0.95/%	模型大小/ MB	模型参数量	运算量/ GFLOPs	FPS/ Hz
√				99.9	100	99.5	95.7	6.3	3 006 038	8.1	86.02
√	√			100.0	100	99.5	96.5	6.3	3 006 038	8.1	141.47
√	√	√		99.7	100	99.4	94.6	2.7	1 188 716	2.8	77.32
√			√	99.9	100	99.5	98.6	6.3	3 006 047	8.1	236.37
√	√	√	√	99.9	100	99.5	96.1	2.6	1 188 725	2.8	216.58

表5 多种模型对比实验结果

Table 5 Comparative experimental results of multiple models

模型	Precision/%	Recall/%	mAP0.5/%	mAP0.5:0.95/%	模型大小/MB	模型参数量	运算量/GFLOPs	FPS/Hz
YOLOv3-tiny	99.3	100	99.5	94.9	24.4	12 128 692	18.9	63.86
YOLOv5	99.6	100	99.5	97.5	5.3	2 503 334	7.1	85.15
YOLOv7	100.0	100	99.8	97.4	74.8	37 201 950	105.1	52.38
YOLOv8n	99.9	100	99.5	95.7	6.3	3 006 038	8.1	86.02
MBI-YOLOv8	99.9	100	99.5	96.1	2.6	1 188 725	2.8	216.58

图13 mAP50:95(B)曲线图

Fig. 13 mAP50:95(B) curve

图14 参数量散点图

Fig. 14 Scatter plot of parameter count

图15 推理时间散点图

Fig. 15 Scatter plot of inference time

图16 煤矸石目标检测结果

Fig. 16 The results of coal gangue object detection

图17 现场验证结果

Fig. 17 Field validation results

参考文献 28

[1]	孙超, 姜琳, 袁广玉. “十四五”期间我国煤炭供需趋势分析[J]. 煤炭工程, 2021, 53(5): 193-196.
	SUN C, JIANG L, YUAN G Y. Trend analysis of China’s coal supply and demand during the 14th Five-Year Plan[J]. Coal Engineering, 2021, 53(5): 193-196. (in Chinese)
[2]	REN Y H, LI R F, GAO H, et al. Research on the evaluation model and suggestions of coal resources mining level in China[J]. IOP Conference Series: Earth and Environmental Science, 2018, 170(2): 022064.
[3]	许永祥, 李申龙, 王国法, 等. 特厚坚硬煤层超大采高综放首采工作面智能化技术[J]. 煤炭科学技术, 2020, 48(7): 186-194.
	XU Y X, LI S L, WANG G F, et al. Intelligent technology of first-mining face of longwall top-coal caving with super large cutting height in extra-thick and hard coal seam[J]. Coal Science and Technology, 2020, 48(7): 186-194. (in Chinese)
[4]	李曼, 段雍, 曹现刚, 等. 煤矸分选机器人图像识别方法和系统[J]. 煤炭学报, 2020, 45(10): 3636-3644.
	LI M, DUAN Y, CAO X G, et al. Image identification method and system for coal and gangue sorting robot[J]. Journal of China Coal Society, 2020, 45(10): 3636-3644. (in Chinese)
[5]	康利, 黄金辉, 刘畅. GDRT型γ射线智能干法分选系统在六家煤矿的应用[J]. 煤炭加工与综合利用, 2017(3): 22-24.
	KANG L, HUANG J H, LIU C. Application of GDRT type gamma-ray intelligent dry sorting system in six coal mines[J]. Coal Processing & Comprehensive Utilization, 2017(3): 22-24. (in Chinese)
[6]	张宁波. 综放开采煤矸自然射线辐射规律及识别研究[D]. 徐州: 中国矿业大学, 2015.
	ZHANG N B. Detection and radiation law of natural gamma ray from coal and roof-rock in the fully mechanized top coal caving mining[D]. Xuzhou: China University of Mining and Technology, 2015. (in Chinese)
[7]	崔斌, 陈林, 亓玉浩, 等. 基于改进YOLO v4的煤矸石识别检测技术研究[J]. 煤炭工程, 2023, 55(12): 161-166.
	CUI B, CHEN L, QI Y H, et al. Coal gangue detection technology based on improved YOLO v4[J]. Coal Engineering, 2023, 55(12): 161-166. (in Chinese)
[8]	吕璐, 程虎, 朱鸿泰, 等. 基于深度学习的目标检测研究与应用综述[J]. 电子与封装, 2022, 22(1): 010307.
	LV L, CHENG H, ZHU H T, et al. Progress of research and application of object detection based on deep learning[J]. Electronics & Packaging, 2022, 22(1): 010307. (in Chinese)
[9]	高昂, 梁兴柱, 夏晨星, 等. 一种改进YOLOv8的密集行人检测算法[J]. 图学学报, 2023, 44(5): 890-898. DOI
	GAO A, LIANG X Z, XIA C X, et al. A dense pedestrian detection algorithm with improved YOLOv8[J]. Journal of Graphics, 2023, 44(5): 890-898. (in Chinese) DOI
[10]	李德永, 王国法, 郭永存, 等. 基于CFS-YOLO算法的复杂工况环境下煤矸图像识别方法[J]. 煤炭科学技术, 2024, 52(6): 226-237.
	LI D Y, WANG G F, GUO Y C, et al. Image recognition method of coal gangue in complex working conditions based on CES-YOLO algorithm[J]. Coal Science and Technology, 2024, 52(6): 226-237. (in Chinese)
[11]	常枫懿, 赵国贞. 基于YOLOv5m改进模型的煤矸识别方法[J]. 煤炭技术, 2023, 42(7): 10-14.
	CHANG F Y, ZHAO G Z. Identification method of coal and Gangue based on YOLOv5m improved model[J]. Coal Technology, 2023, 42(7): 10-14. (in Chinese)
[12]	韦小龙, 王方田, 何东升, 等. 基于CSPNet-YOLOv7目标检测算法的煤矸图像识别模型[J]. 煤炭科学技术, 2024, 52(S1): 238-248.
	WEI X L, WANG F T, HE D S, et al. Coal gangue image recognition model based on CSPNet-YOLOv7 target detection algorithm[J]. Coal Science and Technology, 2024, 52(S1): 238-248. (in Chinese)
[13]	HOWARD A, SANDLER M, CHEN B, et al. Searching for MobileNetV3[C]// 2019 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2019: 1314-1324.
[14]	LI Y H, CHEN Y T, WANG N Y, et al. Scale-aware trident networks for object detection[C]// 2019 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2019: 6053-6062.
[15]	ZHANG H, XU C, ZHANG S J. Inner-IoU: more effective intersection over union loss with auxiliary bounding box[EB/OL]. (2023-11-06) [2024-06-16]. https://arxiv.org/abs/2311.02877.
[16]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2016: 779-788.
[17]	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 7464-7475.
[18]	VARGHESE R, SAMBATH M. YOLOv8: a novel object detection algorithm with enhanced performance and robustness[C]// 2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems. New York: IEEE Press, 2024: 1-6.
[19]	朱强军, 胡斌, 汪慧兰, 等. 基于轻量化YOLOv8s交通标志的检测[J]. 图学学报, 2024, 45(3): 422-432. DOI
	ZHU Q J, HU B, WANG H L, et al. Detection of traffic signs based on lightweight YOLOv8s[J]. Journal of Graphics, 2024, 45(3): 422-432. (in Chinese) DOI
[20]	LIU S, QI L, QIN H F, et al. Path aggregation network for instance segmentation[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 8759-8768.
[21]	翟永杰, 赵晓瑜, 王璐瑶, 等. IDD-YOLOv7: 一种用于输电线路绝缘子多缺陷的轻量化检测方法[J]. 图学学报, 2024, 45(1): 90-101. DOI
	ZHAI Y J, ZHAO X Y, WANG L Y, et al. IDD-YOLOv7: a lightweight method for multiple defect detection of insulators in transmission lines[J]. Journal of Graphics, 2024, 45(1): 90-101. (in Chinese) DOI
[22]	葛道辉, 李洪升, 张亮, 等. 轻量级神经网络架构综述[J]. 软件学报, 2020, 31(9): 2627-2653.
	GE D H, LI H S, ZHANG L, et al. Survey of lightweight neural network[J]. Journal of Software, 2020, 31(9): 2627-2653. (in Chinese)
[23]	HOWARD A G, ZHU M L, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[EB/OL]. (2017-04-17) [2024-06-16]. https://arxiv.org/abs/1704.04861.
[24]	SANDLER M, HOWARD A, ZHU M L, et al. MobileNetV2: inverted residuals and linear bottlenecks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 4510-4520.
[25]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 7132-7141.
[26]	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2017: 936-944.
[27]	ZHENG Z H, WANG P, LIU W, et al. Distance-IoU loss: faster and better learning for bounding box regression[EB/OL]. [2024-05-25]. https://ojs.aaai.org/index.php/AAAI/article/view/6999.
[28]	REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. (2018-04-09) [2024-06-16]. https://arxiv.org/abs/1804.02767.

基于MBI-YOLOv8的煤矸石目标检测算法研究

Research on gangue target detection algorithm based on MBI-YOLOv8

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 22

参考文献 28

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李琼, 考月英, 张莹, 徐沛. 面向无人机航拍图像的目标检测研究综述[J]. 图学学报, 2024, 45(6): 1145-1164.
[2]	闫建红, 冉同霄. 基于YOLOv8的轻量化无人机图像目标检测算法[J]. 图学学报, 2024, 45(6): 1328-1337.
[3]	胡凤阔, 叶兰, 谭显峰, 张钦展, 胡志新, 方清, 王磊, 满孝锋. 一种基于改进YOLOv8的轻量化路面病害检测算法[J]. 图学学报, 2024, 45(5): 892-900.
[4]	姜晓恒, 段金忠, 卢洋, 崔丽莎, 徐明亮. 融合先验知识推理的表面缺陷检测[J]. 图学学报, 2024, 45(5): 957-967.
[5]	章东平, 魏杨悦, 何数技, 徐云超, 胡海苗, 黄文君. 特征融合与层间传递：一种基于Anchor DETR改进的目标检测方法[J]. 图学学报, 2024, 45(5): 968-978.
[6]	李建华, 韩宇, 石开铭, 张可嘉, 郭红领, 方东平, 曹佳明. 施工现场小目标工人检测方法[J]. 图学学报, 2024, 45(5): 1040-1049.
[7]	孙己龙, 刘勇, 周黎伟, 路鑫, 侯小龙, 王亚琼, 王志丰. 基于DCNv2和Transformer Decoder的隧道衬砌裂缝高效检测模型研究[J]. 图学学报, 2024, 45(5): 1050-1061.
[8]	李大湘, 吉展, 刘颖, 唐垚. 改进YOLOv7遥感图像目标检测算法[J]. 图学学报, 2024, 45(4): 650-658.
[9]	胡欣, 常娅姝, 秦皓, 肖剑, 程鸿亮. 基于改进YOLOv8和GMM图像点集匹配的双目测距方法[J]. 图学学报, 2024, 45(4): 714-725.
[10]	牛为华, 郭迅. 基于改进YOLOv8的船舰遥感图像旋转目标检测算法[J]. 图学学报, 2024, 45(4): 726-735.
[11]	曾志超, 徐玥, 王景玉, 叶元龙, 黄志开, 王欢. 基于SOE-YOLO轻量化的水面目标检测算法[J]. 图学学报, 2024, 45(4): 736-744.
[12]	宫永超, 沈旭昆. 一种用于互惠目标检测与实例分割的深层架构[J]. 图学学报, 2024, 45(4): 745-759.
[13]	张相胜, 杨骁. 基于改进YOLOv7-tiny的橡胶密封圈缺陷检测方法[J]. 图学学报, 2024, 45(3): 446-453.
[14]	胡欣, 胡帅, 马丽军, 司利云, 肖剑, 袁晔. 基于融合MBAM与YOLOv5的PCB缺陷检测方法[J]. 图学学报, 2024, 45(1): 47-55.
[15]	魏陈浩, 杨睿, 刘振丙, 蓝如师, 孙希延, 罗笑南. 具有双层路由注意力的YOLOv8道路场景目标检测方法[J]. 图学学报, 2023, 44(6): 1104-1111.