基于红外图像的矿石传送带托辊异常检测

doi:10.11872/j.issn.1005-2518.2023.01.099

Abstract

Abstract:

With the development of intelligent construction of mine，the detection of mine equipment is becoming more and more intelligent.The ore conveyor belt is one of the important production equipment in the mine，and the abnormal detection of the conveyor roller is one of the important contents of mine safety.At present，most of the ore conveyor roller inspection is manual inspection，and depends on the personal experience of the inspector to judge the working status of the roller，which will lead to problems such as the damage of the roller is not found in time.Therefore，it is urgent to study a more objective，intelligent and efficient method for abnormal detection of rollers.In order to solve the problems of low efficiency and poor real-time performance of the traditional ore conveyor roller anomaly detection method，an optimization model of ore conveyor roller anomaly detection based on infrared image recognition was proposed.The infrared image of the roller was collected on the spot and the infrared image data set of the roller was expanded by using label smoothing and Mosaic data enhancement processing to prevent the overfitting of the detection model and reduce the training cost of the model.In the feature extraction module，it was proposed to use GhostNet backbone feature extraction network，which can effectively reduce the image redundancy produced by feature extraction，accelerate the learning speed of the model，and further optimize the backbone feature extraction network through LeakyReLU activation function to improve the learning ability of the model.In the feature fusion module，multi-dimensional feature fusion was realized through the feature pyramid structure and the bottom-up feature fusion layer，and the SPP-Net module was used to optimize the PaNet feature fusion network to increase the effective receptive field of the model.And through the depth separable convolution block to simplify the model structure，reduce the amount of calculation and the number of parameters of the model.The experimental results show that，compared with the mainstream detection model，the detection model can more effectively identify rollers and distinguish between normal and abnormal rollers.In the actual detection，the detection accuracy of the idlers is 96.2%，the recall rate is 95.9%，and the average detection accuracy is 94.9%，in which the accuracy of abnormal rollers is 99.6%，the accuracy of normal rollers is 90.2%，the detection speed is 39.2 FPS，and the number of model parameters is only 1.1×10⁷.The method provides a guarantee for accurate and efficient inspection of mine conveyor rollers.

Key words: machine vision, infrared image recognition, deep learning network, network structure optimization, roller detection, anomaly detection

CLC Number:

X986

Shunling RUAN,Yankang RUAN,Caiwu LU,Qinghua GU. Detection of Ore Conveyer Roller Based on Infrared Image[J].Gold Science and Technology, 2023, 31(1): 123-132.

Figures/Tables 15

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Fig.9

References 0

	Bochkovskiy A， Wang C Y， Liao H Y M，2020.Yolov4：Optimal speed and accuracy of object detection［J］.arXiv，2004.10934.
	Cao Guanqiang，2020.Fault detection method for belt conveyor roller［J］.Industry and Mine Automation，46（6）：81-86.
	Guo Qinghua，2018.Research on roller fault identification algorithm of belt conveyor system based on fiber temperature measurement technology［J］.Coal Mine Machinery，39（8）：157-160.
	Han K， Wang Y， Tian Q，et al，2020.Ghostnet：More features from cheap operations［C］//IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Seattle：IEEE.
	Hao Hongtao， Ni Fanfan， Ding Wenjie，2019.Fault diagnosis method of rollers based on sound signals ［J］.Noise and Vibration Control，39（3）：187-192.
	He K， Zhang X， Ren S，et al，2015.Spatial pyramid pooling in deep convolutional networks for visual recognition［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence，37（9）：1904-1916.
	Ioffe S， Szegedy C，2015.Batch normalization：Accelerating deep network training by reducing internal covariate shift［J］.arXiv，1502.03167.
	Lin T Y， Goyal P， Girshick R，et al，2017.Focal loss for dense object detection［C］//IEEE International Conference on Computer Vision（ICCV）.Venice：IEEE.
	Liu Fen，2020.Fault diagnosis of roller of belt conveyor based on big data technology［J］.Coal Mine Machinery，41（8）：177-179.
	Liu S， Qi L， Qin H，et al，2018.Path aggregation network for instance segmentation［C］//IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Salt Lake City：IEEE.
	Liu W， Anguelov D， Erhan D，et al，2016.Ssd：Single shot multibox detector［C］//European Conference on Computer Vision（ECCV）.Amsterdam：IEEE.
	Ravikumar S， Kanagasabapathy H， Muralidharan V，2019.Fault diagnosis of self-aligning troughing rollers in belt conveyor system using k-star algorithm［J］.Measurement，133：341-349.
	Redmon J， Divvala S， Girshick R，et al，2016.You only look once：Unified，real-time object detection［C］// IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Las Vegas：IEEE.
	Ren S， He K， Girshick R，et al，2017.Faster r-cnn：Towards real-time object detection with region proposal networks［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence，39（6）：1137-1149.
	Song Tianxiang， Yang Mingjin， Yang Linshun，et al，2019.Fault diagnosis for roller based on spectral clustering analysis［J］.Electronic Measurement Technology，42（5）：144-150.
	Su Hui， Niu Linkai， Zhang Kun，2018.Design of roller fault monitoring system based on ZigBee wireless sensor network［J］.Coal Engineering，50（7）：14-17.
	Sun Wei， Diao Dongmei，2016.Roller fault detection of belt convey or based on φ-OTDR technology ［J］.Industry and Mine Automation，42（8）：9-12.
	Szegedy C， Vanhoucke V， Ioffe S，et al，2016.Rethinking the inception architecture for computer vision［C］//IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Las Vegas：IEEE.
	Wang C Y， Liao H Y M， Wu Y H，et al，2020.CSPNet：A new backbone that can enhance learning capability of CNN［C］//IEEE Conference on Computer Vision and Pattern Recognition（CVPR）.Seattle：IEEE.
	Xie Miao， Zhu Zhen， Lu Jinnan，2020.Research on detection method of roller jam based on infrared image processing technology［J］.Machine Design and Research，36（5）：152-157.
	Yang M， Zhou W， Song T，2020.Audio-based fault diagnosis for belt conveyor rollers［J］.Neurocomputing，397：447-456.
	Yi Xin， Yang Mingjin， Yang Linshun，et al，2020.The KNN and SVM-based 2-level comprehensive health indicators diagnosis method for detecting the failure of belt conveyor’s idlers［J］.Coal Preparation Technology，（5）：94-102.
	Zhang X， Wan S， He Y，et al，2021.Teager energy spectral kurtosis of wavelet packet transform and its application in locating the sound source of fault bearing of belt conveyor［J］.Measurement，173：108367.
	Zheng Z， Wang P， Ren D，et al，2021.Enhancing geometric factors in model learning and inference for object detection and instance segmentation［J］.Transactions on Cybernetics，52（8）：8574-8586.
	曹贯强，2020.带式输送机托辊故障检测方法［J］.工矿自动化，46（6）：81-86.
	郭清华，2018.基于光纤测温技术的带式输送机托辊故障识别算法研究［J］.煤矿机械，39（8）：157-160.
	郝洪涛，倪凡凡，丁文捷，2019.基于声音信号的托辊故障诊断方法［J］.噪声与振动控制，39（3）：187-192.
	刘芬，2020.基于大数据技术的带式输送机托辊故障诊断［J］.煤矿机械，41（8）：177-179.
	宋天祥，杨明锦，杨林顺，等，2019.基于谱聚类分析的托辊故障诊断［J］.电子测量技术，42（5）：144-150.
	苏辉，牛蔺楷，张琨，2018.基于ZigBee无线传感网络的托辊卡死故障监测系统设计［J］.煤炭工程， 50（7）：14-17.
	孙维，刁冬梅，2016.基于φ-OTDR技术的带式输送机托辊故障检测［J］.工矿自动化，42（8）：9-12.
	谢苗，朱振，卢进南，2020.基于红外图像处理技术的托辊卡阻检测方法［J］.机械设计与研究，36（5）：152-157.
	伊鑫，杨明锦，杨林顺，等，2020.基于KNN与SVM两级综合健康指标的托辊故障诊断方法［J］.选煤技术，（5）：94-102.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

Comments

Recommended 0

No Suggested Reading articles found!

骨干网络	精准度/%	召回率/%	MAP/%	参数量 /（×10⁷）	速度 /FPS
CSPDarkNet53	90.5	91.6	89.4	3.6	31.7
MobileNetv1	91.0	91.6	85.9	1.3	60.7
MobileNetv2	93.9	95.3	93.6	1.1	48.4
MobileNetv3	94.1	94.4	91.8	1.2	44.3
GhostNet	95.6	95.3	93.8	1.1	36.4

激活函数	精准度 /%	召回率 /%	AP1 /%	AP2 /%	MAP /%	速度 /FPS
ReLU	95.6	95.3	97.7	89.8	93.8	36.4
Tanh	93.8	94.1	98.8	87.4	93.1	37.9
Sigmoid	94.7	94.7	99.4	81.2	90.3	40.0
RReLU	96.9	96.9	99.8	90.0	94.9	37.9
LeakyReLU	96.2	95.9	99.6	90.2	94.9	39.2

标签平滑	MOSAIC增强	精准度/%	召回率/%	AP1/%	AP2/%	MAP/%
		90.4	91.3	97.8	84.3	91.1
√		93.1	92.3	97.7	88.2	92.9
	√	94.3	94.0	99.0	84.9	92.0
√	√	96.2	95.9	99.6	90.2	94.9

模型	精准度 /%	召回率 /%	AP1 /%	AP2 /%	MAP /%	参数量 /（×10⁷）	速度 /FPS
Faster R-CNN	60.8	87.3	99.9	79.7	84.8	14	28.7
Retomamet	86.0	86.0	97.0	70.3	83.4	3.7	28.1
SSD	86.9	86.9	95.1	62.2	78.7	2.6	99.8
YOLOV3	90.4	90.7	96.0	75.0	85.5	6.2	26.2
YOLOV4	95.0	95.0	99.3	85.0	92.1	6.4	35.3
本文模型	96.2	95.9	99.6	90.2	94.9	1.1	39.2

Detection of Ore Conveyer Roller Based on Infrared Image

RichHTML

PDF (PC)

Abstract

Cite this article

share this article

Figures/Tables 15

References 0

Related Articles 1

Metrics

Comments

Recommended 0