CN111652873A - Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning - Google Patents

Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning Download PDF

Info

Publication number
CN111652873A
CN111652873A CN202010498373.1A CN202010498373A CN111652873A CN 111652873 A CN111652873 A CN 111652873A CN 202010498373 A CN202010498373 A CN 202010498373A CN 111652873 A CN111652873 A CN 111652873A
Authority
CN
China
Prior art keywords
iron
permanent magnet
discarding
detection
conveyor belt
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010498373.1A
Other languages
Chinese (zh)
Inventor
李现国
刘晓
苗长云
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin Polytechnic University
Original Assignee
Tianjin Polytechnic University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tianjin Polytechnic University filed Critical Tianjin Polytechnic University
Priority to CN202010498373.1A priority Critical patent/CN111652873A/en
Publication of CN111652873A publication Critical patent/CN111652873A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Abstract

The invention discloses a permanent magnet self-discharging iron remover scrap conveyor belt fracture detection method based on transfer learning, belongs to the field of nondestructive detection, and solves the problems that an existing permanent magnet self-discharging iron remover scrap conveyor belt fracture detection method is not accurate, a detection device is easy to damage, and installation is complex. Firstly, setting an RIO area through a template matching technology in OpenCV, and performing matching segmentation on an iron-discarding conveyer belt of an iron remover to enhance picture characteristics; then, carrying out feature extraction and classification on the conveyer belt pictures by using a transfer learning method of the Resnet18 network; and finally, deploying the ResNet18 network to a Jetson TX2 embedded development platform, forming a breakage detection system of the iron-discarding conveying belt of the permanent magnet self-discharging iron remover by using Jetson TX2, and immediately alarming when the breakage of the iron-discarding conveying belt is detected. The method and the system can be effectively used in the working environment of the permanent magnet self-unloading iron remover, realize the automatic analysis of the field monitoring video, have high accuracy of fault detection and are simple and convenient to arrange.

Description

Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning
Technical Field
The invention belongs to the field of nondestructive testing, and particularly relates to a permanent magnet self-unloading iron remover scrap conveyor belt fracture detection method based on transfer learning.
Background
The permanent magnet self-unloading iron remover can remove ferromagnetic impurities mixed in nonmagnetic materials, generally comprises a permanent magnet core, a discarded iron conveying belt, a speed reducing motor, a frame, a roller and the like, is used together with a belt conveyor, and is widely applied to occasions such as mines, coal preparation plants, ports and the like. The discarded iron conveyer belt is a consumable in a permanent magnet self-unloading iron remover, and is broken frequently due to various reasons, and if the discarded iron conveyer belt cannot be timely treated, accidents such as equipment damage, material piling, production stop and the like can be caused.
The permanent magnet self-discharging iron remover generally hangs above a material conveyer belt to work, the size is relatively small, and the existing method for detecting the fracture of the iron-discarding conveyer belt is not suitable for the fracture detection of the permanent magnet self-discharging iron remover in the aspects of installation and arrangement and long-term use. The current deep learning automatic feature extraction technology is widely applied by people, is a research hotspot in the aspect of video detection, can fully explore the potential of the existing monitoring facilities of the permanent magnet self-unloading iron remover by applying the deep learning technology in the aspect of conveyer belt detection, can timely reduce the occurrence of accidents, and ensures the normal operation of operation.
Disclosure of Invention
The invention aims to provide a permanent magnet self-discharging iron remover conveyor belt fracture detection method based on transfer learning, and solves the problems that an existing permanent magnet self-discharging iron remover iron-discarding conveyor belt fracture detection method is not accurate, and a detection device is easy to damage and complex to install.
In order to solve the problems, the technical scheme adopted by the invention is as follows:
a permanent magnet self-unloading iron remover iron scrap conveyor belt fracture detection method based on transfer learning specifically comprises a training stage and an application stage, wherein the training stage specifically comprises the following operation steps:
step 1, applying a ResNet18 network structure and parameters for image classification on an ImgeNet data set to a training stage of a ResNet18 network model for detecting breakage of a discarded iron conveyor belt by adopting a transfer learning method;
step 2, collecting a monitoring video in a working scene of the permanent magnet self-unloading iron remover, converting the video into a picture, setting a Region of Interest (RIO) by utilizing a template matching technology in OpenCV according to the actual requirement of the breakage detection of the iron-discarding conveyer belt, and performing matching segmentation on the image of the iron-discarding conveyer belt of the iron remover;
step 3, expanding the divided pictures by methods of rotation, cutting, brightness conversion and the like into a training set and a test set;
step 4, training a break detection ResNet18 network of the iron scrap conveyor belt by using a training set, and testing the actual effect of the break detection model of the iron scrap conveyor belt of the permanent magnet self-discharging iron remover obtained after training by using a test set;
and 5, analyzing whether the actual requirement of the fracture detection of the iron-discarding conveying belt of the permanent magnet self-unloading iron remover is met or not according to the tested actual effect, performing targeted adjustment and improvement, executing the step 4 again until the actual requirement of the fracture detection of the conveying belt of the permanent magnet self-unloading iron remover is met, and constructing a final fracture detection model of the iron-discarding conveying belt of the permanent magnet self-unloading iron remover.
The ResNet18 network structure and parameters used for image classification on the ImgeNet data set are applied by adopting the transfer learning method and the data set is expanded by utilizing image transformation, so that the problem of poor learning effect caused by insufficient sample quantity in the process of feature extraction of the ResNet18 network for detecting the breakage of the iron-saving conveyor belt is solved.
The method is characterized in that a RIO area is set by utilizing an OpenCV template matching technology to perform matching segmentation on the iron scrap conveyor belt of the iron remover, and aims to reduce irrelevant information interference and enhance image characteristics when a ResNet18 network for detecting the breakage of the iron scrap conveyor belt is used for feature extraction.
In the application stage, an England Jetson TX2 embedded development platform and a monitoring camera are mainly used for building a fracture detection system of the iron-discarding conveyer belt of the permanent magnet self-unloading iron remover, and whether the iron-discarding conveyer belt in a monitoring area is fractured or not is monitored in real time, and the method comprises the following operation steps:
step 1, converting a ResNet18 network model for detecting the breakage of the abandoned iron conveyor belt into a model identified by a C + + interface by TorchScript;
step 2, redeploying a ResNet18 network model based on C + + on Jetson TX2 by utilizing Libtrch;
step 3, monitoring the conveyer belt in the monitoring area in real time by using a discarded iron conveyer belt fracture detection system of the permanent magnet self-unloading iron remover based on Jetson TX 2;
and 4, when the conveyor belt is detected to be broken, the Jetson TX2 immediately sends a signal to an alarm to give an alarm.
The TorchScript adopted in the steps is used for converting the iron scrap conveyor belt breakage detection ResNet18 network model into a C + + interface identification model, and the purpose is to control GPIO on Jetson TX2 when the iron scrap conveyor belt breakage detection ResNet18 network is deployed on Jetson TX2 for application.
The invention has the beneficial effects that: (1) the invention can be effectively applied to the working environment of the permanent magnet self-unloading iron remover; (2) according to the method, the process of traditional manual target feature extraction is replaced by a deep learning technology, and the image features are obtained by using a convolutional neural network, so that the accuracy and the real-time performance of the detection of the fracture of the conveying belt are improved, and the actual requirement of the detection of the fracture of the conveying belt is met; (3) the invention can be directly deployed in the monitoring server terminal of an enterprise, does not need to add extra equipment by utilizing the existing monitoring facility, and has low application cost. (4) The method for detecting the breakage of the conveying belt is low in probability of being damaged by the external environment, and the service life is greatly prolonged.
Drawings
FIG. 1 is an overall flow diagram of the method of the present invention.
FIG. 2 is a graph of the template matching results of the method of the present invention.
FIG. 3 is a graph showing the results of the detection by the method of the present invention
Detailed Description
Specific embodiments of the present disclosure are described below with reference to the accompanying drawings.
The overall flow of the proposed method of the present invention is shown in fig. 1. The invention aims to provide a permanent magnet self-discharging iron remover scrap conveyor belt fracture detection method based on transfer learning, and solves the problems that an existing permanent magnet self-discharging iron remover scrap conveyor belt fracture detection method is not accurate, a detection device is easy to damage, and installation is complex.
The method specifically comprises a training stage and an application stage, wherein the training stage comprises the following operation steps:
step 1, applying a ResNet18 network structure and parameters for image classification on an ImgeNet data set to a training stage of a ResNet18 network model for detecting breakage of a discarded iron conveyor belt by adopting a transfer learning method;
step 2, collecting a monitoring video in a working scene of the permanent magnet self-unloading iron remover, converting the video into a picture, setting a Region of Interest (RIO) by utilizing a template matching technology in OpenCV according to the actual requirement of the breakage detection of the iron-discarding conveyer belt, and performing matching segmentation on the image of the iron-discarding conveyer belt of the iron remover;
step 3, expanding the divided pictures by methods of rotation, cutting, brightness conversion and the like into a training set and a test set;
step 4, training a break detection ResNet18 network of the iron scrap conveyor belt by using a training set, and testing the actual effect of the break detection model of the iron scrap conveyor belt of the permanent magnet self-discharging iron remover obtained after training by using a test set;
and 5, analyzing whether the actual requirement of the fracture detection of the iron-discarding conveying belt of the permanent magnet self-unloading iron remover is met or not according to the tested actual effect, performing targeted adjustment and improvement, executing the step 4 again until the actual requirement of the fracture detection of the conveying belt of the permanent magnet self-unloading iron remover is met, and constructing a final fracture detection model of the iron-discarding conveying belt of the permanent magnet self-unloading iron remover.
Specifically, in the training stage, the RIO area is set by using a template matching technology in OpenCV to perform matching segmentation on the iron remover conveyor belt, and the method is characterized in that when the characteristics of the ResNet18 network are extracted by detecting the breakage of the iron-removing conveyor belt, the interference of irrelevant information is reduced, and the image characteristics are enhanced. The template matching result of the method is shown in fig. 2, the working environment of the permanent magnet self-discharging iron remover is complex, the target image in the monitored image is small, and the accuracy of subsequent work can be ensured only by amplifying the target and enhancing the characteristics.
Specifically, in the training phase, the method adopting the transfer learning in step 1 and step 3 of the invention applies the structure and parameters of the ResNet18 network for image classification on the ImgeNet data set and expands the data set by using image transformation, and is characterized by being used for solving the problem of poor learning effect caused by insufficient sample quantity during feature extraction of the ResNet18 network for detecting the breakage of the iron-discarding conveyor belt.
The detection result of the method is shown in fig. 3, the method can be effectively applied to the working field of the permanent magnet self-discharging iron remover, and can also detect when the fracture of the conveying belt is not obvious and only cracks or conveying belt looseness and other abnormal conditions exist in the picture, so that the accuracy of the method is shown.
The application stage mainly utilizes the control of enterprise to establish permanent magnetism self-discharging de-ironing separator conveyer belt fracture detecting system, and whether real-time supervision monitoring area conveyer belt splits, when detecting the fracture, triggers the alarm and reports to the police, handles in advance the conveyer belt, prevents that the production length from pausing, prevents serious accidents such as machine damage, coal piling, wherein contains following operating procedure:
step 1, converting a ResNet18 network model for detecting the breakage of the abandoned iron conveyor belt into a model identified by a C + + interface by TorchScript;
step 2, redeploying a ResNet18 network model based on C + + on Jetson TX2 by utilizing Libtrch;
step 3, monitoring the conveyer belt in the monitoring area in real time by using a discarded iron conveyer belt fracture detection system of the permanent magnet self-unloading iron remover based on Jetson TX 2;
and 4, when the conveyor belt is detected to be broken, the Jetson TX2 immediately sends a signal to an alarm to give an alarm.
Specifically, in the application stage step 1 of the invention, TorchScript is adopted to convert the iron scrap conveyor belt breakage detection ResNet18 network model into a C + + interface identification model, and the method is characterized in that when the iron scrap conveyor belt breakage detection ResNet18 network is deployed on Jetson TX2 for application, the method is used for controlling GPIO on Jetson TX2 and controlling GPIO on a Jetson TX2 development board to operate in C/C + + language, so that the Python language ResNet18 model needs to be converted into the C + + language ResNet18 model.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the true spirit and scope of the invention.

Claims (4)

1. The utility model provides a permanent magnetism self-discharging de-ironing separator abandons iron conveyer belt fracture detection method based on migration learning, contains training phase and application phase, its characterized in that:
in the training stage, matching and segmenting the images of the iron-removing device iron-discarding conveyor belt by adopting a template matching technology in OpenCV, constructing a permanent magnet self-unloading iron-removing device iron-discarding conveyor belt fracture image data set, and training an iron-discarding conveyor belt fracture detection ResNet18 network model by utilizing a transfer learning method;
in the application stage, TorchScript is adopted to convert a ResNet18 network model for detecting the breakage of the abandoned iron conveying belt into a C + + based ResNet18 network model, Jetson TX2 is utilized to form a breakage detection system for the abandoned iron conveying belt of the permanent magnet self-unloading iron remover, and an alarm is given immediately when the breakage of the abandoned iron conveying belt is detected.
2. The transfer learning method as claimed in claim 1, wherein the ResNet18 network structure and parameters used for image classification on the ImgeNet dataset are applied to the training phase of the ResNet18 network model for iron-rejection conveyor belt breakage detection.
3. The method for matching and segmenting the iron-discarding conveyer belt image of the iron remover by using the template matching technique in OpenCV as claimed in claim 1, wherein the interference of irrelevant information can be reduced, and the image characteristics can be enhanced.
4. The use of TorchScript to convert the discarded iron conveyor belt break detection ResNet18 network model to a C + + based ResNet18 network model as claimed in claim 1, wherein the Python language model is converted to a C + + language model.
CN202010498373.1A 2020-06-04 2020-06-04 Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning Pending CN111652873A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010498373.1A CN111652873A (en) 2020-06-04 2020-06-04 Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010498373.1A CN111652873A (en) 2020-06-04 2020-06-04 Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning

Publications (1)

Publication Number Publication Date
CN111652873A true CN111652873A (en) 2020-09-11

Family

ID=72344657

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010498373.1A Pending CN111652873A (en) 2020-06-04 2020-06-04 Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning

Country Status (1)

Country Link
CN (1) CN111652873A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113221688A (en) * 2021-04-28 2021-08-06 南京南瑞继保电气有限公司 Disconnecting link state identification method and device and storage medium
CN113682762A (en) * 2021-08-27 2021-11-23 中国矿业大学 Belt tearing detection method and system based on machine vision and deep learning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108510488A (en) * 2018-03-30 2018-09-07 安徽理工大学 Four kinds of damage detecting methods of conveyer belt based on residual error network
CN109740598A (en) * 2018-12-29 2019-05-10 中国铁道科学研究院集团有限公司基础设施检测研究所 Object localization method and device under structuring scene
CN109879005A (en) * 2019-04-15 2019-06-14 天津美腾科技有限公司 Device for detecting belt tearing and method
CN111079511A (en) * 2019-10-25 2020-04-28 湖北富瑞尔科技有限公司 Document automatic classification and optical character recognition method and system based on deep learning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108510488A (en) * 2018-03-30 2018-09-07 安徽理工大学 Four kinds of damage detecting methods of conveyer belt based on residual error network
CN109740598A (en) * 2018-12-29 2019-05-10 中国铁道科学研究院集团有限公司基础设施检测研究所 Object localization method and device under structuring scene
CN109879005A (en) * 2019-04-15 2019-06-14 天津美腾科技有限公司 Device for detecting belt tearing and method
CN111079511A (en) * 2019-10-25 2020-04-28 湖北富瑞尔科技有限公司 Document automatic classification and optical character recognition method and system based on deep learning

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张泰源: "TorchScript 如何实现Python -> C++ 代码转换", 《知乎》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113221688A (en) * 2021-04-28 2021-08-06 南京南瑞继保电气有限公司 Disconnecting link state identification method and device and storage medium
CN113221688B (en) * 2021-04-28 2024-03-19 南京南瑞继保电气有限公司 Knife switch state identification method, device and storage medium
CN113682762A (en) * 2021-08-27 2021-11-23 中国矿业大学 Belt tearing detection method and system based on machine vision and deep learning

Similar Documents

Publication Publication Date Title
CN109879005B (en) Belt tearing detection device and method
CN104535356B (en) A kind of roller steel wire rope rope fault monitoring method and system based on machine vision
CN110127479B (en) Elevator door switch abnormity detection method based on video analysis
CN109631848B (en) Transmission line foreign matter intrusion detection system and detection method
CN107133592B (en) Human body target feature detection algorithm for power substation by fusing infrared thermal imaging and visible light imaging technologies
CN108280855A (en) A kind of insulator breakdown detection method based on Fast R-CNN
CN111652873A (en) Permanent magnet self-unloading iron remover iron-discarding conveyer belt fracture detection method based on transfer learning
CN111127448B (en) Method for detecting air spring fault based on isolated forest
UA88919C2 (en) Device and method for collection of diagnostic data
CN100469138C (en) Power transformer draught fan state recognizing method based on video monitoring and image recognition
CN113128555A (en) Method for detecting abnormality of train brake pad part
CN104680787B (en) A kind of congestion in road detection method
CN107124577A (en) A kind of real-time alarm system for guarding against theft based on moving object detection
Wang et al. Visual defect detection for substation equipment based on joint inspection data of camera and robot
CN116231504A (en) Remote intelligent inspection method, device and system for booster station
CN110969081B (en) Power transmission line external force damage detection method based on KL divergence of multi-module division
CN103489179A (en) Method for eliminating cement crack interference in airport runway impurity detection
CN115049623A (en) Device for analyzing pantograph contour through visual segmentation
CN110991286A (en) Automatic electric shovel bucket defect identification device and method based on convolutional neural network
CN105590329B (en) The Intelligent Measurement algorithm of moving target in a kind of oil gas field video monitoring
CN114511551A (en) Ground damage identification system based on machine vision
CN110728679A (en) Intelligent monitoring method for water leakage of pipe joint
Liu et al. Detection of railway tunnel lining based on adaptive background learning
CN112714284A (en) Power equipment detection method and device and mobile terminal
CN110264532A (en) Underground coal mine driving face sky top job identifying method based on intelligent video identification technology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200911

WD01 Invention patent application deemed withdrawn after publication