CN114187237A - Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium - Google Patents
Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium Download PDFInfo
- Publication number
- CN114187237A CN114187237A CN202111341853.8A CN202111341853A CN114187237A CN 114187237 A CN114187237 A CN 114187237A CN 202111341853 A CN202111341853 A CN 202111341853A CN 114187237 A CN114187237 A CN 114187237A
- Authority
- CN
- China
- Prior art keywords
- image data
- neural network
- image
- tearing
- deviation
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000003860 storage Methods 0.000 title claims abstract description 18
- 230000002159 abnormal effect Effects 0.000 claims abstract description 91
- 238000003062 neural network model Methods 0.000 claims abstract description 81
- 238000012706 support-vector machine Methods 0.000 claims abstract description 74
- 238000012549 training Methods 0.000 claims abstract description 69
- 239000013598 vector Substances 0.000 claims abstract description 50
- 238000001514 detection method Methods 0.000 claims description 41
- 230000006870 function Effects 0.000 claims description 30
- 238000012545 processing Methods 0.000 claims description 24
- 238000004590 computer program Methods 0.000 claims description 12
- 230000004044 response Effects 0.000 claims description 7
- 230000000694 effects Effects 0.000 abstract description 2
- 239000003245 coal Substances 0.000 description 8
- 238000010606 normalization Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 238000012360 testing method Methods 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
-
- 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
- G06T2207/10016—Video; Image sequence
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a method, a device, equipment and a storage medium for detecting tearing and deviation of a conveying belt, wherein the method comprises the steps of acquiring video data of the conveying belt and constructing an image database; inputting data of an image database into a deep neural network model for training, and outputting abnormal image data characteristics; training a hyperparametric combination which enables the highest classification accuracy rate based on the support vector machine model to obtain the support vector machine model under the hyperparametric combination; enabling the deep neural network model and the support vector machine model to detect video data of the conveying belt; when a preset condition is met, the deep neural network model outputs a characteristic vector z, and the current video data of the conveying belt is judged to be abnormal; the support vector machine model outputs an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z. The problem of the lower rate of accuracy of discernment of conveyer belt tear or off tracking scene is solved. The method and the device have the effect of improving the identification accuracy.
Description
Technical Field
The application relates to the technical field of industrial automation, in particular to a method, a device, equipment and a storage medium for detecting tearing and deviation of a conveying belt.
Background
At present, a conveying belt is an important tool for carrying power coal in a power plant and is a necessary condition for safe and reliable operation of the power plant. However, in the long-term coal transportation process, due to the reasons of load alternation, fatigue damage and the like, the conveying belt has the problems of tearing, deviation and the like, and the normal operation of a factory is influenced under serious conditions, even the operation safety of workers is threatened.
Specifically, the tearing of the conveying belt means that the tearing phenomenon of the conveying belt caused by friction damage, metal blockage, external pressure, high-speed operation and other reasons in the process of conveying coal of the conveying belt can cause the leakage of power coal and the continuous expansion of cracks, and serious accidents such as machine failure, production stagnation, casualties and the like can be caused in serious cases, so that a great threat can be formed on the safe operation of a coal mine conveying system. The existing conveyor belt tearing detection methods mainly comprise manual detection, sensor detection and tearing detection methods based on image processing, but the manual detection method cannot find tearing in time and stop transportation; the sensor detection is easy to damage, so that the detection is invalid, and the accuracy, the stability and the reliability are low; the tearing detection method based on image processing can only detect tearing under a single scene generally, early prediction of tearing trend is lacked, and universality and reliability are low.
The deviation of the conveying belt means that the distance between the actual central line and the theoretical central line of the conveying belt is too large in the running process of the conveying belt, so that the power coal is gathered at the tail end to form side leakage, the environmental pollution is caused, the cleaning workload is increased, the production efficiency and the service life of the conveying belt are influenced, and the safety production accident is seriously caused. The existing conveying belt deviation detection method mainly comprises manual detection, photoelectric detection and visual image detection methods, but the manual detection method has low efficiency and cannot find deviation in time and carry out shutdown and overhaul; photoelectric detection equipment is expensive in manufacturing cost and difficult to popularize; the visual image detection method depends on a mathematical model and is greatly influenced by a complex environment, so that the detection accuracy is low.
For the related technologies, the inventor thinks that the defect that the recognition accuracy of the existing tearing or deviation scene of the conveying belt is low exists.
Disclosure of Invention
In order to improve the identification accuracy of a tearing or deviation scene of a conveying belt, the application provides a tearing and deviation detection method, a tearing and deviation detection device, equipment and a storage medium for the conveying belt.
In a first aspect, the application provides a method for detecting tearing and deviation of a conveying belt, which has the characteristic of accurately detecting the tearing and deviation of the conveying belt.
The application is realized by the following technical scheme:
a method for detecting tearing and deviation of a conveying belt comprises the following steps:
acquiring video data of a conveyer belt, and constructing an image database;
inputting data of the image database into a deep neural network model for training, and outputting abnormal image data characteristics;
inputting a hyper-parameter combination which enables the classification accuracy to be highest based on a support vector machine model, and enabling the marked tearing image data characteristics and the marked deviation image data characteristics of the abnormal image data characteristics to the support vector machine model for training to obtain the support vector machine model under the hyper-parameter combination;
enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect video data of a conveying belt;
when a preset condition is met, the deep neural network model outputs a characteristic vector z, and the current video data of the conveying belt is judged to be abnormal;
and outputting an abnormal result of tearing or deviation of the conveying belt by the support vector machine model based on the characteristic vector z.
The present application may be further configured in a preferred example to: the preset condition comprises that the distance R between the characteristic vector z output by the deep neural network model and a preset central vector c is greater than a preset threshold value R.
The present application may be further configured in a preferred example to: the step of inputting the data of the image database into a deep neural network model for training comprises the following steps:
presetting a contrast loss function, wherein the contrast loss function comprises,
wherein, I1Representing a normal image data set, I2Representing an abnormal image data set, I representing a normal image data set I1Or abnormal image data set I2One image of ZiFeature vector representing image I, I1\ I denotes a normal image data set I1Is not in a certain image iAll other image data sets, j representing the normal image data set I1Or abnormal image data set I2One image other than the certain image i, ZjFeature vector, I, representing image j2\ I represents an abnormal image data set I2T is a normal image data set I1Or abnormal image data set I2An image of ZtIs the feature vector of the image t; and pre-training a deep neural network model by using the contrast loss function.
The present application may be further configured in a preferred example to: after the deep neural network model is pre-trained, the method further comprises the following steps:
presetting a training loss function, the training loss function comprising,
wherein z isiAn output vector of an input image i on the deep neural network model is obtained, and c is an output vector mean value obtained by pre-training;
and optimizing the pre-trained deep neural network model by using the training loss function.
The present application may be further configured in a preferred example to: the step of obtaining the hyper-parameter combination of the support vector machine model which enables the classification accuracy rate to be highest comprises the following steps:
presetting hyper-parameter combinations including kernel functions, punishment parameters and relaxation vectors of a plurality of groups of support vector machine models, training the support vector machine models based on the tearing image data characteristics and the deviation image data characteristics, and obtaining classification accuracy corresponding to each group of hyper-parameter combinations;
constructing a response relation between the hyper-parameter combination and the classification accuracy;
and acquiring the hyper-parameter combination with the highest classification accuracy by combining the gain expectation.
The present application may be further configured in a preferred example to: before inputting the data of the image database into the deep neural network model for training, the method further comprises the following steps:
performing data enhancement processing on torn image data and off-tracking image data of the abnormal category image data of the divided image database;
and inputting the normal category image data of the image database, the tearing image data and the deviation image data which are subjected to data enhancement processing into a depth neural network model.
The present application may be further configured in a preferred example to: before the step of inputting the normal category image data of the image database, the tearing image data and the deviation image data which are subjected to data enhancement processing into a depth neural network model, the method further comprises the following steps:
and normalizing the torn image data and the off-tracking image data subjected to data enhancement processing and the normal type image data, and inputting the normalized data into a deep neural network model.
In a second aspect, the application provides a conveyer belt tearing and deviation detecting device, which has the characteristic of accurately detecting the tearing and deviation of the conveyer belt.
The application is realized by the following technical scheme:
a conveyor belt tear and run deviation detection device comprising:
the data acquisition module is used for acquiring video data of the conveyer belt and constructing an image database;
the first training module is used for inputting the data of the image database into a deep neural network model for training and outputting abnormal image data characteristics;
the second training module is used for enabling a hyper-parameter combination with the highest classification accuracy to be achieved based on a support vector machine model, enabling the marked tearing image data features and the marked deviation image data features of the abnormal image data features to be input into the support vector machine model for training, and obtaining the support vector machine model under the hyper-parameter combination;
the detection module is used for enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect video data of a conveying belt;
the anomaly sensing module is used for enabling the deep neural network model to output a characteristic vector z when a preset condition is met, and judging the current video data of the conveying belt to be abnormal;
and the abnormal category early warning module is used for enabling the support vector machine model to output an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z.
In a third aspect, the present application provides an apparatus having the features of performing tear and off tracking accurate detection on a conveyor belt.
The application is realized by the following technical scheme:
an apparatus comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of a conveyor belt tear and lane deviation detection method as described above when executing the computer program.
In a fourth aspect, the present application provides a storage medium having features for accurately detecting tearing and off tracking of a conveyor belt.
The application is realized by the following technical scheme:
a storage medium storing a computer program which, when executed by a processor, implements the steps of a conveyor belt tear and lane deviation detection method described above.
The application comprises at least one of the following beneficial technical effects:
1. a method for detecting tearing and deviation of a conveying belt is based on a visual detection method of a neural network to detect faults of the conveying belt, train a deep neural network model to extract image characteristics of the conveying belt, classify normal conditions and abnormal conditions of the conveying belt, and output abnormal image data characteristics when the abnormal conditions are judged; marking the tearing image data characteristics and the deviation image data characteristics based on the abnormal image data characteristics, acquiring a hyper-parameter combination of a support vector machine model which enables the classification accuracy to be highest, and training the support vector machine model to further subdivide the tearing condition and the deviation condition of the abnormal images of the conveying belt; the condition that the recognition accuracy rate of the existing tearing or deviation scene of the conveying belt is low is improved, the recognition accuracy rate of the tearing or deviation scene of the conveying belt is improved, the defects of timeliness, efficiency, accuracy, reliability and universality of the traditional detection method are overcome, the high-efficiency and accurate detection of the conveying belt can be realized, the equipment is stopped and maintained in time, and the method has important significance for guaranteeing the safe and reliable operation of an energy factory;
2. the contrast loss function is used for pre-training the deep neural network model so as to improve the extraction capability of the deep neural network model on image characteristics, effectively filter interference characteristics such as image background and noise and improve the training efficiency of the model;
3. optimizing the deep neural network model pre-trained by the comparison loss function by using the training loss function to enable the output characteristic of the normal image of the conveying belt to be close to the center c and the output characteristic of the abnormal image of the conveying belt to be far away from the center c so as to accurately classify the normal image and the abnormal image of the conveying belt;
4. the method comprises the steps of obtaining a hyperparametric combination training SVM model of a support vector machine model, enabling the classification accuracy to be highest, adopting the SVM model under the optimal hyperparametric combination, further classifying the abnormal image data feature types of a conveying belt, effectively avoiding the serious overfitting problem caused by less abnormal data, improving the generalization performance of the model, namely avoiding the deviation problem caused by classification by using a deep neural network model, and improving the judgment precision of abnormal image detection of the conveying belt; meanwhile, the dimensionality of input data is reduced, the detection efficiency of a class image with a large proportion of normal image data of the conveying belt can be improved, and the response speed of the SVM model is improved;
5. the divided tearing image data and the divided deviation image data are subjected to data enhancement processing and then input into a deep neural network model, so that abnormal image samples of the conveying belt in deviation and tearing states are increased, the condition that image sample data of the conveying belt are unbalanced is relieved, and the identification accuracy rate of tearing and deviation scenes of the conveying belt is improved;
6. and after normalization processing is carried out on the tearing image data and the deviation image data which are subjected to data enhancement processing and the normal category image data, the tearing image data and the deviation image data are input into the deep neural network model so as to reduce the data calculation amount during deep neural network model training and facilitate subsequent deep neural network model training.
Drawings
Fig. 1 is a flowchart illustrating a model training process of a method for detecting tearing and deviation of a conveyor belt according to an embodiment of the present disclosure.
FIG. 2 is a training schematic of a deep neural network model and a support vector machine model.
Fig. 3 is a flowchart of a method for detecting tearing and deviation of a conveyor belt according to an embodiment of the present disclosure.
Fig. 4 is a block diagram of a conveyor belt tearing and deviation detecting device according to an embodiment of the present disclosure.
Detailed Description
The present embodiment is only for explaining the present application, and it is not limited to the present application, and those skilled in the art can make modifications of the present embodiment without inventive contribution as needed after reading the present specification, but all of them are protected by patent law within the scope of the claims of the present application.
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
In addition, the term "and/or" herein is only one kind of association relationship describing an associated object, and means that there may be three kinds of relationships, for example, a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship, unless otherwise specified.
The embodiments of the present application will be described in further detail with reference to the drawings attached hereto.
With reference to fig. 1, the present application provides a method for detecting tearing and deviation of a conveyor belt, and the main steps of the method are described as follows.
Acquiring video data of a conveyer belt, and constructing an image database;
inputting data of an image database into a deep neural network model for training, and outputting abnormal image data characteristics;
inputting a hyper-parameter combination which enables the classification accuracy to be highest based on a support vector machine model, and enabling the tearing image data characteristics and the deviation image data characteristics of the marked abnormal image data characteristics to the support vector machine model for training to obtain the support vector machine model under the hyper-parameter combination;
enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect the video data of the conveying belt;
when a preset condition is met, the deep neural network model outputs a characteristic vector z, and the current video data of the conveying belt is judged to be abnormal;
the support vector machine model outputs an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z.
Specifically, video data of the power coal conveying belt of the power plant are collected, and one image is taken at intervals of a plurality of frames to construct an image database. In this embodiment, one image is taken every 5 frames in the video on the conveyor belt, and the images are aggregated to form an image data set.
Then, the images in the image data set are marked into a normal category and an abnormal category to obtain normal category image data and abnormal category image data; and the category image data is further subdivided and marked as tearing image data and deviation image data to construct an image database.
Furthermore, by analyzing the image information in the image database, it is found that a lot of image samples of the conveying belt are available in the normal state, and a few abnormal samples are available in the off tracking and tearing states, so that the problem of unbalance of the image samples occurs, and the identification accuracy of the off tracking and off tracking scenes of the conveying belt is further influenced.
Therefore, data enhancement processing is carried out on the tearing image data and the deviation image data so as to expand the training data set of the abnormal type image and improve the accuracy of the model. In the embodiment, the tearing image data and the deviation image data can be enhanced by adopting enhancement processing modes such as rotation, scaling, color space adjustment, mosaic enhancement and the like, so that abnormal samples of the power plant power coal conveying belt in deviation and tearing states are increased, the condition that image sample data of the conveying belt is unbalanced is relieved, and the identification accuracy of tearing and deviation scenes of the conveying belt is improved.
Furthermore, the tearing image data and the deviation image data which are subjected to data enhancement processing and the normal category image data of the image database are input into the deep neural network model for training after normalization processing. In this embodiment, a maximum and minimum normalization method may be adopted, and the specific formula is as follows:
wherein x isiThe method comprises the steps of representing original pixel points of an image to be processed, representing min (x) the minimum value of image pixels, and representing max (x) the maximum value of the image pixels.
The normalization processing does not change the information storage of the image, but the value range of the pixel value of the image after the normalization processing is converted from 0-255 to 0-1, so that the data calculation amount during the deep neural network model training is reduced, and the subsequent deep neural network model training is facilitated.
In this embodiment, the deep neural network model adopts a ResNet18-SVDD deep neural network model to extract image data features of the conveyor belt.
Specifically, referring to fig. 2, a ResNet18 deep neural network model is constructed, the main part of the model adopts a ResNet18 structure, and then is followed by a plurality of full connection layers FC. In this embodiment, the number of the full connection layers FC may be 3.
And pre-training the deep neural network model by using a contrast learning method through a preset contrast loss function.
Wherein the contrast loss function may include, among other things,
wherein,
I1a normal set of image data is represented,
I2representing an abnormal image data set, I representing a normal image data set I1Or abnormal image data set I2One image of ZiFeature vector representing image I, I1\ I denotes a normal image data set I1J represents the normal image data set I1Or abnormal image data set I2One image other than the certain image i, ZjFeature vector, I, representing image j2\ I represents an abnormal image data set I2T is a normal image data set I1Or abnormal image data set I2An image of ZtIs the feature vector of the image t.
Based on the normalized image data, the contrast loss function is used for pre-training the deep neural network model so as to improve the extraction capability of the deep neural network model on the image characteristics, effectively filter the interference characteristics such as image background, noise and the like, and improve the training efficiency of the model. At the moment, the output of the deep neural network model obtains an output vector zi。
Further, a training loss function is preset, which may include,
wherein, I1Representing a normal image data set, I2Representing an abnormal image data set, ziThe output vector of the input image i on the deep neural network model is shown, c is the average value of the output vectors obtained through comparison learning, and R is a preset training set.
Based on a normal image data set I1Abnormal image data set I2And the output vector z of the input image i on the deep neural network modeliAnd optimizing the deep neural network model pre-trained by the comparison loss function by using the training loss function to obtain the optimized ResNet18-SVDD deep neural network model.
By adopting the optimized ResNet18-SVDD deep neural network model, because of Support Vector Data Description (SVDD), namely, a hypersphere with the minimum radius is found through training a loss function, so that the inside is a positive sample and the outside is a negative sample, a sample just positioned on the hypersphere is a Support Vector for constructing the hypersphere, further the output characteristic of a normal image of the conveying belt is close to a center c, and the output characteristic of an abnormal image of the conveying belt is far away from the center c, so that the normal image and the abnormal image of the conveying belt are accurately classified.
And because the data volume of the abnormal images of the conveying belt is small, even if the optimized ResNet18-SVDD deep neural network model is used for classifying the video data of the conveying belt, the overfitting problem still easily occurs to influence the identification accuracy of the tearing or deviation scene of the conveying belt, and a Support Vector Machine (SVM) is more excellent in the classification problem of a small amount of data, wherein the support vector machine is a two-class model, and the basic model of the support vector machine is a linear classifier with the maximum interval defined on a feature space. The learning strategy of the SVM is interval maximization, can be formalized into a problem of solving convex quadratic programming, and is also equivalent to the minimization problem of a regularized hinge loss function. The learning algorithm of the SVM is an optimization algorithm for solving convex quadratic programming.
Therefore, according to the scheme, the abnormal image data features output by the deep neural network model are input into the support vector machine model for training, so that the types of the abnormal images of the conveying belt are further subdivided.
Referring to fig. 1 and 2, specifically, an output vector z of an optimized ResNet18-SVDD deep neural network model output is obtainediAll abnormal image output vectors z in2And outputting a characteristic data set as an abnormal image, and marking the abnormal image output characteristic data set as a tearing image data characteristic and a deviation image data characteristic.
And (3) carrying out normalization processing on the tearing image data characteristics and the deviation image data characteristics, and dividing 80% of the data set into a training set and 20% of the data set into a testing set so as to preset the training set and the testing set. Wherein the training set is used for training the model; the test set is used to test the models trained by the training set to find the best performing model.
Training an SVM model by using the data set of the tear image data characteristics and the off-tracking image data characteristics after normalization processing, and acquiring the optimal hyper-parameter combination of the SVM by using a Kiring response surface method so as to enable the classification accuracy to be highest, thereby obtaining the SVM model under the optimal hyper-parameter combination.
The step of obtaining the hyperparameter combination of the SVM model which enables the classification accuracy to be highest comprises the following steps: presetting hyper-parameter combinations including kernel functions, punishment parameters and relaxation vectors of a plurality of groups of support vector machine models, training the support vector machine models based on the tearing image data characteristics and the deviation image data characteristics, and obtaining classification accuracy corresponding to each group of hyper-parameter combinations; the classification accuracy is an index for evaluating the performance of the classifier, and the classification accuracy is the ratio of the number of correctly classified samples of the classifier to the total number of samples for a given test data set, that is, the formula of the classification accuracy P is as follows:
wherein, TP is used for predicting positive samples into positive numbers, FP is used for predicting negative samples into positive numbers;
constructing a response relation between the hyper-parameter combination and the classification accuracy, namely a Kriging response surface method, and obtaining a function expression of the hyper-parameter combination and the classification accuracy through a Kriging interpolation method to construct an SVM model;
and searching the hyper-parameter combination with the highest classification accuracy by a genetic algorithm in combination with the expected gain (EI) to obtain the hyper-parameter combination with the highest classification accuracy as the optimal hyper-parameter combination of the SVM, and training to obtain the SVM model under the hyper-parameter combination.
The SVM model under the optimal hyper-parameter combination can further classify the abnormal image data feature types of the conveying belt, so that the problem of serious overfitting caused by less abnormal data is effectively avoided, the generalization performance of the model is improved, namely, the deviation problem caused by classification by using a deep neural network model is avoided, and the judgment precision of the abnormal image detection of the conveying belt is improved; meanwhile, the dimensionality of input data is reduced, the detection efficiency of a class image with a large proportion of normal image data of the conveying belt can be improved, and the response speed of the SVM model is improved.
In this embodiment, the hyper-parameter combinations may be randomly set.
Referring to fig. 3, video data of a conveyor belt is obtained in real time, an image is extracted every a plurality of frames, the image is uploaded to a server, and the image data is input into a trained deep neural network model after being normalized.
And enabling the trained deep neural network model to detect the input image data in real time.
And when the preset condition is met, the deep neural network model outputs the characteristic vector z, and the current video data of the conveying belt is judged to be abnormal. In this embodiment, the preset condition includes that a distance R between a feature vector z output by the deep neural network model and a preset central vector c is greater than a preset threshold R, where R | | z-c | | | ^ 2.
In this embodiment, the preset threshold R may be a radius value of a hypersphere obtained by training the ResNet18-SVDD deep neural network model.
When the distance R is larger than a preset threshold value R, the deep neural network model outputs a characteristic vector z and judges that the video data of the current conveying belt is abnormal; and when the distance R is smaller than or equal to the preset threshold value R, the deep neural network model judges that the video data of the current conveying belt is normal, and continues to detect the next image.
Meanwhile, outputting the characteristic vector z to an SVM model under the optimal hyper-parameter combination, outputting an abnormal result of tearing or deviation of the conveying belt by the SVM model based on the characteristic vector z to obtain the type of the abnormal condition of the conveying belt, and sending an abnormal type prompt.
Further, a conveyor belt tearing and deviation detection method is based on a neural network vision detection method to detect faults of the conveyor belt, train a deep neural network model to extract image characteristics of the conveyor belt, classify normal conditions and abnormal conditions of the conveyor belt, and output abnormal image data characteristics when the abnormal conditions are judged; marking the tearing image data characteristics and the deviation image data characteristics based on the abnormal image data characteristics, acquiring a hyper-parameter combination of a support vector machine model which enables the classification accuracy to be highest, and training the support vector machine model to further subdivide the tearing condition and the deviation condition of the abnormal images of the conveying belt; the condition that the recognition accuracy rate of the existing tearing or deviation scene of the conveying belt is low is improved, the recognition accuracy rate of the tearing or deviation scene of the conveying belt is improved, and the defects of timeliness, efficiency, accuracy, reliability and universality of the traditional detection method are overcome; the conveying belt detection device can realize efficient and accurate conveying belt detection, timely stops equipment for maintenance, and has important significance for guaranteeing safe and reliable operation of an energy factory.
It should be understood that the execution sequence of each process in the above embodiments should be determined by the function and the inherent logic thereof, and should not constitute any limitation to the implementation process of the embodiments of the present application.
Referring to fig. 4, an embodiment of the present application further provides a conveyor belt tearing and deviation detecting device, which corresponds to the conveyor belt tearing and deviation detecting method and system in the foregoing embodiments one to one. This conveyer belt is torn and off tracking detection device includes:
the data acquisition module is used for acquiring video data of the conveyer belt and constructing an image database;
the first training module is used for inputting data of the image database into the deep neural network model for training and outputting abnormal image data characteristics;
the second training module is used for enabling a hyper-parameter combination with the highest classification accuracy to be achieved based on the support vector machine model, enabling the tearing image data characteristics and the deviation image data characteristics of the marked abnormal image data characteristics to be input into the support vector machine model for training, and obtaining the support vector machine model under the hyper-parameter combination;
the detection module is used for enabling the trained deep neural network model and the support vector machine model under the optimal hyper-parameter combination to detect the video data of the conveying belt;
the anomaly sensing module is used for enabling the deep neural network model to output a characteristic vector z when a preset condition is met, and judging the current video data of the conveying belt to be abnormal;
and the abnormal type early warning module is used for enabling the support vector machine model to output an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z and sending an abnormal type prompt.
Further, a conveyer belt is torn and off tracking detection device still includes:
the data enhancement processing module is used for carrying out data enhancement processing on the tearing image data and the deviation image data of the abnormal category image data of the divided image database;
and the data normalization processing module is used for normalizing the tearing image data and the deviation image data subjected to the data enhancement processing and the normal category image data of the image database.
For the specific limitation of the conveyor belt tearing and deviation detecting device, reference may be made to the above limitation on the conveyor belt tearing and deviation detecting method, and details are not repeated here. All or part of each module in the conveyer belt tearing and deviation detecting device can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, an apparatus is provided that includes a computer device, which may be a server. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a conveyor belt tear and lane deviation detection method.
In one embodiment, a storage medium is provided, the storage medium comprising a computer-readable storage medium including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring video data of a conveyer belt, and constructing an image database;
inputting data of an image database into a deep neural network model for training, and outputting abnormal image data characteristics;
inputting a hyper-parameter combination which enables the classification accuracy to be highest based on a support vector machine model, and enabling the tearing image data characteristics and the deviation image data characteristics of the marked abnormal image data characteristics to the support vector machine model for training to obtain the support vector machine model under the hyper-parameter combination;
enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect the video data of the conveying belt;
when a preset condition is met, the deep neural network model outputs a characteristic vector z, and the current video data of the conveying belt is judged to be abnormal;
the support vector machine model outputs an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the system is divided into different functional units or modules to perform all or part of the above-mentioned functions.
Claims (10)
1. A method for detecting tearing and deviation of a conveying belt is characterized by comprising the following steps:
acquiring video data of a conveyer belt, and constructing an image database;
inputting data of the image database into a deep neural network model for training, and outputting abnormal image data characteristics;
inputting a hyper-parameter combination which enables the classification accuracy to be highest based on a support vector machine model, and enabling the marked tearing image data characteristics and the marked deviation image data characteristics of the abnormal image data characteristics to the support vector machine model for training to obtain the support vector machine model under the hyper-parameter combination;
enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect video data of a conveying belt;
when a preset condition is met, the deep neural network model outputs a characteristic vector z, and the current video data of the conveying belt is judged to be abnormal;
and outputting an abnormal result of tearing or deviation of the conveying belt by the support vector machine model based on the characteristic vector z.
2. The conveyor belt tearing and deviation detecting method according to claim 1, wherein the preset condition includes that a distance R between a feature vector z output by the deep neural network model and a preset central vector c is greater than a preset threshold R.
3. The conveyor belt tearing and deviation detecting method as claimed in claim 1, wherein: the step of inputting the data of the image database into a deep neural network model for training comprises the following steps:
presetting a contrast loss function, wherein the contrast loss function comprises,
wherein, I1Representing a normal image data set, I2Representing an abnormal image data set, I representing a normal image data set I1Or abnormal image data set I2One image of ZiFeature vector representing image I, I1\ I denotes a normal image data set I1J represents the normal image data set I1Or abnormal image data set I2One image other than the certain image i, ZjFeature vector, I, representing image j2\ I represents an abnormal image data set I2T is a normal image data set I1Or abnormal image data set I2An image of ZtIs the feature vector of the image t;
and pre-training a deep neural network model by using the contrast loss function.
4. The conveyor belt tearing and deviation detecting method as claimed in claim 3, wherein: after the deep neural network model is pre-trained, the method further comprises the following steps:
presetting a training loss function, the training loss function comprising,
wherein z isiAn output vector of an input image i on the deep neural network model is obtained, and c is an output vector mean value obtained by pre-training;
and optimizing the pre-trained deep neural network model by using the training loss function.
5. The conveyor belt tearing and deviation detecting method according to claim 1, wherein the step of obtaining the hyper-parameter combination of the support vector machine model which enables the highest classification accuracy rate comprises:
presetting hyper-parameter combinations including kernel functions, punishment parameters and relaxation vectors of a plurality of groups of support vector machine models, training the support vector machine models based on the tearing image data characteristics and the deviation image data characteristics, and obtaining classification accuracy corresponding to each group of hyper-parameter combinations;
constructing a response relation between the hyper-parameter combination and the classification accuracy;
and acquiring the hyper-parameter combination with the highest classification accuracy by combining the gain expectation.
6. The conveyor belt tearing and deviation detecting method according to any one of claims 1-5, wherein before inputting the data of the image database into a deep neural network model for training, the method further comprises the following steps:
performing data enhancement processing on torn image data and off-tracking image data of the abnormal category image data of the divided image database;
and inputting the normal category image data of the image database, the tearing image data and the deviation image data which are subjected to data enhancement processing into a depth neural network model.
7. The method for detecting the tearing and the deviation of the conveying belt according to claim 6, wherein before the step of inputting the normal category image data of the image database and the torn image data and the deviation image data which are processed by data enhancement into the deep neural network model, the method further comprises the following steps:
and normalizing the torn image data and the off-tracking image data subjected to data enhancement processing and the normal type image data, and inputting the normalized data into a deep neural network model.
8. The utility model provides a conveyer belt is torn and off tracking detection device which characterized in that includes:
the data acquisition module is used for acquiring video data of the conveyer belt and constructing an image database;
the first training module is used for inputting the data of the image database into a deep neural network model for training and outputting abnormal image data characteristics;
the second training module is used for enabling a hyper-parameter combination with the highest classification accuracy to be achieved based on a support vector machine model, enabling the marked tearing image data features and the marked deviation image data features of the abnormal image data features to be input into the support vector machine model for training, and obtaining the support vector machine model under the hyper-parameter combination;
the detection module is used for enabling the trained deep neural network model and the support vector machine model under the hyper-parameter combination to detect video data of a conveying belt;
the anomaly sensing module is used for enabling the deep neural network model to output a characteristic vector z when a preset condition is met, and judging the current video data of the conveying belt to be abnormal;
and the abnormal category early warning module is used for enabling the support vector machine model to output an abnormal result of tearing or deviation of the conveying belt based on the characteristic vector z.
9. An apparatus comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to perform the steps of the method of any one of claims 1 to 7.
10. A storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111341853.8A CN114187237A (en) | 2021-11-12 | 2021-11-12 | Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111341853.8A CN114187237A (en) | 2021-11-12 | 2021-11-12 | Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114187237A true CN114187237A (en) | 2022-03-15 |
Family
ID=80539952
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111341853.8A Pending CN114187237A (en) | 2021-11-12 | 2021-11-12 | Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114187237A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821305A (en) * | 2022-04-02 | 2022-07-29 | 华南理工大学 | Safety belt identification method, device, equipment and storage medium for electric power operation site |
| CN117585399A (en) * | 2024-01-19 | 2024-02-23 | 汶上义桥煤矿有限责任公司 | Coal mine conveyor belt tearing detection method based on image processing |
-
2021
- 2021-11-12 CN CN202111341853.8A patent/CN114187237A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821305A (en) * | 2022-04-02 | 2022-07-29 | 华南理工大学 | Safety belt identification method, device, equipment and storage medium for electric power operation site |
| CN114821305B (en) * | 2022-04-02 | 2024-06-11 | 华南理工大学 | Method, device, equipment and storage medium for identifying safety belt of electric power operation site |
| CN117585399A (en) * | 2024-01-19 | 2024-02-23 | 汶上义桥煤矿有限责任公司 | Coal mine conveyor belt tearing detection method based on image processing |
| CN117585399B (en) * | 2024-01-19 | 2024-04-09 | 汶上义桥煤矿有限责任公司 | Coal mine conveyor belt tearing detection method based on image processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2023197461A1 (en) | Gearbox fault early warning method and system based on working condition similarity evaluation | |
| CN114187237A (en) | Method, device and equipment for detecting tearing and deviation of conveyer belt and storage medium | |
| CN111027058B (en) | Method for detecting data attack of power system, computer equipment and storage medium | |
| CN112241699B (en) | Object defect type identification method, object defect type identification device, computer equipment and storage medium | |
| CN113297885A (en) | Belt conveyor surface state detection method and device based on convolutional neural network | |
| US20210182615A1 (en) | Alexnet-based insulator self-explosion recognition method | |
| CN109766992A (en) | Industry control abnormality detection and attack classification based on deep learning | |
| CN112017204A (en) | Tool state image classification method based on edge marker graph neural network | |
| CN115496384A (en) | Monitoring management method and device for industrial equipment and computer equipment | |
| CN115526515A (en) | Safety monitoring system of gate for water conservancy and hydropower | |
| CN115762059A (en) | Intelligent flood prevention early warning method and device, storage medium and electronic equipment | |
| CN115170882A (en) | Optimization method of rail wagon part detection network and guardrail breaking fault identification method | |
| CN110674991A (en) | OCSVM (online charging management system VM) -based method for detecting abnormality of primary fan of thermal power plant | |
| CN116714437B (en) | Hydrogen fuel cell automobile safety monitoring system and monitoring method based on big data | |
| CN117485842A (en) | Method and system for monitoring attitude of wheel bucket cross beam of gate type bucket wheel machine in real time | |
| CN112381110A (en) | Coal mine production multidimensional data abnormal probability prediction method based on fuzzy rule | |
| CN116577345A (en) | Method and system for detecting number of tabs of lithium battery | |
| CN115965625A (en) | Instrument detection device based on visual identification and detection method thereof | |
| CN111126490A (en) | Steel plate shape abnormity identification method based on deep random forest | |
| Li et al. | The batch process fault monitoring using adversarial auto-encoder and k-nearest neighbor rule | |
| CN108388232B (en) | Method for monitoring operation mode fault in crude oil desalting process | |
| CN115837862A (en) | Abnormal data detection method and device, electronic equipment and storage medium | |
| CN113283339A (en) | Machine vision-based method for monitoring material shortage abnormity of industrial mineral aggregate conveying belt | |
| Cai et al. | Inductive conformal out-of-distribution detection based on adversarial autoencoders | |
| CN112037197A (en) | Hot-rolled bar cold-shearing material accumulation detection method, system and medium |
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 |