CN117953313B - Method and system for realizing anomaly identification of mine data based on artificial intelligence - Google Patents
Method and system for realizing anomaly identification of mine data based on artificial intelligence Download PDFInfo
- Publication number
- CN117953313B CN117953313B CN202410347664.9A CN202410347664A CN117953313B CN 117953313 B CN117953313 B CN 117953313B CN 202410347664 A CN202410347664 A CN 202410347664A CN 117953313 B CN117953313 B CN 117953313B
- Authority
- CN
- China
- Prior art keywords
- mine
- data
- feature
- features
- periodic
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 16
- 230000000737 periodic effect Effects 0.000 claims abstract description 59
- 238000012549 training Methods 0.000 claims abstract description 45
- 230000004927 fusion Effects 0.000 claims abstract description 15
- 238000003062 neural network model Methods 0.000 claims abstract description 12
- 239000011159 matrix material Substances 0.000 claims description 56
- 230000002159 abnormal effect Effects 0.000 claims description 42
- 239000013598 vector Substances 0.000 claims description 33
- 238000000605 extraction Methods 0.000 claims description 29
- 238000006243 chemical reaction Methods 0.000 claims description 24
- 238000011176 pooling Methods 0.000 claims description 24
- 238000004422 calculation algorithm Methods 0.000 claims description 22
- 230000009467 reduction Effects 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 8
- 230000003595 spectral effect Effects 0.000 claims description 5
- 238000005516 engineering process Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 31
- 230000008569 process Effects 0.000 description 9
- 230000005856 abnormality Effects 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 238000001914 filtration Methods 0.000 description 4
- 238000005065 mining Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000007500 overflow downdraw method Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000005422 blasting Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to an artificial intelligence technology, and discloses an anomaly identification method and system for realizing mine data based on artificial intelligence, wherein the method comprises the following steps: acquiring image data and text data of a mine through a preset acquisition unit, extracting characteristics of the acquired image data and text data to obtain mine image characteristics and mine periodic characteristics, carrying out multi-mode fusion on the mine image characteristics and the mine periodic characteristics to obtain mine characteristic data, training an anomaly identification model according to published standard mine data by utilizing a neural network model, carrying out anomaly identification on the obtained mine characteristic data through the anomaly identification model, and judging whether potential safety hazards exist in the mine according to an anomaly identification result. The method can predict whether the potential safety hazard exists in the mine.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an anomaly identification method and system for realizing mine data based on artificial intelligence.
Background
Mining is a common modern industrial engineering, mining of mines is achieved through blasting, excavating and other methods, and a large number of workers in China engage in mining industries. In the mine exploitation process, if an abnormal situation occurs, a safety accident is most likely to be caused, so that a method capable of timely finding potential safety hazards of the mine is needed to ensure production safety.
Disclosure of Invention
The invention provides an anomaly identification method, system, equipment and computer readable storage medium for realizing mine data based on artificial intelligence, which mainly aim to predict potential safety hazards of mines in the process of mining and ensure production safety.
In order to achieve the above purpose, the invention provides an anomaly identification method for realizing mine data based on artificial intelligence, comprising the following steps:
Acquiring mine data through an acquisition unit preset in a mine;
Extracting features of the mine image in the mine data by a feature extraction algorithm to obtain mine image features, wherein the extracting features of the mine image in the mine data by the feature extraction algorithm comprises the following steps:
and extracting the characteristics of the mine image by using the following characteristic extraction algorithm:
；
Wherein, Represents the/>, in the mine dataThe coordinates of the pixel center points of the individual mine images,Represents the/>, in the mine dataFeatures of individual mine images,/>Representing the number of neighboring pixel points,Gray value of pixel points around central pixel point,/>The gray value representing the center point of the pixel,The method is a binarization function, the pixel point with the gray value larger than or equal to the central pixel point is assigned as 1, otherwise, the pixel point is assigned as 0;
Extracting characteristics of text data in the mine data through a word bag model to obtain mine periodic characteristics, and performing multi-mode fitting on the mine image characteristics and the mine periodic characteristics to obtain mine characteristic data;
carrying out anomaly identification on the mine according to the mine characteristic data by utilizing a pre-trained anomaly identification model to obtain an anomaly value;
judging whether the mine has potential safety hazard according to the abnormal value.
Optionally, the extracting the periodic feature of the text data included in the mine data to obtain the mine periodic feature includes:
The text data is subjected to direct current component removal processing to obtain preprocessed text data;
constructing a frequency domain conversion matrix, and multiplying the frequency domain conversion matrix with the preprocessed text data to obtain frequency domain information;
And carrying out frequency spectrum characteristic analysis on the frequency domain information to obtain mine periodic characteristics.
Optionally, the constructing the frequency domain transformation matrix includes: calculating the data length of the preprocessed text data;
determining a basis function for constructing a frequency domain conversion matrix;
Creating a complex matrix according to the data length;
determining sampling frequency for constructing a frequency domain conversion matrix;
and filling the preprocessed text data into the complex matrix according to the basis function and the sampling frequency to obtain a frequency domain conversion matrix.
Optionally, the performing multi-mode fitting on the mine image feature and the mine periodic feature to obtain mine feature data includes:
converting the mine image features and the mine periodic features into vector forms to obtain mine image feature vectors and mine periodic feature vectors;
mapping the mine image feature vector and the mine periodic feature vector to the same dimension through dimension reduction treatment to obtain dimension reduction mine image features and dimension reduction mine periodic features;
And fusing the dimension-reduced mine image characteristics and the dimension-reduced mine periodic characteristics through weighted fusion to obtain mine characteristic data.
Optionally, the method further includes, before performing anomaly identification on the mine according to the mine characteristic data by using a pre-trained anomaly identification model to obtain an anomaly value:
acquiring published standard mine data in a mine data platform, performing feature extraction on the standard mine data to obtain mine standard feature data, and taking the mine standard feature data as a training data set;
Performing anomaly identification on the mine according to the training data set by utilizing a neural network model to obtain training anomaly parameters;
calculating the difference value of the training abnormal parameter and the preset standard abnormal parameter of the training data set;
Judging whether the difference value is larger than or equal to a preset difference threshold value or not;
If the difference value is greater than or equal to the difference threshold value, after the bias item of the neural network model is adjusted, the step of analyzing the mine by using the neural network model according to the training data set and carrying out anomaly identification is executed again;
And if the difference value is smaller than the difference threshold value, confirming that model training is completed, and obtaining an abnormal recognition model.
Optionally, the calculating the difference value of the training abnormal parameter and the preset standard abnormal parameter of the training data set includes:
the variance value is calculated using the following formula:
；
Wherein, Representing the/>, in the training datasetDifference value of training abnormal parameters of group data and preset standard abnormal parameters of training data set,/>For the training anomaly parameter,/>Presetting standard abnormal parameters for the training data set.
Optionally, the performing anomaly identification on the mine according to the mine characteristic data by using a pre-trained anomaly identification model to obtain an anomaly value includes:
performing data conversion processing on the mine characteristic data by using the anomaly identification model to obtain a characteristic matrix;
performing convolution operation on the feature matrix through a convolution layer by using the anomaly identification model to obtain convolution features;
performing maximum pooling operation on the convolution characteristics through a pooling layer by utilizing the anomaly identification model to obtain pooling characteristics;
And carrying out anomaly identification on the mine through the full-connection layer according to the pooling characteristics by utilizing the anomaly identification model to obtain an anomaly value.
Optionally, the performing convolution operation on the feature matrix by using the anomaly identification model through a convolution layer to obtain a convolution feature includes:
the convolution operation is expressed using the following formula:
；
Wherein, For the/>, in the convolution featureLocal features/>Represents the/>, of the feature matrixLine/>Is a convolution kernel parameter,/>Is an offset term,/>Is a ReLU function.
Optionally, the anomaly identification of the mine by using the anomaly identification model according to the pooling feature through a full connection layer, to obtain an anomaly value, includes:
the outlier is calculated using the following formula:
；
Wherein, For the outlier,/>Is the weight parameter of the full connection layer,/>For the pooling feature,/>Is an offset term,/>As a softmax function.
In order to solve the above problems, the present invention further provides an anomaly identification system for implementing mine data based on artificial intelligence, the system comprising:
And a data acquisition module: acquiring mine data through an acquisition unit preset in a mine;
And the feature extraction module is used for: extracting features of the mine image in the mine data through a feature extraction algorithm to obtain mine image features, and extracting periodic features of text data contained in the mine data to obtain mine periodic features, wherein the extracting features of the mine image in the mine data through the feature extraction algorithm comprises the following steps:
and extracting the characteristics of the mine image by using the following characteristic extraction algorithm:
；
Wherein, Represents the/>, in the mine dataThe coordinates of the pixel center points of the individual mine images,Represents the/>, in the mine dataFeatures of individual mine images,/>Representing the number of neighboring pixel points,Gray value of pixel points around central pixel point,/>The gray value representing the center point of the pixel,The method is a binarization function, the pixel point with the gray value larger than or equal to the central pixel point is assigned as 1, otherwise, the pixel point is assigned as 0;
and a feature fusion module: carrying out multi-mode fitting on the mine image characteristics and the mine periodic characteristics to obtain mine characteristic data;
An anomaly identification module: and carrying out anomaly identification on the mine according to the mine characteristic data by utilizing a pre-trained anomaly identification model to obtain an anomaly value, and judging whether potential safety hazards exist in the mine according to the anomaly value.
According to the embodiment of the invention, the image data and the text data of the mine are acquired through the preset acquisition unit, the acquired image data and text data are subjected to feature extraction to obtain the mine image features and the mine periodic features, the mine image features and the mine periodic features are subjected to multi-mode fusion to obtain the mine feature data, an anomaly identification model is trained according to the published standard mine data by utilizing the neural network model, the obtained mine feature data is subjected to anomaly identification through the anomaly identification model, and whether the mine has potential safety hazards or not is judged according to the anomaly identification result. Therefore, the mine data anomaly identification method and device, the electronic equipment and the computer readable storage medium can predict whether the mine has potential safety hazards.
Drawings
FIG. 1 is a schematic flow chart of a method for identifying anomalies in mine data according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a multi-mode fusion process according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of anomaly identification for mine feature data by an anomaly identification model according to an embodiment of the present invention;
FIG. 4 is a functional block diagram of a mine data anomaly identification device according to an embodiment of the present invention;
the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The embodiment of the application provides an anomaly identification method based on artificial intelligence for mine data and an execution subject including but not limited to at least one of a server, a terminal and the like which can be configured to execute the method provided by the embodiment of the application. In other words, the mine data anomaly identification method may be performed by software or hardware installed in a terminal device or a server device, and the software may be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like. The server may be an independent server, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and basic cloud computing services such as big data and artificial intelligence platforms.
Referring to fig. 1, a flow chart of a mine data anomaly identification method according to an embodiment of the invention is shown. In this embodiment, the mine data anomaly identification method includes:
s1, acquiring mine data through an acquisition unit preset in a mine.
In the embodiment of the invention, the acquisition unit comprises a camera, a laser scanner, a drilling machine and the like.
In the embodiment of the invention, the mine data comprise the topography information, the geological structure information and the like of the mine.
In detail, the topographic information includes height, gradient, degree of relief, type of relief, etc.
Further, the geologic structure information includes formation structure, ore distribution, formation information, and the like.
In the embodiment of the invention, the information such as the landform information, the geological structure information and the like of the mine is acquired by utilizing the preset acquisition unit, so that necessary data is provided for subsequent analysis.
And S2, extracting characteristics of the mine image in the mine data through a characteristic extraction algorithm to obtain the characteristics of the mine image.
In the embodiment of the present invention, the feature extraction of the image data in the mine data by the feature extraction algorithm, to obtain the mine image feature, includes:
denoising the image in the image data to obtain a first preprocessed image;
Graying treatment is carried out on the first preprocessed image, so that a second preprocessed image is obtained;
And carrying out feature extraction on the second preprocessed image by using the feature extraction algorithm to obtain mine image features.
In the embodiment of the invention, the denoising processing is to replace the pixel value of the noise point with the average value or weighted average value of surrounding pixels by processing the pixel value of the image, or reduce the influence of the noise point by other algorithms, and common denoising processing methods comprise mean value filtering, weighted average filtering, median filtering, self-adaptive wiener filtering and the like.
In the embodiment of the invention, the graying process is a process of converting a color image into a gray image, the graying process can simplify color information of the image and reduce complexity of the image, so that the processing efficiency of the image is improved.
In the embodiment of the invention, the dimension reduction processing is to convert the high-dimension data into a low-dimension form so as to simplify the data structure, reduce the calculation cost of an algorithm, improve the data processing speed and simultaneously reserve the important characteristics and rules in the data.
Specifically, the feature extraction of the mine image in the mine data by the feature extraction algorithm includes:
and extracting the characteristics of the mine image by using the following characteristic extraction algorithm:
；
Wherein, Represents the/>, in the mine dataThe coordinates of the pixel center points of the individual mine images,Represents the/>, in the mine dataFeatures of individual mine images,/>Representing the number of neighboring pixel points,Gray value of pixel points around central pixel point,/>The gray value representing the center point of the pixel,The binary function is used for assigning 1 to the pixel point with the gray value larger than or equal to the central pixel point, or assigning 0 to the pixel point.
In the embodiment of the invention, the mine image characteristics are obtained by extracting the characteristics of the image data of the mine, so that the subsequent further analysis is convenient according to the mine image characteristics.
And S3, extracting periodic characteristics of text data contained in the mine data to obtain mine periodic characteristics, and performing multi-mode fitting on the mine image characteristics and the mine periodic characteristics to obtain mine characteristic data.
In the embodiment of the invention, the text data contained in the mine data is extracted to obtain the mine periodic characteristics, which comprises the following steps:
The text data is subjected to direct current component removal processing to obtain preprocessed text data;
constructing a frequency domain conversion matrix, and multiplying the frequency domain conversion matrix with the preprocessed text data to obtain frequency domain information;
And carrying out frequency spectrum characteristic analysis on the frequency domain information to obtain mine periodic characteristics.
In detail, the direct current component processing is a part of the data where the frequency is zero.
In detail, the spectral characteristic analysis is one method of researching frequency domain information of data. The method mainly focuses on energy distribution and phase information of data on different frequencies, and spectral characteristic analysis can reveal the characteristics of frequency distribution, peak frequency, bandwidth and the like of the data.
In detail, the constructing the frequency domain transformation matrix includes: calculating the data length of the preprocessed text data;
determining a basis function for constructing a frequency domain conversion matrix;
Creating a complex matrix according to the data length;
determining sampling frequency for constructing a frequency domain conversion matrix;
and filling the preprocessed text data into the complex matrix according to the basis function and the sampling frequency to obtain a frequency domain conversion matrix.
In detail, the basis functions may be orthogonal basis functions, sine functions, and cosine functions.
In detail, the sampling frequency is a period of sampling, which is set to 24 hours here.
In the embodiment of the invention, since the current acquisition time is recorded when the mine data is acquired, the text data contained in the mine data contains time information.
In detail, since a mine is in a natural environment, data such as an ambient temperature and an ambient humidity of the mine periodically changes with time, and the period is 24 hours.
Furthermore, when the characteristics of the text data are extracted, the periodic characteristics of the text data are required to be extracted, so that images caused by the change of the ambient temperature and the ambient humidity can be eliminated better, and the safety analysis of the mine data can be performed.
In the embodiment of the present invention, the performing multi-modal fitting on the mine image feature and the mine periodic feature to obtain mine feature data includes:
S21, converting the mine image features and the mine periodic features into vector forms to obtain mine image feature vectors and mine periodic feature vectors;
S22, mapping the mine image feature vector and the mine periodic feature vector to the same dimension through dimension reduction processing to obtain dimension reduction mine image features and dimension reduction mine periodic features;
and S23, fusing the dimension-reduced mine image characteristics and the dimension-reduced mine periodic characteristics through weighted fusion to obtain mine characteristic data.
In the embodiment of the invention, the multi-mode fusion refers to a process of integrating information from two or more modes to predict. For example, in the fields of video classification, emotion analysis, speech recognition, etc., data of different modalities such as images, videos, speech, text, etc. can be fused to extract a more comprehensive feature representation. The method of multi-modal fusion can be broadly divided into a joint architecture, a collaborative architecture, and a codec architecture. The joint architecture maps the multi-modal space into the shared semantic subspace to fuse the characteristics of a plurality of modalities; the collaboration framework is used for realizing mutual collaboration of various single modes under the action of some constraints; the codec architecture maps one modality into a multi-modality conversion task of another modality. The multi-mode fusion method has wide application in the field of machine learning.
In detail, the weighted fusion is a data fusion method, which performs weighted summation on information of a plurality of data sources to obtain a more accurate and reliable result, and the core idea of the weighted fusion is to assign different weights to information of different data sources and then perform weighted average. The weight determination needs to consider a plurality of factors such as the reliability of the data source, the precision of the data, the timeliness of the data and the like.
In the embodiment of the invention, the mine image characteristic and the mine periodic characteristic can be converted into vector forms through a Convolutional Neural Network (CNN).
According to the embodiment of the invention, the mine periodic characteristics are obtained by carrying out continuous Fourier transform on the mine text data, and the text characteristics and the mine image characteristics are subjected to multi-mode fusion to obtain the mine characteristic data, so that the subsequent further analysis according to the mine characteristic data is facilitated.
S4, carrying out anomaly identification on the mine according to the mine characteristic data by utilizing a pre-trained anomaly identification model to obtain an anomaly value.
In the embodiment of the invention, the anomaly identification model is also a neural network model.
In the embodiment of the invention, the method further comprises the steps of:
acquiring published standard mine data in a mine data platform, performing feature extraction on the standard mine data to obtain mine standard feature data, and taking the mine standard feature data as a training data set;
Performing anomaly identification on the mine according to the training data set by utilizing a neural network model to obtain training anomaly parameters;
calculating the difference value of the training abnormal parameter and the preset standard abnormal parameter of the training data set;
Judging whether the difference value is larger than or equal to a preset difference threshold value or not;
If the difference value is greater than or equal to the difference threshold value, after the bias item of the neural network model is adjusted, the step of analyzing the mine by using the neural network model according to the training data set and carrying out anomaly identification is executed again;
And if the difference value is smaller than the difference threshold value, confirming that model training is completed, and obtaining an abnormal recognition model.
In the embodiment of the invention, the difference value of the training abnormal parameter and the preset standard abnormal parameter of the training data set is calculated through a difference value algorithm.
In detail, the difference value algorithm may be expressed by the following formula:
；
Wherein, Representing the/>, in the training datasetDifference value of training abnormal parameters of group data and preset standard abnormal parameters of training data set,/>For the training anomaly parameter,/>Presetting standard abnormal parameters for the training data set.
In the embodiment of the invention, the bias term refers to a parameter added in neural network learning and used for controlling the fitting degree of a model to input data. The bias term may move the model up or down at the best fit location to better accommodate different data distributions. By adjusting the bias term, the prediction accuracy of the model can be improved.
In the embodiment of the present invention, the method for performing anomaly identification on a mine by using a pre-trained anomaly identification model according to the mine feature data to obtain an anomaly value includes:
s31, carrying out data conversion processing on the mine characteristic data by using the anomaly identification model to obtain a characteristic matrix;
s32, carrying out convolution operation on the feature matrix through a convolution layer by utilizing the anomaly identification model to obtain convolution features;
S33, carrying out maximum pooling operation on the convolution characteristics through a pooling layer by utilizing the anomaly identification model to obtain pooling characteristics;
s34, carrying out anomaly identification on the mine through the full-connection layer according to the pooling characteristics by utilizing the anomaly identification model to obtain an anomaly value.
In the embodiment of the invention, the data conversion processing refers to representing text features in the mine feature data into word vectors, and then splicing the word vectors to form a matrix, as follows:
；
Wherein, For the spliced matrix,/>Is the length of text,/>The dimension of a word vector refers to the number of parameters used in converting words or characters into a vector matrix.
In detail, text features in the mine feature data are converted into a matrix, and then the matrix is spliced with image features in the mine feature data, so that a feature matrix is obtained.
In the embodiment of the invention, the convolution operation refers to the convolution operation of the feature matrix and the convolution operation according to the convolution kernel parameter, and partial features in the feature matrix are extracted.
Specifically, a feature matrix and a convolution kernel with a size h×d are subjected to convolution operation to obtain a feature map Ci, where h is the height of the convolution kernel and d is the dimension of the word vector. Then, the convolution kernel is moved to the right by one step length, and convolution operation is carried out on the vector matrix and the convolution kernel, so that another feature map Ci+1 is obtained. The process is repeated until the convolution kernel moves to the right end of the input matrix, and finally, the feature mapping c= [ C1, C2, …, ck ] is obtained. Here, k is the number of convolution kernels, which means that the convolution operation is performed using k convolution kernels of different sizes, and in detail, the convolution operation can be expressed by the following formula:
；
Wherein, Is the/>, of the convolution featureLocal features/>Represents the/>, of the feature matrixLine/>Is a convolution kernel parameter,/>Is an offset term,/>To activate the function, a ReLU function is typically used.
In detail, the ReLU function is a commonly used activation function, generally referred to as a ramp function in mathematics, in the neural network, the ReLU function is used as an activation function of a neuron, and a nonlinear output result of the neuron after linear transformation is defined, so that the method has the advantages of high calculation speed, simplicity, clarity, easiness in implementation, small occupied memory and the like. In addition, since the ReLU function has an output proportional to an input when the input is greater than 0, it can effectively solve the gradient vanishing problem.
In the embodiment of the invention, the maximum pooling operation refers to taking the maximum value of each channel in the feature mapping as the output of the channel, and finally obtaining a feature vector after dimension reduction.
Specifically, each feature map in the feature map C is respectively subjected to maximum pooling, and the maximum value pi in each feature map Ci is extracted, because the feature map c= [ C1, C2, …, ck ]. The pooled feature vector can be expressed as:
p=[p1,p2,p3,…,pk]；
Wherein k is the number of convolution kernels and is the same as the number of feature mappings output by the convolution layer.
In the embodiment of the present invention, the fully connected layer is used for mapping the feature vector p output by the pooling layer to the dimension of the classification result, and in detail, the outlier can be calculated by the following formula:
；
Wherein, For the outlier,/>Is the weight parameter of the full connection layer,/>For the pooling feature, the weight parameter is determined according to the actual task requirement,/>Is an offset term,/>To activate the function, a softmax function is typically used.
In detail, the softmax function is a generalization of the logic function, which is used in the probabilistic and related fields to "compress" a K-dimensional vector z containing arbitrary real numbers into another K-dimensional real vector. The feature of this function is that it can compress each element of the vector to within the range of (0, 1) and the sum of all elements is equal to 1.
In the embodiment of the invention, the abnormal recognition model has the implicit characteristic learning capability, and the abnormal recognition model can continuously update the parameters of the model through a back propagation algorithm in the training process, so that the model can automatically learn the implicit characteristics in the data. In the problem of anomaly identification for mines, these implicit features may be abstract features associated with anomalies, such as a particular waveform shape, spectral distribution, etc. Through the deep neural network, the characteristics with higher level and higher expression capacity can be gradually extracted, so that the abnormal state of the mine can be better described.
In detail, the performance of the anomaly identification model is largely dependent on the available data quality and the representativeness of the sample. Thus, collecting sufficient, representative and accurately labeled mine data is critical to ensuring model accuracy. In addition, updating and maintaining the model in time is also an important factor for ensuring the effectiveness of the model under different running environments and data changes.
According to the embodiment of the invention, the standard mine data in the disclosed mine data platform is subjected to characteristic extraction to obtain the mine standard characteristic data, the mine standard characteristic data is used as a training data set to train out an abnormal recognition model, the collected mine data is subjected to abnormal recognition by using the abnormal recognition model, whether the mine has potential safety hazard or not is judged, and the efficiency of abnormal recognition according to the mine data is improved.
S5, judging whether the mine has potential safety hazards according to the abnormal value.
In the embodiment of the invention, the judging whether the mine has potential safety hazard according to the abnormal value comprises the following steps:
Judging whether the abnormal value is larger than or equal to a preset abnormal threshold value or not;
If the abnormal value is greater than or equal to the abnormal threshold value, judging that the mine has potential safety hazard;
And if the abnormal value is smaller than the abnormal threshold value, judging that the mine has no potential safety hazard.
In the embodiment of the invention, the abnormal threshold value is required to be set after analysis and research on the mine according to the existing mine data.
According to the embodiment of the invention, firstly, the mine data is acquired through the preset acquisition unit, the mine data is subjected to characteristic extraction to obtain the mine characteristic data, the mine characteristic data is subjected to abnormality identification through the pre-trained abnormality identification model, the abnormal value is obtained, whether the mine has potential safety hazard or not is judged through judging the magnitude relation between the abnormality and the preset abnormality threshold value, and the efficiency of abnormality identification on the mine is improved.
Fig. 4 is a functional block diagram of a mine data anomaly identification system according to an embodiment of the present invention.
The mine data abnormality recognition device 100 according to the present invention may be mounted in an electronic apparatus. The mine data anomaly identification device 100 may include a data acquisition module 101, a feature extraction module 102, a feature fusion module 103, and an anomaly identification module 104, depending on the functions implemented. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.
In the present embodiment, the functions concerning the respective modules/units are as follows:
the data acquisition module 101: acquiring mine data through an acquisition unit preset in a mine;
the feature extraction module 102: extracting features of the mine image in the mine data through a feature extraction algorithm to obtain mine image features, and extracting periodic features of text data contained in the mine data to obtain mine periodic features, wherein the extracting features of the mine image in the mine data through the feature extraction algorithm comprises the following steps:
and extracting the characteristics of the mine image by using the following characteristic extraction algorithm:
；
Wherein, Represents the/>, in the mine dataThe coordinates of the pixel center points of the individual mine images,Represents the/>, in the mine dataFeatures of individual mine images,/>Representing the number of neighboring pixel points,Gray value of pixel points around central pixel point,/>The gray value representing the center point of the pixel,The method is a binarization function, the pixel point with the gray value larger than or equal to the central pixel point is assigned as 1, otherwise, the pixel point is assigned as 0;
the feature fusion module 103: carrying out multi-mode fitting on the mine image characteristics and the mine periodic characteristics to obtain mine characteristic data;
The anomaly identification module 104: and carrying out anomaly identification on the mine according to the mine characteristic data by utilizing a pre-trained anomaly identification model to obtain an anomaly value, and judging whether potential safety hazards exist in the mine according to the anomaly value.
In detail, each module in the mine data anomaly identification device 100 in the embodiment of the present invention adopts the same technical means as the mine data anomaly identification method described in fig. 1 to 3, and can produce the same technical effects, and will not be described in detail here.
In addition, each functional module in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.
The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The blockchain (Blockchain), essentially a de-centralized database, is a string of data blocks that are generated in association using cryptographic methods, each of which contains information from a batch of network transactions for verifying the validity (anti-counterfeit) of its information and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Wherein artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) is the theory, method, technique, and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend, and expand human intelligence, sense the environment, acquire knowledge, and use knowledge to obtain optimal results.
Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. A plurality of units or means recited in the system claims can also be implemented by means of software or hardware by means of one unit or means. The terms first, second, etc. are used to denote a name, but not any particular order.
Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (6)
1. An anomaly identification method for realizing mine data based on artificial intelligence is characterized by comprising the following steps:
Acquiring mine data through an acquisition unit preset in a mine;
extracting characteristics of the mine image contained in the mine data through a characteristic extraction algorithm to obtain mine image characteristics;
Extracting periodic characteristics of text data contained in the mine data to obtain mine periodic characteristics, wherein the extracting the periodic characteristics of the text data contained in the mine data to obtain the mine periodic characteristics comprises the following steps: removing direct current components from the text data to obtain preprocessed text data, calculating the data length of the preprocessed text data, determining a basis function for constructing a frequency domain conversion matrix, creating a complex matrix according to the data length, determining the sampling frequency for constructing the frequency domain conversion matrix, filling the preprocessed text data into the complex matrix according to the basis function and the sampling frequency to obtain a frequency domain conversion matrix, multiplying the frequency domain conversion matrix with the preprocessed text data to obtain frequency domain information, and performing spectral feature analysis on the frequency domain information to obtain mine periodic features;
Performing multi-mode fitting on the mine image feature and the mine periodic feature to obtain mine feature data, wherein the performing multi-mode fitting on the mine image feature and the mine periodic feature to obtain mine feature data comprises the following steps: converting the mine image features and the mine periodic features into vector forms to obtain mine image feature vectors and mine periodic feature vectors, mapping the mine image feature vectors and the mine periodic feature vectors to the same dimension through dimension reduction processing to obtain dimension reduction mine image features and dimension reduction mine periodic features, and fusing the dimension reduction mine image features and the dimension reduction mine periodic features through weighted fusion to obtain mine feature data;
performing data conversion processing on the mine feature data by using an anomaly identification model to obtain a feature matrix, performing convolution operation on the feature matrix by using the anomaly identification model through a convolution layer to obtain convolution features, performing maximum pooling operation on the convolution features by using the anomaly identification model through a pooling layer to obtain pooling features, performing anomaly identification on a mine by using the anomaly identification model through a full connection layer according to the pooling features to obtain anomaly values, wherein the feature matrix is obtained by performing stitching on the text features in the mine feature data after converting the text features into the matrix and the image features in the mine feature data;
judging whether the mine has potential safety hazard according to the abnormal value.
2. The anomaly identification method of claim 1, wherein prior to anomaly identification of the mine from the mine characteristic data using a pre-trained anomaly identification model, the method further comprises:
acquiring published standard mine data in a mine data platform, performing feature extraction on the standard mine data to obtain mine standard feature data, and taking the mine standard feature data as a training data set;
Performing anomaly identification on the mine according to the training data set by utilizing a neural network model to obtain training anomaly parameters;
calculating the difference value of the training abnormal parameter and the preset standard abnormal parameter of the training data set;
Judging whether the difference value is larger than or equal to a preset difference threshold value or not;
If the difference value is greater than or equal to the difference threshold value, after the bias item of the neural network model is adjusted, the step of analyzing the mine by using the neural network model according to the training data set and carrying out anomaly identification is executed again;
And if the difference value is smaller than the difference threshold value, confirming that model training is completed, and obtaining an abnormal recognition model.
3. The anomaly identification method of claim 2, wherein the calculating of the difference value of the training anomaly parameter and the preset standard anomaly parameter of the training dataset comprises:
the variance value is calculated using the following formula:
，
Wherein, Representing the/>, in the training datasetDifference value of training abnormal parameters of group data and preset standard abnormal parameters of training data set,/>For the training anomaly parameter,/>Presetting standard abnormal parameters for the training data set.
4. The anomaly identification method of claim 2, wherein the convolving the feature matrix with the anomaly identification model via a convolution layer to obtain a convolution feature comprises:
the convolution operation is expressed using the following formula:
，
Wherein, For the/>, in the convolution featureLocal features/>Represents the/>, of the feature matrixLine/>Is a convolution kernel parameter,/>Is an offset term,/>Is a ReLU function.
5. The anomaly identification method according to claim 2, wherein the anomaly identification of the mine by using the anomaly identification model according to the pooling feature through a full connection layer to obtain an anomaly value comprises:
the outlier is calculated using the following formula:
，
Wherein, For the outlier,/>Is the weight parameter of the full connection layer,/>For the pooling feature,/>Is an offset term,/>As a softmax function.
6. An anomaly identification system for implementing mine data based on artificial intelligence, the system comprising:
And a data acquisition module: acquiring mine data through an acquisition unit preset in a mine;
and the feature extraction module is used for: extracting characteristics of mine images contained in the mine data through a characteristic extraction algorithm to obtain mine image characteristics, and extracting periodic characteristics of text data contained in the mine data to obtain mine periodic characteristics;
The extracting the periodic characteristic of the text data contained in the mine data to obtain the periodic characteristic of the mine comprises the following steps: removing direct current components from the text data to obtain preprocessed text data, calculating the data length of the preprocessed text data, determining a basis function for constructing a frequency domain conversion matrix, creating a complex matrix according to the data length, determining the sampling frequency for constructing the frequency domain conversion matrix, filling the preprocessed text data into the complex matrix according to the basis function and the sampling frequency to obtain a frequency domain conversion matrix, multiplying the frequency domain conversion matrix with the preprocessed text data to obtain frequency domain information, and performing spectral feature analysis on the frequency domain information to obtain mine periodic features;
And a feature fusion module: performing multi-mode fitting on the mine image feature and the mine periodic feature to obtain mine feature data, wherein the performing multi-mode fitting on the mine image feature and the mine periodic feature to obtain mine feature data comprises the following steps: converting the mine image features and the mine periodic features into vector forms to obtain mine image feature vectors and mine periodic feature vectors, mapping the mine image feature vectors and the mine periodic feature vectors to the same dimension through dimension reduction processing to obtain dimension reduction mine image features and dimension reduction mine periodic features, and fusing the dimension reduction mine image features and the dimension reduction mine periodic features through weighted fusion to obtain mine feature data;
an anomaly identification module: performing data conversion processing on the mine feature data by using an anomaly identification model to obtain a feature matrix, performing convolution operation on the feature matrix by using the anomaly identification model through a convolution layer to obtain convolution features, performing maximum pooling operation on the convolution features by using the anomaly identification model through a pooling layer to obtain pooling features, performing anomaly identification on a mine by using the anomaly identification model through a full connection layer according to the pooling features to obtain anomaly values, wherein the feature matrix is obtained by performing stitching on the text features in the mine feature data after converting the text features into the matrix and the image features in the mine feature data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410347664.9A CN117953313B (en) | 2024-03-26 | 2024-03-26 | Method and system for realizing anomaly identification of mine data based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410347664.9A CN117953313B (en) | 2024-03-26 | 2024-03-26 | Method and system for realizing anomaly identification of mine data based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117953313A CN117953313A (en) | 2024-04-30 |
| CN117953313B true CN117953313B (en) | 2024-06-18 |
Family
ID=90799700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410347664.9A Active CN117953313B (en) | 2024-03-26 | 2024-03-26 | Method and system for realizing anomaly identification of mine data based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117953313B (en) |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2002025588A2 (en) * | 2000-09-21 | 2002-03-28 | Md Online Inc. | Medical image processing systems |
| CN114936392A (en) * | 2022-05-09 | 2022-08-23 | 无为华塑矿业有限公司 | Method for controlling pre-cracking holes of slope angle in non-coal mine mining |
| CN116363502B (en) * | 2023-01-31 | 2023-10-20 | 中国科学院地理科学与资源研究所 | Mining land multidimensional information acquisition method and device integrating multisource geographic big data |
| CN116975775B (en) * | 2023-06-29 | 2024-01-30 | 中信重工开诚智能装备有限公司 | Deep learning method for coal mine gas outburst prediction based on multi-mode data fusion |
| CN117409541A (en) * | 2023-10-17 | 2024-01-16 | 中铁十九局集团矿业投资有限公司 | Strip mine slope monitoring and early warning method and system |
| CN117574317A (en) * | 2023-12-08 | 2024-02-20 | 国网湖南省电力有限公司 | Mountain fire monitoring method and device based on multi-mode data fusion of sky and ground |
-
2024
- 2024-03-26 CN CN202410347664.9A patent/CN117953313B/en active Active
Non-Patent Citations (2)
| Title |
|---|
| 多模态交互下的露天矿安全双控管理系统交互设计方法研究;何畔;《中国优秀硕士学位论文全文数据库：工程科技Ⅰ辑》;20240315(第3期);1-135 * |
| 多模态内容安全识别关键技术研究;杜鹏飞;《中国优秀博士学位论文全文数据库：信息科技辑》;20230914(第9期);1-130 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117953313A (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110929622A (en) | Video classification method, model training method, device, equipment and storage medium | |
| CN113327279B (en) | Point cloud data processing method and device, computer equipment and storage medium | |
| CN114550053A (en) | Traffic accident responsibility determination method, device, computer equipment and storage medium | |
| CN110941978B (en) | Face clustering method and device for unidentified personnel and storage medium | |
| CN116797787B (en) | Remote sensing image semantic segmentation method based on cross-modal fusion and graph neural network | |
| CN116579616B (en) | Risk identification method based on deep learning | |
| CN114241459B (en) | Driver identity verification method and device, computer equipment and storage medium | |
| CN112418292A (en) | Image quality evaluation method and device, computer equipment and storage medium | |
| CN111738054A (en) | Behavior anomaly detection method based on space-time self-encoder network and space-time CNN | |
| JP2024513596A (en) | Image processing method and apparatus and computer readable storage medium | |
| CN114332473A (en) | Object detection method, object detection device, computer equipment, storage medium and program product | |
| CN114550051A (en) | Vehicle loss detection method and device, computer equipment and storage medium | |
| CN114611672A (en) | Model training method, face recognition method and device | |
| CN117036843A (en) | Target detection model training method, target detection method and device | |
| CN114612902A (en) | Image semantic segmentation method, device, equipment, storage medium and program product | |
| CN111402156A (en) | Restoration method and device for smear image, storage medium and terminal equipment | |
| CN114943937A (en) | Pedestrian re-identification method and device, storage medium and electronic equipment | |
| CN113781462A (en) | Human body disability detection method, device, equipment and storage medium | |
| CN110555406B (en) | Video moving target identification method based on Haar-like characteristics and CNN matching | |
| CN117953313B (en) | Method and system for realizing anomaly identification of mine data based on artificial intelligence | |
| CN112183303A (en) | Transformer equipment image classification method and device, computer equipment and medium | |
| CN114692715A (en) | Sample labeling method and device | |
| CN117173731B (en) | Model training method, image processing method and related device | |
| Wang et al. | Keyframe image processing of semantic 3D point clouds based on deep learning | |
| CN117058498B (en) | Training method of segmentation map evaluation model, and segmentation map evaluation method and device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |