AU2020103251A4 - Method and system for identifying metallic minerals under microscope based on bp nueral network - Google Patents
Method and system for identifying metallic minerals under microscope based on bp nueral network Download PDFInfo
- Publication number
- AU2020103251A4 AU2020103251A4 AU2020103251A AU2020103251A AU2020103251A4 AU 2020103251 A4 AU2020103251 A4 AU 2020103251A4 AU 2020103251 A AU2020103251 A AU 2020103251A AU 2020103251 A AU2020103251 A AU 2020103251A AU 2020103251 A4 AU2020103251 A4 AU 2020103251A4
- Authority
- AU
- Australia
- Prior art keywords
- metallic mineral
- neural network
- characteristic data
- unknown
- image
- 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.)
- Ceased
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention discloses a method and system for identifying metallic minerals
under a microscope based on a BP neural network, and relates to the technical field of
identification of the metallic minerals. The method mainly includes the steps: acquiring an
unknown metallic mineral image photographed by a reflection polarizing microscope; processing
the unknown metallic mineral image by using MATLAB software to obtain primary
characteristic data of the unknown metallic mineral image; determining environmental difference
characteristics; calculating final characteristic data of the unknown metallic mineral image
according to the primary characteristic data of the unknown metallic mineral image and the
environmental difference characteristics; and inputting the final characteristic data of the
unknown metallic mineral image into a trained BP neural network model, and determining the
type of the unknown metallic mineral. The present invention automatically identifies the image
through the computer software and the reflection polarizing microscope, so that on the basis of
ensuring the low identification cost of the metallic minerals, the personal and environmental
interference can be prevented, and the identification accuracy can be improved.
Drawings of Description
Acquire an unknown metallic mineral image photographed by a reflection polarizing microscope
Process the unknown metallic mineral image by using MATLAB software to obtain primary
characteristic data of the unknown metallic mineral image 102
Determine environmental difference characteristics
103
alculate fial characteristic data ofthe unknown metallic mineral image according to the primary
characteristic data of the unknown metallic mineral image and the environmental difference
characteristics
Input the final characteristic dataof the unknown metallic mineral image into a trained BP neural
network model, and determine the type of the unknown metallic mineral
Fig. 1
1
Description
Drawings of Description
Acquire an unknown metallic mineral image photographed by a reflection polarizing microscope
Process the unknown metallic mineral image by using MATLAB software to obtain primary characteristic data of the unknown metallic mineral image 102
Determine environmental difference characteristics 103
alculate fial characteristic data ofthe unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristics
Input the final characteristic dataof the unknown metallic mineral image into a trained BP neural network model, and determine the type of the unknown metallic mineral
Fig. 1
Description
Technical Field
The present invention relates to the technical field of identification of metallic minerals, and
particularly relates to a method and system for identifying metallic minerals under a microscope
based on a BP neural network.
Background With the development of economy and society, the demand on metal is also continuously increasing, and higher requirements are put forward for the mining and identification of metal ores. There are a lot of identification methods for the metal ores. The electronic probe analysis technology and the reflection polarizing microscope technology are conventional identification methods. With the development of science and information, the electronic probe analysis technology is constantly reformed and innovated, and has been continuously upgraded and improved on the basis of the previous multi-element analysis technology. However, the electronic probe analysis technology has the following disadvantages: electronic probe analysis instruments are expensive; the unit price of one electronic probe analysis instrument is up to one million Yuan, and special maintenance and services are required in the later period, so the electronic probe analysis instrument is not applicable to ordinary study and work in universities; the electronic probe analysis instruments are heavy and inconvenient to carry, thereby limiting the application range; and the electronic probe analysis instrument shall be operated by people with certain experience. The reflection polarizing microscope technology utilizes a reflection polarizing microscope to identify the metallic minerals and mainly identifies the metallic minerals by utilizing the optical properties and physical properties of the metallic minerals. Generally, for the minerals with similar reflectivity, the level of the reflectivity is difficult to determine by a simple comparison method. The minerals with similar reflection colors are difficult to describe in precise language. Furthermore, due to the individual difference of people, especially for people with color blindness or color weakness, the color of the minerals under the microscope is more difficult to observe; for example, the colors such as creamy yellow, light yellow, light brown,
Description
dark brown and chocolate brown are difficult to distinguish sometimes; and even the experienced practitioners cannot guarantee 100% accuracy, and the error is large.
Summary The objective of the present invention is to provide a method and system for identifying metallic minerals under a microscope based on a BP neural network, which utilizes a machine learning way to solve the problems of high cost and large error in identification of some metallic minerals in daily geological activities. To realize the above objective, the present invention adopts the following technical solutions: A method for identifying metallic minerals under a microscope based on a BP neural network includes the following steps: acquiring an unknown metallic mineral image photographed by a reflection polarizing microscope; processing the unknown metallic mineral image by using MATLAB software to obtain primary characteristic data of the unknown metallic mineral image; determining environmental difference characteristics, wherein the environmental difference characteristics are differences between first characteristic data and second characteristic data; the first characteristic data is characteristic data of a first metallic mineral image under illumination intensity when the unknown metallic mineral image is photographed; the second characteristic data is the characteristic data of a second metallic mineral image under the illumination intensity when a BP neural network model is trained; and the illumination density of the first metallic mineral image is different from the illumination intensity of the second metallic mineral image, and the metallic mineral types are same; calculating final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic; inputting the final characteristic data of the unknown metallic mineral image into the trained BP neural network model, and determining the type of the unknown metallic mineral, wherein the input of the trained BP neural network model is the final characteristic data of the unknown metallic mineral image, and the output of the trained BP neural network model is the type of the metallic mineral.
Description
Optionally, processing the unknown metallic mineral image by using the MATLAB software to obtain primary characteristic data of the unknown metallic mineral image specifically includes: reading the unknown metallic mineral image into the MATLAB software by using an imread function, wherein a storage form of the unknown metallic mineral image in the MATLAB software is a three-dimensional array form, and the three-dimensional array form contains RGB color information of the unknown metallic mineral image; sequentially performing the average filtering and normalization for the unknown metallic mineral image; extracting the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image. Optionally, extracting the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image specifically includes: performing the dimension separation for the normalized unknown metallic mineral image to obtain three one-dimension arrays, wherein the three one-dimension arrays are respectively a first one-dimension array, a second one-dimension array, and a third one-dimension array; calculating an average value of each of the one-dimension arrays by using a mean function; determining the primary characteristic data of the unknown metallic mineral image according to the average value of all the one-dimension arrays, wherein the primary characteristic data of the unknown metallic mineral image includes four elements; the first element is the average value of the first one-dimension array; the second element is the average value of the second one-dimension array; the third element is the average value of the third one-dimension array; and the fourth element is the type of the unknown metallic mineral. Optionally, calculating final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic specifically includes: superimposing the environmental difference characteristics into the primary characteristic data of the unknown metallic mineral image to obtain the final characteristic data of the unknown metallic mineral image. Optionally, training the BP neural network model includes the following steps: acquiring different types of historical metallic mineral images under the same illumination intensity;
Description
processing all historical metallic mineral images by using the MATLAB software; grouping the processed historical metallic mineral images to obtain a training set and a test set; training the BP neural network model by using the data in the training set, stopping training the BP neural network model after a set training stop condition is satisfied, and obtaining a primary BP neural network model; testing the primary BP neural network model by using the data in the test set, stopping training the BP neural network model after a set test condition is satisfied, and saving the primary BP neural network model satisfying the test condition, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model. A system for identifying metallic minerals under a microscope based on a BP neural network includes: an unknown metallic mineral image acquiring module used to acquire an unknown metallic mineral image photographed by a reflection polarizing microscope; an unknown metallic mineral image primary characteristic data calculation module used to process the unknown metallic mineral image by using MATLAB software to obtain primary characteristic data of the unknown metallic mineral image; an environmental difference characteristic determination module used to determine environmental difference characteristics, wherein the environmental difference characteristics are differences between the first characteristic data and the second characteristic data; the first characteristic data is characteristic data of the first metallic mineral image under illumination intensity when the unknown metallic mineral image is obtained; the second characteristic data is the characteristic data of a second metallic mineral image under the illumination intensity when a BP neural network model is trained; and the illumination density of the first metallic mineral image is different from the illumination intensity of the second metallic mineral image, and the metallic mineral types are same; an unknown metallic mineral image final characteristic data calculation module used to calculate final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic; an unknown metallic mineral type determination module used to input the final characteristic data of the unknown metallic mineral image into the trained BP neural network model, and determine the type of the unknown metallic mineral, wherein the input of the trained
Description
BP neural network model is the final characteristic data of the unknown metallic mineral image, and the output of the trained BP neural network model is the type of the metallic mineral. Optionally, the unknown metallic mineral image primary characteristic data calculation module specifically includes: a reading unit used to read the unknown metallic mineral image into MATLAB software by using an imread function, wherein a storage form of the unknown metallic mineral image in the MATLAB software is a three-dimensional array form, and the three-dimensional array form contains RGB color information of the unknown metallic mineral image; a processing unit used to sequentially perform the average filtering and normalization for the unknown metallic mineral image; an unknown metallic mineral image primary characteristic data determination unit used to extract the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image. Optionally, the unknown metallic mineral image primary characteristic data determination unit specifically includes: a dimension separation subunit used to perform the dimension separation for the normalized unknown metallic mineral image to obtain three one-dimension arrays, wherein the three one-dimension arrays are respectively a first one-dimension array, a second one-dimension array, and a third one-dimension array; an average value calculation subunit used to calculate an average value of each one-dimension array by using a mean function; an unknown metallic mineral image primary characteristic data determination subunit used to determine the primary characteristic data of the unknown metallic mineral image according to the average values of the one-dimension arrays, wherein the primary characteristic data of the unknown metallic mineral image includes four elements; the first element is the average value of the first one-dimension array; the second element is the average value of the second one-dimension array; the third element is the average value of the third one-dimension array, and the fourth element is the type of the unknown metallic mineral. Optionally, the unknown metallic mineral image final characteristic data calculation module specifically includes: an unknown metallic mineral image final characteristic data calculation unit used to superimpose the environmental difference characteristic into the primary characteristic data of
Description
the unknown metallic mineral image to obtain the final characteristic data of the unknown metallic mineral image. Optionally, the identification system also includes a BP neural network model training module, wherein the BP neural network model training module specifically includes: a historical metallic mineral image acquiring unit used to acquire different types of historical metallic mineral images under the same illumination intensity; a historical metallic mineral image processing unit used to process all historical metallic mineral images by using the MATLAB software; a training set and test set determination unit used to group the processed historical metallic mineral images to obtain a training set and a test set; a primary BP neural network model determination unit used to train the BP neural network model by using data in the training set, stop training the BP neural network model after a set training stop condition is satisfied, and obtain a primary BP neural network model; a trained BP neural network model determination unit used to test the primary BP neural network model by using data in the test set, stop training the BP neural network model after a set test condition is satisfied, and save the primary BP neural network model satisfying the test condition, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model. According to specific embodiments provided by the present invention, the present invention has the following technical effects: The method and system for identifying the metallic minerals under the microscope based on the BP neural network in the present invention automatically identify the image through the computer software and the reflection polarizing microscope, so that on the basis of ensuring the low identification cost of the metallic minerals, the personal and environmental interference can be prevented, the identification accuracy can be improved, and the application range is extended.
Description of Drawings To more clearly describe the technical solutions in the embodiments of the present invention or in prior art, the drawings required to be used in the embodiments will be simply presented below. Apparently, the drawings in the following description are merely some embodiments of the present invention, and for those ordinary skilled in the art, other drawings can also be obtained according to these drawings without contributing creative labor.
Description
Fig. 1 is a flow chart of a method for identifying metallic minerals under a microscope based on a BP neural network according to embodiments of the present invention; Fig. 2 is a structural diagram of a system for identifying metallic minerals under a microscope based on a BP neural network according to embodiments of the present invention; Fig. 3 is a gelenite image under a 40-times microscope according to embodiments of the present invention; Fig. 4 is a brightness distribution image of a surface of a gelenite picture according to embodiments of the present invention; Fig. 5 is a ludwigite image under the 40-times microscope according to embodiments of the present invention; Fig. 6 is a sphalerite image under the 40-times microscope according to embodiments of the present invention; and Fig. 7 is a structural diagram of a BP neural network model according to embodiments of the present invention.
Detailed Description The technical solutions in the embodiments of the present invention will be clearly and fully described below in combination with the drawings in the embodiments of the present invention. Apparently, the described embodiments are merely part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those ordinary skilled in the art without contributing creative labor will belong to the protection scope of the present invention. To make the above purpose, features and advantages of the present invention more clear and understandable, the present invention will be further described below in detail in combination with the drawings and specific embodiments. Embodiment 1 As shown in Fig. 1, a method for identifying metallic minerals under a microscope based on a BP neural network provided by the present embodiment includes the following steps: Step 101: an unknown metallic mineral image photographed by a reflection polarizing microscope is acquired. Step 102: the unknown metallic mineral image is processed by MATLAB software to obtain primary characteristic data of the unknown metallic mineral image.
Description
Step 103: environmental difference characteristics are determined, wherein the environmental difference characteristics are differences between the first characteristic data and the second characteristic data; the first characteristic data is characteristic data of a first metallic mineral image under illumination intensity when the unknown metallic mineral image is photographed; the second characteristic data is the characteristic data of a second metallic mineral image under the illumination intensity when a BP neural network model is trained; and the illumination density of the first metallic mineral image is different from the illumination intensity of the second metallic mineral image, and the metallic mineral types are same; Step 104: according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic, final characteristic data of the unknown metallic mineral image is calculated; Step 105: the final characteristic data of the unknown metallic mineral image is inputted into the trained BP neural network model, and the type of the unknown metallic mineral is determined, wherein the input of the trained BP neural network model is thefinal characteristic data of the unknown metallic mineral image, and the output of the trained BP neural network model is the type of the metallic mineral. The step 102 specifically includes: The unknown metallic mineral image is read into MATLAB software by using an imread function, wherein a storage form of the unknown metallic mineral image in the MATLAB software is a three-dimensional array form, and the three-dimensional array form contains RGB color information of the unknown metallic mineral image. The unknown metallic mineral image is subjected to average filtering and normalization sequentially. The characteristics of the normalized unknown metallic mineral image are extracted to obtain the primary characteristic data of the unknown metallic mineral image. Specifically, in step I, the normalized unknown metallic mineral image is subjected to dimension separation to obtain three one-dimension arrays. The three one-dimension arrays are respectively a first one-dimension array, a second one-dimension array and a third one-dimension array; in step II, a mean function is used to calculate an average value of each of the one-dimension arrays; in step III, according to all the average values of the one-dimension arrays, the primary characteristic data of the unknown metallic mineral image is determined. The primary characteristic data of the unknown metallic mineral image includes four elements; the first element is the average value of the first one-dimension array; the second element is the average value of the second
Description
one-dimension array; the third element is the average value of the third one-dimension array; and the fourth element is the type of the unknown metallic mineral. The step 104 specifically includes: The environmental difference characteristic is superimposed into the primary characteristic data of the unknown metallic mineral image to obtain the final characteristic data of the unknown metallic mineral image. Optionally, the present invention also includes a step of training the BP neural network model, which is specifically as follows: Different types of historical metallic mineral images under the same illumination intensity are acquired. The MATLAB software is used to process all historical metallic mineral images. The processed historical metallic mineral images are grouped to obtain a training set and a test set; The BP neural network model is trained by using data in the training set; the training of the BP neural network model is stopped after a set training stopping condition is satisfied; and a primary BP neural network model is obtained. The primary BP neural network model is tested by using the data in the test set; the training of the BP neural network model is stopped after a set test condition is satisfied; and the primary BP neural network model satisfying the test condition is saved, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model. Embodiment II As shown in Fig. 2, the present embodiment also provides a system for identifying metallic minerals under a microscope based on a BP neural network, which includes: an unknown metallic mineral image acquiring module 201 used to acquire an unknown metallic mineral image photographed by a reflection polarizing microscope; an unknown metallic mineral image primary characteristic data calculation module 202 used to process the unknown metallic mineral image by using the MATLAB software to obtain primary characteristic data of the unknown metallic mineral image; an environmental difference characteristic determination module 203 used to determine environmental difference characteristics, wherein the environmental difference characteristics are differences between the first characteristic data and the second characteristic data; the first characteristic data is characteristic data of a first metallic mineral image under illumination intensity when the unknown metallic mineral image is photographed; the second characteristic
Description
data is the characteristic data of a second metallic mineral image under the illumination intensity when a BP neural network model is trained; and the illumination density of the first metallic mineral image is different from the illumination intensity of the second metallic mineral image, and the metallic mineral types are same; an unknown metallic mineral image final characteristic data calculation module 204 used to calculate final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic; an unknown metallic mineral type determination module 205 used to input the final characteristic data of the unknown metallic mineral image into the trained BP neural network model, and determine the type of the unknown metallic mineral, wherein the input of the trained BP neural network model is the final characteristic data of the unknown metallic mineral image, and the output of the trained BP neural network model is the type of the metallic mineral. The unknown metallic mineral image primary characteristic data calculation module 202 specifically includes: a reading unit used to read the unknown metallic mineral image into the MATLAB software by using an imread function, wherein a storage form of the unknown metallic mineral image in the MATLAB software is a three-dimensional array form, and the three-dimensional array form contains RGB color information of the unknown metallic mineral image; a processing unit used to sequentially perform the average filtering and normalization for the unknown metallic mineral image; an unknown metallic mineral image primary characteristic data determination unit used to extract the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image. The unknown metallic mineral image primary characteristic data determination unit specifically includes: a dimension separation subunit used to perform the dimension separation for the normalized unknown metallic mineral image to obtain three one-dimension arrays, wherein the three one-dimension arrays are respectively a first one-dimension array, a second one-dimension array, and a third one-dimension array; an average value calculation subunit used to calculate an average value of each of the one-dimension arrays by using a mean function;
Description
an unknown metallic mineral image primary characteristic data determination subunit used to determine the primary characteristic data of the unknown metallic mineral image according to the average values of the one-dimension arrays, wherein the primary characteristic data of the unknown metallic mineral image includes four elements; the first element is the average value of the first one-dimension array; the second element is the average value of the second one-dimension array; the third element is the average value of the third one-dimension array; and the fourth element is the type of the unknown metallic mineral. The unknown metallic mineral image final characteristic data calculation module 204 specifically includes: an unknown metallic mineral image final characteristic data calculation unit used to superimpose the environmental difference characteristics into the primary characteristic data of the unknown metallic mineral image to obtain the final characteristic data of the unknown metallic mineral image. Preferably, the identification system also includes a BP neural network model training module, wherein the BP neural network model training module specifically includes: a historical metallic mineral image acquiring unit used to acquire different types of historical metallic mineral images under the same illumination intensity; a historical metallic mineral image processing unit used to process all historical metallic mineral images by using MATLAB software; a training set and test set determination unit used to group the processed historical metallic mineral images to obtain a training set and a test set; a primary BP neural network model determination unit used to train the BP neural network model by using the data in the training set, stop training the BP neural network model after a set training stop condition is satisfied, and obtain a primary BP neural network model; a trained BP neural network model determination unit used to test the primary BP neural network model by using the data in the test set, stop training the BP neural network model after a set test condition is satisfied, and save the primary BP neural network model satisfying the test condition, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model. Embodiment III Introduction of identification principle The RGB color mode is a color standard in the industry. Various colors are obtained by changing three color channels of red (R), green (G) and blue (B) and superimposing them onto
Description
each other. RGB represents the colors of the red, green and blue three channels. This standard includes almost all colors that can be perceived by human vision and is one of the most widely used color systems. Under normal circumstances, RGB each has 256 levels of brightness, which are represented by numbers from 0, 1, 2... to 255, wherein 0 means the darkest, 255 means the brightest, and the value changes with the brightness. The change is linear. The 256 levels of RGB colors can be combined into about 16.78 million colors, that is 256x256x256=16777216. Since there are no metal minerals with exactly the same color brightness, when all image information can be recorded during photographing, there can be about 16.78 million types of color brightness information of metal minerals, which sets the foundation for RGB color components serving as the identification characteristics. The RGB color components in the computer MATLAB software are stored in a three-dimensional matrix form. Each dimension stores one color component. For example, a picture of gelenite (as shown in Fig. 3) is read to study its brightness (the brightness is reflected by the RGB color component) distribution (as shown in Fig. 4). It is discovered that the surface brightness of the gelenite is relatively uniform. Actually, the reflectivity (positively correlated to the brightness) of the metallic minerals such as the gelenite is a determined value, which means that the brightness distribution of the metallic mineral is relatively uniform. After the brightness is averaged, the brightness is fluctuated in a very small range. Introduction of the neural network principle The BP (Back Propagation) network was proposed by a group of scientists led by Rumelhart and McCelland in 1986. It is a multi-layer feedforward network trained by an error back propagation algorithm and is currently one of the most widely used neural network models. The BP network can learn and store a large number of input-output pattern mapping relationships without revealing mathematical equations describing this mapping relationship in advance. Its learning rule is to use a steepest descent method to continuously adjust the weight and threshold of the network through back propagation to minimize the sum of squared errors. A BP neural network model topological structure includes an input layer, a hide layer and an output layer. Each layer is composed of units (also called neural nodes). The input layer has the characteristics of the training set and is passed to the next layer through the weights of connection points. The output of one layer is the input of the next layer, and the number of hide layers is arbitrary. Each layer must be weighted and summed and outputs the sum according to a
Description
nonlinear transformation equation. Through the BP (Backpropagation) algorithm, an error between a predicted value and a true value of the neural network output layer is compared, and the error is minimized in the opposite direction (from the output layer => hide layer => input layer) to update the weight of each connection. At present, the BP network has already been widely applied to the fields such as classification identification, approximation, regression and compression. Based on the above principle, a method for identifying metallic minerals under a microscope based on a BP neural network provided by the present embodiment mainly includes the following steps: Step I: an image is collected. The step I specifically includes: an OLYMPUS-BX5IM reflection polarizing microscope is used for setting the exposure time of 750[tS, sensitivity (ISO) of 200 and resolution of 1600x1200. A digital camera of the reflection polarizing microscope is used to separately collect minerals such as gelenite, sphalerite and boronite. Each mineral was photographed for about 20 pictures, as shown in Fig. 3, Fig. 5 and Fig. 6. Step II: image processing and characteristic extraction The image is placed at a root path specified by the MATLAB software, and is inputted by using a function of the software. The storage form of the image in the MATLAB software is one three-dimension array at this time. The three-dimension array is the RGB color information of the image. The image processing specifically includes: 1. Digitalization: the imread function is used to read the image into the MATLAB software, and the image is changed to a three-dimension matrix, wherein the size of each element in the matrix is 0-256. 2. Average filtering: a fspecial ('average') function in the MATLAB software is used to filter the image. The averaging filtering is also called linear filtering, which mainly adopts a neighborhood averaging method. The core concept of the averaging filter is to regard the entire image as including a plurality of blocks with a constant gray level. The correlation of the adjacent pixels is high, but the noise has the statistics independence. Therefore, the neighborhood average value can be used to substitute each pixel value in the original image. The above processing is mainly to eliminate the noise. 3. Normalization: a premnmax (P, T) function in the MATLAB software is used to perform the normalization, wherein P and T are respectively the original input and output data. The
Description
normalization is to change the number to a decimal between (0, 1) in order to eliminate dimensions and facilitate the data processing. The characteristic extraction specifically includes: 1. Dimension separation: since the read image is a three-dimension array, the three-dimension array needs to be separated into three one-dimension arrays. 2. Averaging: each dimension of the three-dimension array is averaged; and the mean function is directly used to average each dimension of arrays. There are three average values in three dimensions. One picture has three average values. These three average values are stored in an array of size 4, wherein the first three store the average value, and the fourth stores the mineral type. For example, gelenite: array{250, 253, 253, gelenite}. Each mineral has 20 pictures, and a total of 20 arrays are obtained. Step III: training of the BP neural network model Firstly, the image characteristic data is divided into a training set and a test set; and each mineral has 20 arrays, of which 15 groups are used as the training sets and 5 groups are used as the test sets. The training sets of all minerals are combined into one training set, and test sets are combined into one test set. Secondly, neural network training: as shown in Fig. 7, the neural network training is actually to minimize the error between the output value of a certain mineral and the mineral mark. Wij is used to represent the weight and 0 indicates the bias. A main line of realizing this process is to update the weight and bias in the training process, thereby obtaining the most suitable weight and bias, because the values of the hide layer and output layer in the neural network mainly depend on the weight and bias. The method specifically includes: 1. Input layer->hide layer->output layer. Three nodes, i.e. xl, x2 and x3 of the input layer are characteristic values of manual input, which are reflected in the MATLAB software function as net=newff(P,...), wherein P is the matrix of the characteristic values xl, x2 and x3 of the input. The characteristic values are propagated backwards from the input layer, and specifically propagated or advanced towards the hide layer. At this time, it is necessary to calculate the values of the two nodes of the hide layer. The value of each input layer is multiplied by the sum of the weights of the corresponding node and then plus the bias to obtain a value which is brought into a conversion function sigmoid(f(x)=1/1+e-x). For example, the value of No. 4 node is f(x)=1/1+e-x, wherein x=xlxw 14 +x2xw24 +x3xw3 4 +01. Similarly, the value of No. 5 node can be obtained. The value of each node is represented by Oi here.
Description After the value of the hide layer is obtained, the neural network continues the backpropagation. Because the values of the No. 4 and No. 5 nodes are already known, the value of the No. 6 node, i.e. the value of the output layer can be obtained. Next, it is necessary to calculate the error of the output layer. 2. Error of the output layer The purpose of the training is to minimize the error between the output value of the certain mineral and the mineral mark. In fact, the true mark of the output layer is already known, which is contained in the training set. For example, the characteristic value {190, 195, 205} of the gelenite is input, the corresponding mark of the gelenite is 1, and the error between the value of the output layer and the corresponding mark should be calculated. The corresponding calculation formula is Errj=Oj (1-Oj )(Tj -Oj). In the above formula, Tj is the corresponding mark, and Oi is the value of the to-be-calculated error node. The error between the prediction value and true value of the output layer can be calculated. If this error is small enough, the training should be stopped, and the current weight and bias of the neural network are saved. Otherwise, the training is continued to update the weight and bias. The updating strategy is the backpropagation, that is, from the output layer to the hide layer to the input layer. 3. Output layer->hide layer->input layer. 4. Error of the hide layer Like the output layer, the hide layer also needs to calculate the error of each node, and the formula is Err, =0,(1-0,) Errw1
In the formula, Errk is the error of the nodes following the calculated node. Wjk is the weight of this node and the subsequent node. For example, the error of the No. 5 node is calculated as Err5=0 5(1-0 5)Err6 X w56. If there are multiple output layer nodes, it can be obtained
from the formula that the error of each output layer node is multiplied by the weight corresponding to the previous layer node and then multiplied by oj (1-Oj). Similarly, the error of the No. 4 node can be obtained. After the errors of the No. 4, No. 5 and No.6 nodes are obtained, the weight and bias need to be updated subsequently. 5. Weight updating Weight updating formulas are:
Description
Awij =(1)Errj Oj; wij =wij +Awij;
Awij is variation of the weight and is obtained by multiplying the error of the node by the value of the node and then multiplying by a learning rate 1 (generally a default value, which does not need to be manually set) and then plus the original weight to obtain a new weight. The weights of all nodes can be calculated through the above formulas. 6. Bias updating Bias updating formula are: A0j =(1)Errj 0j =0j +A0j The bias updating formula is similar to the weight updating formula, but the value of AO is equal to the learning rate multiplied by the error, and there is no value of the node. All biases are updated through the above formulas. The above process is a complete training process. After the weights and biases are updated, another set of characteristic values is input to the neural network for repeating the above process. The stopping condition of the training process: the weight is less than a certain set threshold value; a predicted error rate is less than a certain threshold value; and a preset cycling times is reached. After the training process is stopped, the optimal weight and bias should be obtained at this time, and the network data can be saved for predicting and classifying the minerals. A core code is to input the training set into the neural network. Then the training is performed. This process is mainly to use the characteristics of the neural network to find an optimal input (first three items of the array) and output (the fourth item of the array) mapping. The main steps are as follows: 1. net=newff(P, [...], {...}) is used to create the neural network, wherein P is the input data, and the ellipsis represents other parameters. 2. net.trainParam.epochs=5000; and the training times is set. 3. net.trainParam.goal=O.0001; and a convergence error is set.
4. [Net, tr]=train(net); and the network is trained. Finally, the data of the test set is used to test the accuracy. After the training is completed, the test set is used as a parameter to be inputted into the trained neural network to study the accuracy of an output result. For example, the test data of
Description
gelenite is inputted to observe whether the output is gelenite. If the accuracy is low, the training and test are repeated. Step IV: the image collected in real time is inputted into the trained BP neural network model for identifying the metallic mineral. The currently extracted three-dimension statistics characteristics are brightness color characteristics of the image, which is related to the illumination intensity (not related to the instrument) during the photographing of the microscope. If the illumination intensity is different, the statistics characteristic of the image may be changed. The following method is adopted: at the beginning of the training of the network, a standard image is collected (marked as Al), and the characteristics of the standard image are extracted: F={x1, x2, x3, x4}. When the unknown mineral image is identified (the illumination intensity is changed during photographing), the image of the standard mineral Al is collected by changing the illumination intensity (or instrument), and the characteristic of the standard mineral (F'={Z, Z2, Z3, Z4}) Al is extracted; and then F'minus F to get a difference Y={yl, y2, y3, y4} (correction). The correction is added to the characteristic of the to-be-identified unknown mineral, thereby avoiding the problem of identification error caused by different illumination intensity or different instruments. Various embodiments of the present invention are described in a progressive manner. Each embodiment focuses on the difference from other embodiments, while same or similar parts of all embodiments can be referred to each other. Systems disclosed in embodiments are simply described because they correspond to the methods disclosed in the embodiments, and relevant information can be referred to the description of the method part. The principle and embodiments of the present invention are described herein through specific examples. The above embodiments are explained to help the understanding of the method and core concept of the present invention. Meanwhile, for those ordinary skilled in the art, according to the concept of the present invention, the specific embodiments and application ranges may be changed. In conclusion, the description of the present invention shall not be construed as limiting the present invention.
Claims (5)
1. A method for identifying metallic minerals under a microscope based on a BP neural network, comprising the following steps: acquiring an unknown metallic mineral image photographed by a reflection polarizing microscope; processing the unknown metallic mineral image by using MATLAB software to obtain primary characteristic data of the unknown metallic mineral image; determining environmental difference characteristics, wherein the environmental difference characteristics are differences between first characteristic data and second characteristic data; the first characteristic data is characteristic data of a first metallic mineral image under illumination intensity when the unknown metallic mineral image is photographed; the second characteristic data is the characteristic data of a second metallic mineral image under the illumination intensity when a BP neural network model is trained; and the illumination density of the first metallic mineral image is different from the illumination intensity of the second metallic mineral image, and the metallic mineral types are same; calculating final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic; inputting the final characteristic data of the unknown metallic mineral image into the trained BP neural network model, and determining the type of the unknown metallic mineral, wherein the input of the trained BP neural network model is the final characteristic data of the unknown metallic mineral image, and the output of the trained BP neural network model is the type of the metallic mineral.
2. The method for identifying metallic minerals under the microscope based on the BP neural network according to claim 1, wherein processing the unknown metallic mineral image by using the MATLAB software to obtain primary characteristic data of the unknown metallic mineral image specifically comprises: reading the unknown metallic mineral image into the MATLAB software by using an imread function, wherein a storage form of the unknown metallic mineral image in the MATLAB software is a three-dimensional array form, and the three-dimensional array form comprises RGB color information of the unknown metallic mineral image; sequentially performing the average filtering and normalization for the unknown metallic mineral image;
Claims
extracting the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image.
3. The method for identifying metallic minerals under the microscope based on the BP neural network according to claim 2, wherein extracting the characteristics of the normalized unknown metallic mineral image to obtain the primary characteristic data of the unknown metallic mineral image specifically comprises: performing the dimension separation for the normalized unknown metallic mineral image to obtain three one-dimension arrays, wherein the three one-dimension arrays are respectively a first one-dimension array, a second one-dimension array, and a third one-dimension array; calculating an average value of each of the one-dimension arrays by using a mean function; determining the primary characteristic data of the unknown metallic mineral image according to the average value of all the one-dimension arrays, wherein the primary characteristic data of the unknown metallic mineral image comprises four elements; the first element is the average value of the first one-dimension array; the second element is the average value of the second one-dimension array; the third element is the average value of the third one-dimension array; and the fourth element is the type of the unknown metallic mineral.
4. The method for identifying metallic minerals under the microscope based on the BP neural network according to claim 1, wherein calculating final characteristic data of the unknown metallic mineral image according to the primary characteristic data of the unknown metallic mineral image and the environmental difference characteristic specifically comprises: superimposing the environmental difference characteristics into the primary characteristic data of the unknown metallic mineral image to obtain the final characteristic data of the unknown metallic mineral image.
5. The method for identifying metallic minerals under the microscope based on the BP neural network according to claim 1, wherein training the BP neural network model comprises the following steps: acquiring different types of historical metallic mineral images under the same illumination intensity; processing all historical metallic mineral images by using the MATLAB software; grouping the processed historical metallic mineral images to obtain a training set and a test set;
Claims training the BP neural network model by using the data in the training set, stopping training the BP neural network model after a set training stop condition is satisfied, and obtaining a primary BP neural network model; testing the primary BP neural network model by using the data in the test set, stopping training the BP neural network model after a set test condition is satisfied, and saving the primary BP neural network model satisfying the test condition, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model; a historical metallic mineral image processing unit is used to process all historical metallic mineral images by using the MATLAB software; a training set and test set determination unit is used to group the processed historical metallic mineral images to obtain a training set and a test set; a primary BP neural network model determination unit is used to train the BP neural network model by using data in the training set, stop training the BP neural network model after a set training stop condition is satisfied, and obtain a primary BP neural network model; a trained BP neural network model determination unit is used to test the primary BP neural network model by using data in the test set, stop training the BP neural network model after a set test condition is satisfied, and save the primary BP neural network model satisfying the test condition, wherein the primary BP neural network model satisfying the test condition is the trained BP neural network model.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010216536.2A CN111460947B (en) | 2020-03-15 | 2020-03-15 | BP neural network-based method and system for identifying metal minerals under microscope |
| CN202010216536.2 | 2020-03-15 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2020103251A4 true AU2020103251A4 (en) | 2021-01-14 |
Family
ID=71685691
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2020103251A Ceased AU2020103251A4 (en) | 2020-03-15 | 2020-11-05 | Method and system for identifying metallic minerals under microscope based on bp nueral network |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111460947B (en) |
| AU (1) | AU2020103251A4 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113128404A (en) * | 2021-04-20 | 2021-07-16 | 中国地质科学院 | Intelligent mineral identification method and system |
| CN113935966A (en) * | 2021-10-13 | 2022-01-14 | 东风本田发动机有限公司 | Method, device, equipment and storage medium for detecting slag point of metal material |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111950427B (en) * | 2020-08-06 | 2022-02-15 | 中国地质科学院矿产资源研究所 | Garnet subclass identification method based on thermal infrared spectrum characteristics and BP neural network model |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1160249C (en) * | 2001-04-02 | 2004-08-04 | 吉林大学 | Mesoporous molecular sieve material with strong acidity and high hydrothermal stability and its prepn |
| US9082166B2 (en) * | 2011-12-30 | 2015-07-14 | Abbott Point Of Care, Inc. | Method and apparatus for automated platelet identification within a whole blood sample from microscopy images |
| KR20200037816A (en) * | 2017-08-02 | 2020-04-09 | 스트롱 포스 아이오티 포트폴리오 2016, 엘엘씨 | Methods and systems for detection in an industrial Internet of Things data collection environment with large data sets |
| CN108573205A (en) * | 2017-09-26 | 2018-09-25 | 中科智文（北京）科技有限公司 | A kind of Khotan jade seed material micro-image intelligent identifying system |
| CN110516954A (en) * | 2019-08-23 | 2019-11-29 | 昆明理工大学 | One kind referring to calibration method based on GA-BP neural network algorithm optimization mineral processing production |
-
2020
- 2020-03-15 CN CN202010216536.2A patent/CN111460947B/en active Active
- 2020-11-05 AU AU2020103251A patent/AU2020103251A4/en not_active Ceased
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113128404A (en) * | 2021-04-20 | 2021-07-16 | 中国地质科学院 | Intelligent mineral identification method and system |
| CN113935966A (en) * | 2021-10-13 | 2022-01-14 | 东风本田发动机有限公司 | Method, device, equipment and storage medium for detecting slag point of metal material |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111460947A (en) | 2020-07-28 |
| CN111460947B (en) | 2022-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2020103251A4 (en) | Method and system for identifying metallic minerals under microscope based on bp nueral network | |
| CN106910192B (en) | Image fusion effect evaluation method based on convolutional neural network | |
| CN106920224B (en) | A method of assessment stitching image clarity | |
| CN106875373B (en) | Mobile phone screen MURA defect detection method based on convolutional neural network pruning algorithm | |
| EP2161555B1 (en) | Coating color database creating method, search method using the database, their system, program, and recording medium | |
| CN101828201B (en) | Image processing device and method, and learning device, method | |
| CN106920215B (en) | Method for detecting registration effect of panoramic image | |
| WO2016197026A1 (en) | Full reference image quality assessment based on convolutional neural network | |
| CN112991371B (en) | Automatic image coloring method and system based on coloring overflow constraint | |
| CN111161224A (en) | Casting internal defect grading evaluation system and method based on deep learning | |
| MSB et al. | Non-destructive watermelon ripeness determination using image processing and artificial neural network (ANN) | |
| CN115018711B (en) | Image super-resolution reconstruction method for warehouse scheduling | |
| CN106651829A (en) | Non-reference image objective quality evaluation method based on energy and texture analysis | |
| CN114021704B (en) | AI neural network model training method and related device | |
| CN116994295B (en) | Wild animal category identification method based on gray sample self-adaptive selection gate | |
| CN113256733A (en) | Camera spectral sensitivity reconstruction method based on confidence voting convolutional neural network | |
| GB2604706A (en) | System and method for diagnosing small bowel cleanliness | |
| US20220309612A1 (en) | Image processing apparatus, image processing method, and storage medium | |
| CN114255193A (en) | Board card image enhancement method, device, equipment and readable storage medium | |
| CN113850335A (en) | Data augmentation method for bathroom ceramic defect detection | |
| Dávila-Meza et al. | Quaternion and split quaternion neural networks for low-light color image enhancement | |
| JP3411847B2 (en) | Color reproduction method and apparatus for input device, and recording medium recording program of this method | |
| CN111091532A (en) | Remote sensing image color evaluation method and system based on multilayer perceptron | |
| CN114549402A (en) | Underwater image quality comparison method without reference image | |
| CN116188797B (en) | Scene light source color estimation method capable of being effectively embedded into image signal processor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| MK22 | Patent ceased section 143a(d), or expired - non payment of renewal fee or expiry |