CN112950632A - Coal quality detection method based on hyperspectral imaging technology and convolutional neural network - Google Patents
Coal quality detection method based on hyperspectral imaging technology and convolutional neural network Download PDFInfo
- Publication number
- CN112950632A CN112950632A CN202110415234.2A CN202110415234A CN112950632A CN 112950632 A CN112950632 A CN 112950632A CN 202110415234 A CN202110415234 A CN 202110415234A CN 112950632 A CN112950632 A CN 112950632A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- output
- coal
- sample
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention provides a coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network, and belongs to the technical field of hyperspectral detection and coal detection. The specific method comprises the following steps: firstly, coal blocks are crushed into coal particles with the particle size of 13mm, the coal particles are uniformly paved at the bottom of a vessel and placed in the center of an objective table, and image acquisition is carried out through a hyperspectral camera. The method comprises the steps of firstly carrying out black and white correction on a hyperspectral image to remove the influence of dark current, and inputting the corrected image into a convolutional neural network to carry out convolution and pooling operation to extract features and classify. And temporarily taking the sample value obtained after classification as the median value of the content cell in which the sample value is located as the input of a BP neural network regression model, taking the measured value obtained by industrial analysis as expected output, updating through reverse iteration until the error reaches a set value, stopping iteration, and outputting the content value. The method realizes the quality detection of coal, has simple operation and provides corresponding technical guidance for the coal selection process.
Description
Technical Field
The invention relates to the technical field of coal quality detection, in particular to a coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network.
Background
Coal occupies a very important position in our daily production life, and although part of coal is replaced by oil as the oil resources are continuously developed at present, the indispensable position of coal is gradually shown as the oil resources are gradually reduced later.
Coal is a non-renewable energy source, which is formed by a series of evolution processes that a large number of plants died hundreds of millions of years ago and were buried underground. Different plants, different conditions and different evolution processes all cause the coal to have different internal component contents and different qualities, thereby causing different heat release when the coal is completely combusted. Therefore, in order to make full use of the coal, which is a non-renewable energy source, it is important to detect the quality of the coal.
The main components of the coal are fixed carbon, and in addition, the coal also has ash content, moisture content, volatile components, sulfur and the like, most of sulfur components in the combustion chamber of the coal can be oxidized into sulfur dioxide, so that the sulfur dioxide can cause certain damage to human bodies and the environment, and organic matters in the coal can be heated and decomposed into combustible gas under certain temperature conditions, namely volatile components. Therefore, the quality of coal is related to the content of each component in the coal.
At present, a plurality of methods exist for detecting the quality of coal. The most traditional quality detection method is to obtain the specific content value of each component in the coal by combining a burning method with a chemical reagent. And coal heat regression by combining a near infrared spectroscopy method and a principal component analysis method, coal heat value quantitative analysis by laser-induced breakdown spectroscopy and the like. Besides, the method also adopts near infrared spectrum to detect ash content and moisture content of coal, and adopts laser-induced breakdown spectroscopy technology to detect volatile components of coal.
Disclosure of Invention
The invention provides a coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network. The method comprises the steps of collecting 13mm raw coal sample images through a hyperspectral imaging technology, removing interference caused by dark current through black and white correction, then performing feature extraction and classification on the images as input of a convolutional neural network, outputting corresponding intervals of ash content, moisture content and volatile components, inputting median values of the intervals corresponding to different components into a BP neural network to establish a regression model, and obtaining specific contents of the ash content, the moisture content and the volatile components of coal, so that the quality detection of the coal is realized. The method adopts the hyperspectral imaging technology for analysis, and the information expression is richer due to the combined characteristic of the hyperspectral images and the maps, so that the reflection spectrum of the sample can be finely depicted; secondly, a convolutional neural network is applied to process a reduced complex preprocessing process, and meanwhile, the relative initial performance of subsequent regression processing is smaller and the convergence is faster due to the result of fine classification.
The invention is realized by the following technical scheme: a coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network comprises the steps of preparing a coal sample, and grinding a raw coal block sample into a raw coal particle sample with the particle size of 13mm through a grinder. All samples with the same number are uniformly paved in a vessel and placed in the center of an objective table, and due to the fact that components in coal are not uniformly distributed, if a local sample is intercepted, all information of the sample cannot be sufficiently expressed, and therefore in order to guarantee integrity and representativeness of sample information, hyperspectral image acquisition needs to be carried out on the whole sample, and operation is carried out through hyperspectral imaging equipment. Firstly, performing black-and-white correction on an acquired image, removing the influence of dark current, transmitting the corrected image to a convolutional neural network for feature extraction, continuously performing convolutional pooling to obtain the features and information of different components of a sample, inputting the features and information into a full-link layer for classification, inputting the median values of intervals corresponding to the different components into a BP neural network for regression, and finally obtaining and outputting the specific content of each component of the coal sample.
A coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network is characterized by mainly comprising the following steps:
s1, grinding the raw coal blocks into raw coal particles with the particle size of 13mm by a grinder, uniformly paving the raw coal particles at the bottom of a container, placing the container in the center of an objective table, placing a hyperspectral imaging device right above a sample, adjusting the height of the device at the same time, ensuring that the hyperspectral imaging device can acquire hyperspectral information of all sample particles, turning on a light source and acquiring hyperspectral images;
s2, in order to overcome the influences of nonuniformity of the intensity of the light source under each wave band and dark current in the collecting sensor, performing black and white correction processing on the collected coal particle sample hyperspectral image;
and S3, classifying the samples according to the ash content, the moisture content and the volatile content which are actually measured in the industrial analysis according to certain interval intervals. Dividing the moisture interval into 5 intervals, wherein the interval is 2% from 0%; dividing an ash interval into 5 intervals, wherein the interval is 10% from 0%; the volatile matter was partitioned into 5 intervals, starting from 0% and the interval was 10%. Establishing a convolutional neural network model, transmitting the black-white corrected hyperspectral image of the coal particle sample to a convolutional neural network for feature extraction, inputting the hyperspectral image into a full-link layer after several convolution and pooling steps, and finally classifying and outputting the hyperspectral image through softmax;
and S4, outputting classification intervals of corresponding ash, moisture and volatile components instead of specific contents after classification by the convolutional neural network, and further regressing to obtain the ash, moisture and volatile components. Because the corresponding interval is a range value, the median output of each interval is selected as a temporary value of the sample component and is used as the input of the BP neural network, and the initial error in iteration is relatively small;
and S5, taking the median of the interval where each component of the classified sample is located as the input of the BP neural network, automatically updating the weight and the threshold value through continuous iteration, gradually reducing the error until the training frequency reaches the maximum or the iteration termination condition is reached, stopping iteration when the training precision meets the requirement, finally obtaining the specific content of each component, and realizing the quality detection of the coal.
Drawings
FIG. 1: structure of hyperspectral imaging system
FIG. 2: convolution neural network model structure chart
FIG. 3: structure diagram of BP neural network model
Detailed Description
A coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network comprises the following specific implementation steps:
s1, crushing a coal briquette sample of raw coal by using a crusher to obtain coal particles 3 with the particle size of 13mm, uniformly paving the coal particle sample in a vessel, placing the coal particle sample in the central position of a hyperspectral objective table 4, placing hyperspectral imaging equipment right above the sample, and adjusting the height of a camera to ensure that the hyperspectral camera can collect all samples. And (3) turning on the light source 2, performing staring shooting by the hyperspectral camera 1 to obtain a hyperspectral image of the coal particle sample, and storing the hyperspectral image into a computer, as shown in figure 1.
S2, since there is some noise interference due to the uneven intensity distribution of the light emitted from the light source 2 and the influence of the dark current of the collection sensor, in order to reduce and eliminate the influence of the noise, the collected hyperspectral image of the coal particle sample is corrected in black and white.
Acquiring a white board calibration image and a blackboard calibration image, calibrating the acquired original coal particle image by using the white board image and the blackboard image, and obtaining a calibrated coal sample image according to the following formula:
c in the above formula is a coal particle sample image obtained after black and white calibration, IwCalibrating the whiteboard for images, IdCalibrating the blackboard for images, IrIs the original coal particle image.
S3, classifying the samples according to the ash content, the moisture content and the volatile content of the samples at certain intervals, establishing a convolutional neural network model, as shown in FIG. 2, transmitting the black-white corrected coal particle sample hyperspectral images to a convolutional neural network for convolution and pooling operation, inputting the coal particle sample hyperspectral images to a full-link layer after several convolution and pooling steps, and finally classifying and outputting the coal particle sample hyperspectral images through softmax;
s3.1, carrying out interval division in advance according to the ash content value, the moisture content value and the volatile component content value which are measured in the practical industrial analysis and are concentrated in the coal sample, dividing the moisture interval into 5 intervals, and starting from 0%, the interval is 2%; dividing an ash interval into 5 intervals, wherein the interval is 10% from 0%; the volatile matter was partitioned into 5 intervals, starting from 0% and the interval was 10%.
And S3.2, inputting the hyperspectral image subjected to black and white correction into a convolutional neural network for feature extraction and classification. The structure of the convolutional neural network is shown in fig. 2, and has 7 layers, and the specific structure is described as follows:
the input layer is a hyperspectral image of coal with 256 channels of size 224 x 224.
The first layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the input coal hyperspectral image by 96 convolution kernels with the size of 7 x 7 and the step length of 2; the pooling layer was pooled using pooling windows of size 3 x 3 with step size 2.
The second layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the pooling result of the previous layer by using 256 convolution kernels with the size of 5 x 5 and the step length of 2; the pooling layer was pooled using pooling windows of size 3 x 3 with step size 2.
The third layer is a convolution layer, and the pooling result of the previous layer is filtered by 384 convolution cores with a size of 3 × 3 and a step size of 1.
The fourth layer is a convolution layer, and the convolution result of the previous layer is filtered by 384 convolution cores with the size of 3 x 3 and the step size of 1.
The fifth layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the convolution result of the previous layer by using 256 convolution kernels with the size of 3 x 3 and the step length of 1; the pooling layer was pooled using pooling windows of size 3 x 3 with step size 2.
The sixth layer is a full link layer FC.
The seventh layer is a full link layer FC.
Wherein, a correction linear unit ReLU activation function is used for the convolution result to obtain an activation result; and using local response normalization operation on the pooling result to obtain a normalization result.
The expression for the correction linearity unit is:
ReLU(x)＝max(0,x)
the specific meaning of the formula is that when x is less than or equal to 0, ReLU (x) is 0; if x > 0, relu (x) x.
The output adopts a soft maximum output function softmax:
in the above formula, x1,x2,…,xnRepresents all model results, exp (x)i) Indicating that the model results are converted to non-negative numbers. The effect of the Softmax function is to convert the model results into probabilities, with the sum of the probabilities for all results being 1.
The loss function uses cross entropy as the loss function:
in the above formula, tiAn indicator variable (0 or 1) which is 1 if the class is the same as the class of the sample, and 0 otherwise; y isiRepresenting the probability of a sample belonging to class i.
And finally, obtaining the classification result of each component obtained by the output of the convolutional neural network, and calculating the median of the classification result in the cell corresponding to each classification result to be used as the temporary content value of the component.
S4, after the classification of the convolutional neural network, corresponding classification intervals of ash, moisture and volatile components are located, but not specific contents, so that specific content values of the components need to be further regressed, wherein the median values of the intervals where the ash, moisture and volatile components are located among the cells output by the classification of the convolutional neural network are temporarily taken as temporary values of the components;
s5, taking the temporary content values of the components of the samples classified by the convolutional neural network model as the input of a BP neural network, establishing a BP neural network regression model for regression, wherein the structural model of the BP neural network is shown in figure 3, updating the weight and the threshold value through reverse iterative learning, gradually reducing the error until the training frequency reaches the maximum or the iteration termination condition is reached, and stopping iteration, namely stopping iteration when the training precision meets the requirement; the method comprises the following concrete steps:
establishing a BP neural network regression model, wherein the BP neural network is divided into three layers, namely an input layer, a hidden layer and an output layer, the input layer is provided with three input neurons, the hidden layer is provided with five hidden neurons, and the output layer is provided with three output neurons.
Firstly, randomly assigning initial values to a weight value from an input layer to a hidden layer, a weight value from the hidden layer to an output layer, a threshold value of the hidden layer in the neural network and a threshold value of the output layer in a (0, 1) interval.
And secondly, giving temporary values of the content of each component output by the convolutional neural network as input, and taking actual content values of the components determined by industrial analysis as reference output. And training the BP neural network by using 255 sample data as a training set and 45 sample data as a test set, reversely iterating and updating the weight and the threshold value according to the error between the output value and the reference output value until the precision requirement is met, and stopping iteration.
As shown in the structure diagram of BP neural network in FIG. 3, the weight from the ith neuron of the input layer to the jth neuron of the hidden layer is vijThe weight value from the jth neuron of the hidden layer to the kth neuron of the output layer is wjkThe threshold for the jth hidden layer neuron is γjThe threshold for the kth output layer neuron is θk。
Wherein the input to the jth hidden neuron is:
the input to the kth output layer neuron is:
wherein, b'j＝f(bj-γj) Is the output of the hidden layer neuron, and the output of the output layer neuron is y'k＝f(yk-θk)。
f [ beta ] takes a differentiable sigmoid function:
calculating an objective function J of the network: p group of samples xp＝[x1，x2，x3]TAt the input, the objective function of the network is
Wherein, ypIs the desired output of the sample, ypAnd (t) is the output of the network after the sample is subjected to the weight adjustment for the t time.
The total objective function of the output of the BP neural network is:
and taking the total objective function of the network as the condition evaluation of network learning, stopping iteration if J (t) is less than or equal to epsilon, finishing the network training, and otherwise, continuing to perform back propagation calculation and adjusting the weight and the threshold.
And the back propagation calculation is carried out layer by layer according to a gradient descent method:
firstly, adjusting the weight and the threshold from the hidden layer to the output layer:
Δwjk＝η1g(t)bj+αΔwjk(t-1)
Δθk＝η1g(t)+αΔθk(t-1)
and then adjusting the weight and the threshold from the input layer to the hidden layer:
Δvij＝η2qj(t)xi+αΔvij(t-1)
Δγj＝η2qj(t)+αΔγj(t-1)
wherein eta is1And η2For the learning rate, g (t) and q (t) are partial derivatives after t iterations, and α is a forgetting factor. In this embodiment, the maximum number of training passes is 1000, the objective function is 0.001, and the learning rate is set to 0.01. In the BP neural network regression process, the coefficient R is determined2As an evaluation index of regression, the expression is:
wherein y'kIs the output value of the network and is,
is the average value of the outputs of the network,
the output expectation value of the network. The output values of the decision coefficients are in the range of (0, 1) interval, and the closer the values are to 1, the better the regression fitting degree is.
Claims (1)
1. A coal quality detection method based on a hyperspectral imaging technology and a convolutional neural network is characterized by comprising the following steps:
s1, crushing a coal briquette sample of raw coal by using a crusher to obtain coal particles with the particle size of 13mm, uniformly paving the coal particle sample in a vessel, placing the coal particle sample at the central position of a hyperspectral objective table, placing hyperspectral imaging equipment right above the sample, and adjusting the height of a camera to ensure that the hyperspectral camera can collect all samples; turning on a light source, performing staring shooting through a hyperspectral camera to obtain a hyperspectral image of the coal particle sample, and storing the hyperspectral image into a computer;
s2, due to the fact that the intensity distribution of light source illumination is uneven and the influence of dark current existing in the collection sensor can cause noise interference, in order to weaken and eliminate the influence of the noise, black and white correction is conducted on the collected coal particle sample hyperspectral image;
acquiring a white board calibration image and a blackboard calibration image, calibrating the acquired original coal particle image by using the white board image and the blackboard image, and obtaining a calibrated coal sample image according to the following formula:
wherein C is coal particle sample image obtained after black and white calibration, IwCalibrating the whiteboard for images, IdCalibrating the blackboard for images, IrIs an original coal particle image;
s3, classifying the samples according to the ash content, the moisture content and the volatile content of the samples at certain intervals, establishing a convolutional neural network model, transmitting the black-white corrected coal particle sample hyperspectral images to a convolutional neural network for convolution and pooling, inputting the coal particle sample hyperspectral images to a full connection layer after several convolution and pooling steps, and finally performing softmax classified output;
s3.1, carrying out interval division in advance according to the ash content value, the moisture content value and the volatile component content value which are measured in the practical industrial analysis and are concentrated in the coal sample, dividing the moisture interval into 5 intervals, and starting from 0%, the interval is 2%; dividing an ash interval into 5 intervals, wherein the interval is 10% from 0%; dividing the volatile components into 5 intervals, wherein the interval is 10% from 0%;
s3.2, inputting the hyperspectral image subjected to black and white correction into a convolutional neural network for feature extraction and classification; the convolutional neural network has 7 layers, and the specific structure is as follows:
the input layer is a hyperspectral image of coal with 256 channels of size 224 x 224;
the first layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the input coal hyperspectral image by 96 convolution kernels with the size of 7 x 7 and the step length of 2; the pooling layer is subjected to pooling by using a pooling window with the size of 3 × 3 and the step length of 2;
the second layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the pooling result of the previous layer by using 256 convolution kernels with the size of 5 x 5 and the step length of 2; the pooling layer is subjected to pooling by using a pooling window with the size of 3 × 3 and the step length of 2;
the third layer is a convolution layer, and 384 convolution cores with the size of 3 x 3 and the step length of 1 are used for filtering the pooling result of the previous layer;
the fourth layer is a convolution layer, and the convolution result of the previous layer is filtered by 384 convolution cores with the size of 3 x 3 and the step length of 1;
the fifth layer comprises a convolution layer and a pooling layer, wherein the convolution layer filters the convolution result of the previous layer by using 256 convolution kernels with the size of 3 x 3 and the step length of 1; the pooling layer is subjected to pooling by using a pooling window with the size of 3 × 3 and the step length of 2;
the sixth layer is a full connection layer FC;
the seventh layer is a full connection layer FC;
wherein, a correction linear unit ReLU activation function is used for the convolution result to obtain an activation result; using local response normalization operation on the pooling result to obtain a normalization result;
the expression for the correction linearity unit is:
ReLU(x)＝max(0,x)
the specific meaning of the formula is that when x is less than or equal to 0, ReLU (x) is 0; if x > 0, relu (x) x;
the output adopts a soft maximum output function softmax:
in the above formula, x1,x2,…,xnRepresents all model results, exp (x)i) The model result is converted into a nonnegative number, the Softmax function is used for converting the model result into the probability, and the sum of the probabilities of all the results is 1;
the loss function uses cross entropy as the loss function:
in the above formula, tiAn indicator variable (0 or 1) which is 1 if the class is the same as the class of the sample, and 0 otherwise; y isiRepresenting the probability of a sample belonging to class i;
finally, obtaining classification results of each component through the output of the convolutional neural network, and solving a median value of the classification results in a cell corresponding to each classification result to be used as a temporary content value of the component;
s4, after the classification of the convolutional neural network, corresponding classification intervals of ash, moisture and volatile components are located, but not specific contents, so that specific content values of the components need to be further regressed, wherein the median values of the intervals where the ash, moisture and volatile components are located among the cells output by the classification of the convolutional neural network are temporarily taken as temporary values of the components;
s5, taking the temporary content values of each component of the sample classified by the convolutional neural network model as the input of a BP neural network, establishing a BP neural network regression model for regression, updating the weight and the threshold value through reverse iterative learning, gradually reducing the error until the training frequency reaches the maximum or the iteration termination condition stops iteration, namely stopping iteration when the training precision reaches the requirement, and specifically implementing the steps as follows;
establishing a BP neural network regression model, wherein the BP neural network is divided into three layers, namely an input layer, a hidden layer and an output layer, the input layer is provided with three input neurons, the hidden layer is provided with five hidden neurons, and the output layer is provided with three output neurons;
firstly, randomly assigning initial values to a weight value from an input layer to a hidden layer, a weight value from the hidden layer to an output layer, a threshold value of the hidden layer in a neural network and a threshold value of the output layer in a (0, 1) interval;
secondly, giving temporary content values of all components output by the convolutional neural network as input, and taking actual content values of the components determined by industrial analysis as reference output; training the BP neural network by using 255 sample data as a training set and 45 sample data as a test set, reversely iterating and updating a weight value and a threshold value according to an error between an output value and a reference output value until a precision requirement is met, and stopping iteration;
the weight value from the ith neuron of the BP neural network input layer to the jth neuron of the hidden layer is vijThe weight value from the jth neuron of the hidden layer to the kth neuron of the output layer is wjkThe threshold for the jth hidden layer neuron is γjThe threshold for the kth output layer neuron is θk；
Wherein the input to the jth hidden neuron is:
the input to the kth output layer neuron is:
wherein, b'j＝f(bj-γj) Is the output of the hidden layer neuron, and the output of the output layer neuron is y'k＝f(yk-θk)；
f [ beta ] takes a differentiable sigmoid function:
calculating an objective function J of the network: p group of samples xp＝[x1，x2，x3]TAt the input, the objective function of the network is
Wherein, ypIs the desired output of the sample, yp(t) is the output of the network after the sample is subjected to the weight adjustment for the t time;
the total objective function of the output of the BP neural network is:
taking the total objective function of the network as the condition evaluation of network learning, if J (t) is less than or equal to epsilon, stopping iteration, completing network training, otherwise, continuing to perform back propagation calculation, and adjusting weight and threshold;
and the back propagation calculation is carried out layer by layer according to a gradient descent method:
firstly, adjusting the weight and the threshold from the hidden layer to the output layer:
Δwjk＝η1g(t)bj+αΔwjk(t-1)
Δθk＝η1g(t)+αΔθk(t-1)
and then adjusting the weight and the threshold from the input layer to the hidden layer:
Δvij＝η2qj(t)xi+αΔvij(t-1)
Δγj＝η2qj(t)+αΔγj(t-1)
wherein eta is1And η2For the learning rate, g (t) and q (t) are partial derivatives after t iterations, and alpha is a forgetting factor; in this embodiment, the maximum training number is 1000, the objective function is 0.001, and the learning rate is set to 0.01; in BP neural network regression processIn order to determine the coefficient R2As an evaluation index of regression, the expression is:
wherein y'kIs the output value of the network and is,
is the average value of the outputs of the network,
an output expected value for the network; the output values of the decision coefficients are in the range of (0, 1) interval, and the closer the values are to 1, the better the regression fitting degree is.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110415234.2A CN112950632B (en) | 2021-04-18 | 2021-04-18 | Coal quality detection method based on hyperspectral imaging technology and convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110415234.2A CN112950632B (en) | 2021-04-18 | 2021-04-18 | Coal quality detection method based on hyperspectral imaging technology and convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112950632A true CN112950632A (en) | 2021-06-11 |
| CN112950632B CN112950632B (en) | 2022-10-14 |
Family
ID=76232970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110415234.2A Active CN112950632B (en) | 2021-04-18 | 2021-04-18 | Coal quality detection method based on hyperspectral imaging technology and convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112950632B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113533220A (en) * | 2021-07-25 | 2021-10-22 | 吉林大学 | Coal type identification method based on hyperspectral imaging technology |
| CN113723534A (en) * | 2021-09-02 | 2021-11-30 | 南京润内克西信息科技有限公司 | Urban raise dust on-line monitoring system based on BP neural network |
| CN113959961A (en) * | 2021-12-22 | 2022-01-21 | 广东省农业科学院动物科学研究所 | Hyperspectral image-based tannin additive anti-counterfeiting detection method and system |
| CN114897835A (en) * | 2022-05-16 | 2022-08-12 | 中国矿业大学 | Coal product ash content real-time detection system based on image |
| CN115375675A (en) * | 2022-10-24 | 2022-11-22 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Coal quality detection method based on image data |
| CN116559119A (en) * | 2023-05-11 | 2023-08-08 | 东北林业大学 | Deep learning-based wood dyeing color difference detection method, system and medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680081A (en) * | 2017-09-08 | 2018-02-09 | 西安电子科技大学 | High spectrum image solution mixing method based on convolutional neural networks |
| CN107727592A (en) * | 2017-10-10 | 2018-02-23 | 中国矿业大学 | A kind of coal-rock interface identification method based on coal petrography high spectrum reflection characteristic |
| CN108491849A (en) * | 2018-03-23 | 2018-09-04 | 上海理工大学 | Hyperspectral image classification method based on three-dimensional dense connection convolutional neural networks |
| CN108596213A (en) * | 2018-04-03 | 2018-09-28 | 中国地质大学（武汉） | A kind of Classification of hyperspectral remote sensing image method and system based on convolutional neural networks |
| CN109766858A (en) * | 2019-01-16 | 2019-05-17 | 中国人民解放军国防科技大学 | Three-dimensional convolution neural network hyperspectral image classification method combined with bilateral filtering |
| CN111007021A (en) * | 2019-12-31 | 2020-04-14 | 北京理工大学重庆创新中心 | Hyperspectral water quality parameter inversion system and method based on one-dimensional convolution neural network |
| US20200298401A1 (en) * | 2019-01-03 | 2020-09-24 | Lucomm technologies, Inc | Semantic Robotic System |
| US20200302224A1 (en) * | 2019-03-21 | 2020-09-24 | Illumina, Inc. | Artificial Intelligence-Based Sequencing |
| CN111754028A (en) * | 2020-06-08 | 2020-10-09 | 吉林大学 | Hyperspectrum-based coal ash content and moisture detection system and method |
| CN112465063A (en) * | 2020-12-11 | 2021-03-09 | 中国矿业大学 | Coal gangue identification method in top coal caving process based on multi-sensor information fusion |
| CN112488082A (en) * | 2020-12-24 | 2021-03-12 | 刘强 | Coal gangue intelligent sorting system based on deep learning |
-
2021
- 2021-04-18 CN CN202110415234.2A patent/CN112950632B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680081A (en) * | 2017-09-08 | 2018-02-09 | 西安电子科技大学 | High spectrum image solution mixing method based on convolutional neural networks |
| CN107727592A (en) * | 2017-10-10 | 2018-02-23 | 中国矿业大学 | A kind of coal-rock interface identification method based on coal petrography high spectrum reflection characteristic |
| CN108491849A (en) * | 2018-03-23 | 2018-09-04 | 上海理工大学 | Hyperspectral image classification method based on three-dimensional dense connection convolutional neural networks |
| CN108596213A (en) * | 2018-04-03 | 2018-09-28 | 中国地质大学（武汉） | A kind of Classification of hyperspectral remote sensing image method and system based on convolutional neural networks |
| US20200298401A1 (en) * | 2019-01-03 | 2020-09-24 | Lucomm technologies, Inc | Semantic Robotic System |
| CN109766858A (en) * | 2019-01-16 | 2019-05-17 | 中国人民解放军国防科技大学 | Three-dimensional convolution neural network hyperspectral image classification method combined with bilateral filtering |
| US20200302224A1 (en) * | 2019-03-21 | 2020-09-24 | Illumina, Inc. | Artificial Intelligence-Based Sequencing |
| CN111007021A (en) * | 2019-12-31 | 2020-04-14 | 北京理工大学重庆创新中心 | Hyperspectral water quality parameter inversion system and method based on one-dimensional convolution neural network |
| CN111754028A (en) * | 2020-06-08 | 2020-10-09 | 吉林大学 | Hyperspectrum-based coal ash content and moisture detection system and method |
| CN112465063A (en) * | 2020-12-11 | 2021-03-09 | 中国矿业大学 | Coal gangue identification method in top coal caving process based on multi-sensor information fusion |
| CN112488082A (en) * | 2020-12-24 | 2021-03-12 | 刘强 | Coal gangue intelligent sorting system based on deep learning |
Non-Patent Citations (2)
| Title |
|---|
| 杨明花等: "基于高光谱图像和卷积神经网络的燃煤热值估计算法", 《中国电力》 * |
| 林赐云等: "GPS数据在线地图匹配与数据分析微实验设计与开发", 《当代教育实践与教学研究》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113533220A (en) * | 2021-07-25 | 2021-10-22 | 吉林大学 | Coal type identification method based on hyperspectral imaging technology |
| CN113723534A (en) * | 2021-09-02 | 2021-11-30 | 南京润内克西信息科技有限公司 | Urban raise dust on-line monitoring system based on BP neural network |
| CN113959961A (en) * | 2021-12-22 | 2022-01-21 | 广东省农业科学院动物科学研究所 | Hyperspectral image-based tannin additive anti-counterfeiting detection method and system |
| CN114897835A (en) * | 2022-05-16 | 2022-08-12 | 中国矿业大学 | Coal product ash content real-time detection system based on image |
| CN114897835B (en) * | 2022-05-16 | 2023-04-28 | 中国矿业大学 | Image-based real-time detection system for ash content of coal products |
| CN115375675A (en) * | 2022-10-24 | 2022-11-22 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Coal quality detection method based on image data |
| CN115375675B (en) * | 2022-10-24 | 2023-01-31 | 山东济矿鲁能煤电股份有限公司阳城煤矿 | Coal quality detection method based on image data |
| CN116559119A (en) * | 2023-05-11 | 2023-08-08 | 东北林业大学 | Deep learning-based wood dyeing color difference detection method, system and medium |
| CN116559119B (en) * | 2023-05-11 | 2024-01-26 | 东北林业大学 | Deep learning-based wood dyeing color difference detection method, system and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112950632B (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112950632B (en) | Coal quality detection method based on hyperspectral imaging technology and convolutional neural network | |
| CN110717368A (en) | Qualitative classification method for textiles | |
| CN111126511B (en) | Vegetation index fusion-based LAI quantitative model establishment method | |
| CN112881306A (en) | Hyperspectral image-based method for rapidly detecting ash content of coal | |
| CN107679569A (en) | Raman spectrum substance automatic identifying method based on adaptive hypergraph algorithm | |
| CN107895136B (en) | Coal mine area identification method and system | |
| CN105678793A (en) | Image-fusion-based early diagnosis method and apparatus for kiwifruit canker | |
| CN106529458A (en) | Deep neural network space spectrum classification method for high-spectral image | |
| CN115812634A (en) | Pre-incubation hatching egg fertilization information nondestructive sorting method and device | |
| CN114764682B (en) | Rice safety risk assessment method based on multi-machine learning algorithm fusion | |
| CN113533220A (en) | Coal type identification method based on hyperspectral imaging technology | |
| Wang et al. | Intelligent detection of hard seeds of snap bean based on hyperspectral imaging | |
| CN114660040A (en) | Method, device, medium and equipment for identifying microbial unicell species | |
| CN116312860B (en) | Agricultural product soluble solid matter prediction method based on supervised transfer learning | |
| Huang et al. | Optimal wavelength selection for hyperspectral scattering prediction of apple firmness and soluble solids content | |
| CN109447009B (en) | Hyperspectral image classification method based on subspace nuclear norm regularization regression model | |
| CN116203001A (en) | Aflatoxin B1 nondestructive testing method and device | |
| CN114998719A (en) | Forest fire prediction method based on deep learning and multi-source remote sensing data | |
| Li et al. | Early drought plant stress detection with bi-directional long-term memory networks | |
| CN114418027A (en) | Hyperspectral image characteristic waveband selection method based on waveband attention mechanism | |
| CN114755189A (en) | Feature-optimized self-attention mechanism hyperspectral satellite LAI inversion method | |
| CN109001244B (en) | Device and method for detecting potato quality based on infrared thermal imaging technology | |
| CN112697746A (en) | Spectrum information-based sweet potato powder alum content detection method | |
| Xiao et al. | Sausage quality classification of hyperspectral multi-data fusion based on machine learning | |
| CN114897835B (en) | Image-based real-time detection system for ash content of coal products |
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 |