A method and system for remote production monitoring of a biomass fuel

Abstract

The application discloses a kind of biomass fuel remote production monitoring method and system, including extracting the first seed feature of coarse particle area and the second seed feature of fine particle area from standard image;Collect the fuel particle image after being finely broken by pulverizer on biomass fuel production line and carry out gray scale processing, obtain first detection image;First seed feature and second seed feature are used to carry out region growth on first detection image respectively, obtain coarse particle area and fine particle area in first detection image, eliminate from first detection image to obtain second detection image;The particle density distribution characteristics of second detection image are calculated, and the coarse particle area and fine particle area of second detection image are divided;Integrate the coarse particle area in first detection image and second detection image, obtain the target coarse particle fuel to be finely broken.The application realizes efficient and accurate monitoring and classification to biomass fuel granularity, and significantly improves production efficiency and finished product quality.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy, in particular to a remote production monitoring method and system for biomass fuel. BACKGROUND

[0002] In the production process of biomass fuel, the monitoring of coarse and fine particle size is crucial. Once relatively coarse particle biomass raw materials are found, secondary crushing is usually required to obtain finer particle materials to avoid affecting the quality of the final product. Currently, the monitoring and screening of particle size mainly rely on manual visual inspection and traditional mechanical screening equipment. The manual visual inspection method is not only inefficient, but also easily affected by personal experience and subjective factors, leading to unstable particle size control. Although the traditional mechanical screening equipment can achieve a certain degree of particle size screening, it lacks intelligentization and real-time monitoring capability, and is difficult to adapt to the fine management and efficient monitoring needs of particle size on modern production lines. SUMMARY

[0003] In order to solve at least one of the above technical problems, the present application provides a remote production monitoring method and system for biomass fuel.

[0004] In a first aspect, the present application provides a remote production monitoring method for biomass fuel, the method comprising:

[0005] establishing a standard image database containing standard particles of different particle sizes, extracting first seed features of coarse particle regions and second seed features of fine particle regions from the standard images;

[0006] collecting fuel particle images after being finely crushed by a crusher on a biomass fuel production line, and performing grayscale processing on the fuel particle images to obtain a first detection image;

[0007] performing region growing segmentation on the first detection image using the first seed features and the second seed features respectively to obtain coarse particle regions and fine particle regions in the first detection image, and removing the coarse particle regions and the fine particle regions from the first detection image to obtain a second detection image to be classified;

[0008] calculating particle density distribution features of the second detection image, and dividing coarse particle regions and fine particle regions of the second detection image according to the particle density distribution features;

[0009] integrating coarse particle regions in the first detection image and the second detection image to obtain target coarse particle fuel to be finely crushed.

[0010] Preferably, the first seed features of coarse particle regions and the second seed features of fine particle regions extracted from the standard images comprise:

[0011] Image features of coarse-grained regions and fine-grained regions are extracted from a standard image. These image features include pixel grayscale features, texture features, and shape features.

[0012] The SAE algorithm is used to perform feature dimensionality reduction on image features in coarse-grained regions and fine-grained regions respectively, and the gradient boosting tree algorithm is used to fit the dimensionality-reduced features.

[0013] Select the features with the highest correlation after fitting, and extract the target image features of coarse-grained regions and fine-grained regions.

[0014] Feature fusion is performed on the target image features in the coarse-grained region to obtain the first seed feature; feature fusion is performed on the target image features in the fine-grained region to obtain the second seed feature.

[0015] Preferably, the step of calculating the particle density distribution characteristics of the second detection image and dividing the second detection image into coarse particle regions and fine particle regions based on the particle density distribution characteristics includes:

[0016] Calculate the pixel grayscale value of each pixel in the second detection image, and use it as the density value at the corresponding location;

[0017] The second detection image is traversed using a sliding window, and the average density within each window is calculated based on the density value.

[0018] ;

[0019] ;

[0020] In the formula, This represents the average density within each window. This indicates the coordinates of the top-left corner of the window. These are the width and height of the window, respectively. For the image in Density value at that location, For the image in The pixel grayscale value at that location;

[0021] The average density of all windows is compared with a preset threshold. When the average density is greater than the preset threshold, the area where the corresponding window is located is divided into a coarse-grained area. When the average density is less than or equal to the preset threshold, the area where the corresponding window is located is divided into a fine-grained area.

[0022] Secondly, the present invention also provides a remote production monitoring system for biomass fuel, the system comprising:

[0023] The feature extraction unit is configured to establish a standard image database containing standard particles of different particle sizes, and extract first seed features of coarse particle regions and second seed features of fine particle regions from the standard images;

[0024] The gray processing unit is configured to collect fuel particle images after the particles are finely crushed by a crusher on a biomass fuel production line, perform gray processing on the fuel particle images to obtain first detection images;

[0025] The image segmentation unit is configured to perform region growing segmentation on the first detection images using the first seed features and the second seed features respectively to obtain coarse particle regions and fine particle regions in the first detection images, and remove the coarse particle regions and the fine particle regions from the first detection images to obtain second detection images to be classified;

[0026] The region division unit is configured to calculate particle density distribution features of the second detection images, and divide coarse particle regions and fine particle regions of the second detection images according to the particle density distribution features;

[0027] The coarse particle integration unit is configured to integrate the coarse particle regions in the first detection images and the second detection images to obtain target coarse particle fuel to be finely crushed.

[0028] Preferably, the feature extraction unit is further configured to:

[0029] extract image features of coarse particle regions and image features of fine particle regions from the standard images, wherein the image features include pixel gray features, texture features and shape features;

[0030] perform feature dimension reduction on the image features of the coarse particle regions and the image features of the fine particle regions respectively using the SAE algorithm, and perform fitting on the reduced features using the gradient boosting tree algorithm;

[0031] screen and select feature quantities with the largest correlation after fitting, and extract target image features of the coarse particle regions and target image features of the fine particle regions;

[0032] perform feature fusion on the target image features of the coarse particle regions to obtain the first seed features, and perform feature fusion on the target image features of the fine particle regions to obtain the second seed features.

[0033] Preferably, the region division unit is further configured to:

[0034] calculate pixel gray values of each pixel point in the second detection images as density values of corresponding positions;

[0035] perform sliding window traversal on the second detection images, and calculate average densities in each window according to the density values:

[0036] ;

[0037] ;

[0038] In the formula, This represents the average density within each window. This indicates the coordinates of the top-left corner of the window. These are the width and height of the window, respectively. For the image in Density value at that location, For the image in The pixel grayscale value at that location;

[0039] The average density of all windows is compared with a preset threshold. When the average density is greater than the preset threshold, the area where the corresponding window is located is divided into a coarse-grained area. When the average density is less than or equal to the preset threshold, the area where the corresponding window is located is divided into a fine-grained area.

[0040] Thirdly, the present invention also provides an electronic device including a processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein when the processor executes the computer instructions, the electronic device performs the method as described in the first aspect above and any possible implementation thereof.

[0041] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor of an electronic device, cause the processor to perform a method as described in the first aspect above and any possible implementation thereof.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] This invention discloses a remote production monitoring method for biomass fuel, comprising: establishing a standard image database containing standard particles of different sizes; extracting a first seed feature for coarse particle regions and a second seed feature for fine particle regions from the standard images; acquiring images of fuel particles after they have been crushed by a pulverizer on a biomass fuel production line; performing grayscale processing on the fuel particle images to obtain a first detection image; using the first and second seed features to perform region growth segmentation on the first detection image to obtain coarse particle regions and fine particle regions in the first detection image, and removing these regions from the first detection image to obtain a second detection image to be classified; calculating the particle density distribution characteristics of the second detection image, and dividing the coarse particle regions and fine particle regions of the second detection image according to the particle density distribution characteristics; and integrating the coarse particle regions in the first and second detection images to obtain the target coarse particle fuel to be crushed.

[0044] The present application greatly improves the efficiency of particle size monitoring by automatic image acquisition and processing technology, and reduces the necessity of manual participation. The image processing technology and region growing segmentation technology are used for particle size classification, which reduces the error caused by human factors, realizes real-time monitoring and intelligent classification of particle size, and improves the accuracy and consistency of the classification results. By combining the gray value, particle density distribution and other characteristics for particle size classification, the accuracy and robustness of the classification are improved, the adaptability to complex production environment is improved, and the particle classification requirements of different materials and sizes can be effectively met, thereby greatly improving the finished product quality and production efficiency of the biomass fuel production line.

[0045] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, but not limiting the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the background art, the drawings needed to be used in the embodiments of the present application or the background art will be described below.

[0047] The drawings herein are incorporated into the specification and form part of the specification, which show embodiments consistent with the present disclosure, and together with the specification, serve to illustrate the technical solutions of the present disclosure.

[0048] Figure 1 A flowchart of a biomass fuel remote production monitoring method provided by the embodiment of the present application;

[0049] Figure 2 A flowchart of a biomass fuel remote production monitoring method provided by the embodiment of the present application; Figure 1 A flowchart of a biomass fuel remote production monitoring method provided by the embodiment of the present application;

[0050] Figure 3 A flowchart of a biomass fuel remote production monitoring method provided by the embodiment of the present application; Figure 1 A flowchart of a biomass fuel remote production monitoring method provided by the embodiment of the present application;

[0051] Figure 4 A structural diagram of a biomass fuel remote production monitoring system provided by the embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor are within the scope of protection of the present application.

[0053] Reference to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. As will be apparent to those of ordinary skill in the art, embodiments described herein can be combinable with other embodiments.

[0054] At present, in the production and processing process of biomass fuel, the traditional manual monitoring method is usually used to screen coarse and fine particles, so that the coarse particles are subjected to secondary crushing, thereby avoiding affecting the quality of the final product. At present, it usually depends on manual detection or machine screening. The former is prone to inconsistency of detection results, time-consuming and difficult to meet the real-time monitoring needs of large-scale production. The latter usually relies on a single feature, such as particle shape and size, when screening, which is also difficult to cope with complex and variable production environment. In order to solve the above problems, the present application aims to provide a remote production monitoring method for biomass fuel, which can quickly and accurately screen coarse and fine particle size regions through intelligent image recognition, thereby improving the energy efficiency of monitoring particle size and further improving production efficiency and product quality.

[0055] Please refer to Figure 1 , Figure 1 The flowchart of a remote production monitoring method for biomass fuel provided by the embodiment of the present application is shown in the figure. Figure 1 As shown in the figure, the method comprises:

[0056] S10, a standard image database containing standard particles of different particle sizes is established, and first seed features of coarse particle regions and second seed features of fine particle regions are extracted from the standard images.

[0057] In step S10, when establishing the standard image database, high-definition images of standard particles of different particle sizes need to be collected to ensure that the image clarity is high enough to clearly distinguish the details of the particles. At the same time, the particle size information in each image is labeled to establish the correspondence between the image and the particle size. Then, the first seed features of the coarse particle regions and the second seed features of the fine particle regions are extracted from the standard images. The first seed features and the second seed features both include pixel gray scale features, texture features and shape features. Through multi-modal feature extraction, more comprehensive particle information can be obtained to improve the accuracy of subsequent classification. Through image processing and automatic feature extraction, the complexity of manual detection and feature extraction can be reduced, and the recognition efficiency is improved.

[0058] S20, the fuel particle images after being finely crushed by the crusher on the biomass fuel production line are collected, and the fuel particle images are subjected to gray scale processing to obtain first detection images.

[0059] In step S20, a high-definition camera is usually installed on the biomass fuel production line to periodically or continuously collect images of fuel particles. When collecting, ensure that the lighting conditions of the image collection environment are stable to avoid fluctuations in image quality due to changes in light. Further, after extracting the fuel particle images, the collected color images need to be grayscale processed to simplify the subsequent processing steps and reduce the computational complexity.

[0060] S30, using the first seed feature and the second seed feature to perform region growing segmentation on the first detection image respectively, to obtain the coarse particle region and the fine particle region in the first detection image, and to remove them from the first detection image to obtain the second detection image to be classified.

[0061] In this step, according to the first seed feature and the second seed feature extracted in step S10, region growing segmentation is performed on the first detection image to obtain coarse particle regions and fine particle regions respectively.

[0062] In one specific embodiment, the region growing process is as follows:

[0063] 1) Select seed points: Select multiple seed points, which should be located in the coarse particle region and fine particle region marked in the standard image. Use the first seed feature and the second seed feature in the standard image to initialize the seed points. For example: select the points with a gray value close to 120 in the image as the seed points of the coarse particle region; select the points with a gray value close to 80 in the image as the seed points of the fine particle region.

[0064] 2) Set the growth threshold: Set the growth threshold, i.e. the gray value difference between the seed points and the similar pixel points. For example, the growth threshold of the coarse particle region can be set to ±15 gray units; the growth threshold of the fine particle region can be set to ±10 gray units. Wherein, the threshold can be adjusted according to the actual gray distribution of the particles.

[0065] 3) Growth process: Starting from the seed points of the coarse particle region, gradually expand the pixel points with a gray value in the range of [105, 135] into the coarse particle region. Starting from the seed points of the fine particle region, gradually expand the pixel points with a gray value in the range of [70, 90] into the fine particle region.

[0066] 4) Stopping condition: When there are no more pixel points around that meet the growth condition, stop growing. For example, the maximum number of growth can be set to 100 times, or the maximum growth area can be set to 20% of the total image area.

[0067] 5) Result verification: Verify the segmentation results by manual sampling or comparison with the standard image to ensure that there is no omission or error.

[0068] Therefore, by the above specific implementation steps, the region growing segmentation can be effectively performed on the first detection image to obtain the coarse particle region and the fine particle region, and the accuracy and integrity of the segmentation result are ensured by setting appropriate growing threshold and stopping condition.

[0069] Generally, the coarse particle region and the fine particle region do not occupy the entire region of the first detection image. After obtaining the coarse particle region and the fine particle region in the first detection image, the segmented coarse particle region and fine particle region are removed from the first detection image to obtain a second detection image to be classified. Therefore, the present embodiment can improve the recognition rate of particles of different sizes by using the multi-scale segmentation technology, and ensure the comprehensiveness of the segmentation result.

[0070] S40, calculating the particle density distribution feature of the second detection image, and dividing the coarse particle region and the fine particle region of the second detection image according to the particle density distribution feature.

[0071] In this step, the particle density distribution feature generally refers to the density value distribution of different pixel point positions, and the number or density of particles in the neighborhood of each pixel point in the second detection image can be calculated. The density distribution can be estimated by counting the number of particles around each pixel point, thereby improving the calculation efficiency of the density distribution feature. After obtaining the particle density distribution feature, the coarse particle region and the fine particle region in the second detection image are divided by setting appropriate threshold.

[0072] Preferably, in some embodiments, when the density distribution feature is used to divide the coarse particle region and the fine particle region, local features can be focused on, such as local region division by sliding window, and a high-dimensional feature vector is formed by splicing the extracted local features together, which can further improve the robustness of classification.

[0073] S50, integrating the coarse particle regions in the first detection image and the second detection image to obtain a target coarse particle fuel to be finely crushed.

[0074] Finally, all the coarse particle regions identified in the first detection image and the second detection image are integrated to obtain a complete coarse particle region set, i.e. the final target coarse particle fuel can be determined, and these target coarse particle fuels need to be further finely crushed and then put into the fuel pressing process to improve the quality of the finished product.

[0075] Referring to Figure 2 In one embodiment, the first seed feature of the coarse particle region and the second seed feature of the fine particle region extracted from the standard image in step S10 include the following sub-steps:

[0076] S101, extract image features of coarse particle regions and image features of fine particle regions from the standard image, the image features including pixel gray level features, texture features and shape features.

[0077] For the convenience of understanding, first of all, the image features in this step are described:

[0078] Pixel gray level features: calculate the average gray level value and the gray level histogram of each particle region. Calculate the gray level co-occurrence matrix (GLCM) of the particle region, and extract statistical features such as energy, contrast, and homogeneity from it.

[0079] Texture features: use local binary pattern (LBP) to extract texture features of the particle region. Calculate the histogram of LBP, which reflects the distribution of local texture. Further features in the gray level co-occurrence matrix (GLCM) can be calculated, such as energy, contrast, and homogeneity.

[0080] Shape features: calculate the aspect ratio and roundness of the particle region. Use edge detection algorithms (such as Sobel operator and Canny operator) to extract edge information of the particle region. Calculate the edge density, which is the proportion of edge pixels.

[0081] Compared with the existing single feature extraction classification, the present embodiment can more comprehensively describe the characteristics of the particles by extracting gray level features, texture features and shape features, and improve the accuracy and robustness of the classification. The combination of multiple features can provide more abundant information, which is helpful to distinguish different types of particles.

[0082] S102, use SAE algorithm to reduce the dimension of the image features of coarse particle regions and the image features of fine particle regions respectively, and use gradient boosting tree algorithm to fit the reduced features.

[0083] In this step, the extracted features are reduced by using stacked auto-encoder (SAE), which removes redundant information and can retain the most important features. In this way, the computational complexity can be reduced and the running speed of the model can be improved. Specifically, multiple auto-encoders can be trained, each of which learns a part of the input features, and finally all the auto-encoders are connected in series to form a deep neural network.

[0084] Further, the gradient boosting tree algorithm is used to fit the reduced features, which specifically includes:

[0085] Suppose the total number of features is m and the number of categories is N, the Gini impurity of the node under the decision tree is calculated:

[0086] ;

[0087] In the formula, is the first Gini impurity of a node v in a decision tree; For a node Under the category The proportion of the occupied;

[0088] Calculate the feature In the node The change amount of Gini impurity before and after branching :

[0089] ;

[0090] In the formula, And Gini impurity of the two new nodes after branching respectively;

[0091] Define the feature In the decision tree The set of nodes appearing under the decision tree is Q, then The importance of the feature in the first Decision tree is:

[0092] ;

[0093] In the formula, Is a node in the node set Q;

[0094] Assume that a total of Trees are generated in the training process, then the importance of the feature In all trees Is:

[0095] ;

[0096] In the formula, Is the feature importance corresponding to each tree;

[0097] Normalization operation is performed to obtain the final importance score of the feature :

[0098] ;

[0099] In the formula, Is the importance under the feature dimension .

[0100] Therefore, by fitting the reduced features through the gradient boosting tree algorithm, the Gini impurity of different nodes under the decision tree is calculated, and the change amount of Gini impurity is calculated to determine the feature importance, and then according to the feature importance score, the most relevant feature quantity is selected from the multi-dimensional features, so that the prediction result is more accurate.

[0101] ​S103, screen the most relevant feature quantity after fitting, extract the target image features of the coarse particle region and the target image features of the fine particle region.

[0102] This step aims to analyze the correlation of the fitted features and screen out the feature quantity with the largest correlation. Preferably, the feature importance score can be used to select the most representative features. Then the target image features of the coarse particle region and the target image features of the fine particle region are obtained respectively. By screening the feature quantity with the largest correlation, the feature set can be further optimized, and the accuracy and robustness of the model can be improved.

[0103] S104, feature fusion is performed on the target image features of the coarse particle region to obtain the first seed feature; feature fusion is performed on the target image features of the fine particle region to obtain the second seed feature.

[0104] Finally, the target image features of the coarse particle region are fused to form a comprehensive feature vector as the first seed feature. The target image features of the fine particle region are fused to form a comprehensive feature vector as the second seed feature. In the fusion, feature splicing method can be used to splice different types of features together. Feature weighting method can also be used to assign different weights to different features according to experimental results. In this way, through feature fusion, a comprehensive feature vector containing multiple information can be formed, improving the classification efficiency and the accuracy and robustness of the classification results.

[0105] Therefore, the embodiment realizes an efficient and accurate biomass fuel particle feature extraction and classification method. It includes extracting multiple image features from standard images, using SAE algorithm for feature dimension reduction, using random forest algorithm for feature fitting, screening the feature quantity with the largest correlation, and performing feature fusion. Not only improves the accuracy and robustness of feature extraction, but also improves the efficiency and generalization ability of the model through feature dimension reduction and fusion, so as to realize the effective classification and monitoring of biomass fuel particle size.

[0106] Referring to Figure 3 In one embodiment, in the step S40 described above, the particle density distribution features of the second detection image are calculated, and the coarse particle region and the fine particle region of the second detection image are divided according to the particle density distribution features, which specifically includes the following sub-steps:

[0107] S401, calculate the pixel gray value of each pixel point in the second detection image as the density value of the corresponding position.

[0108] S402, traverse the second detection image using a sliding window, and calculate the average density in each window according to the density value:

[0109] ;

[0110] ;

[0111] wherein, denotes the average density within each window, denotes the top-left corner coordinate of the window, are the width and height of the window, respectively, is the density value of the image at is the pixel gray value of the image at is the pixel gray value of the image at .

[0112] S403, compare the average density of all windows with a preset threshold value, when the average density is greater than the preset threshold value, the area where the corresponding window is located is divided into coarse particle area, when the average density is less than or equal to the preset threshold value, the area where the corresponding window is located is divided into fine particle area.

[0113] Before step S401 is performed, in order to ensure the accuracy of the calculation result, it is usually necessary to appropriately pretreat the image, such as denoising and image enhancement, etc. In image processing, the gray value usually reflects the brightness or intensity of the pixel point. For fuel particle images, the more or more dense the particles are, the darker they usually appear (i.e. the higher the gray value), because the particles block the passage of light. Therefore, to some extent, the gray value can approximately represent the particle density at that position. A higher gray value means that more particles are accumulated at the same position, and vice versa. When the fuel particles are relatively uniformly distributed in the image, it can be assumed that the actual physical space represented by each pixel point in the image is the same. In this case, the gray value can directly reflect the number or density of particles.

[0114] Therefore, directly using the gray value instead of the density value in step S401 can simplify the calculation process, and since the complex density calculation step is omitted, the entire processing flow becomes more efficient, reducing the demand for computing resources.

[0115] Further, in step S402, the second detection image is traversed by using a sliding window, and the average density in each window is calculated according to the density value. By means of the sliding window, the density change in the image can be analyzed in detail. In this way, the local features of the particle distribution in the image can be captured, which is helpful for subsequent classification. Finally, when the average densities of all the windows are obtained, only a comparison with a preset threshold value is needed. When the average density is greater than the preset threshold value, the region corresponding to the window is divided into a coarse particle region; when the average density is less than or equal to the preset threshold value, the region corresponding to the window is divided into a fine particle region. Through the threshold comparison, the coarse particle region and the fine particle region can be accurately distinguished. In this way, the classification of the coarse and fine particle regions of the fuel particle image can be effectively performed, so as to improve the classification accuracy and efficiency. Therefore, through the division of the coarse and fine particle regions in multiple scales and the final integration and connection, compared with the single-scale division mode, the recognition accuracy of the coarse and fine particle regions can be greatly improved, and finally the finished product quality of the biomass fuel production line is improved.

[0116] In summary, the remote production monitoring method of the biomass fuel provided by the present application greatly improves the efficiency of particle size monitoring and reduces the necessity of manual participation by using automatic image acquisition and processing technology. The particle size classification is performed by using image processing technology and region growing segmentation technology, which reduces errors caused by human factors, realizes real-time monitoring and intelligent classification of particle size, and improves the accuracy and consistency of the classification results. The particle size classification is performed by combining multiple features such as gray value and particle density distribution, which improves the accuracy and robustness of the classification, improves the adaptability to complex production environments, and can effectively meet the particle classification requirements of different materials and sizes, thereby greatly improving the finished product quality and production efficiency of the biomass fuel production line.

[0117] Reference Figure 4 In one embodiment, the present application also provides a remote production monitoring system of biomass fuel, which comprises:

[0118] The feature extraction unit 100 is configured to establish a standard image database containing standard particles of different particle sizes, and extract first seed features of coarse particle regions and second seed features of fine particle regions from the standard images.

[0119] The gray processing unit 200 is configured to acquire fuel particle images after being finely crushed by a crusher on a biomass fuel production line, and perform gray processing on the fuel particle images to obtain a first detection image.

[0120] The image segmentation unit 300 is configured to perform region growing segmentation on the first detection image by using the first seed features and the second seed features respectively, to obtain coarse particle regions and fine particle regions in the first detection image, and to remove the coarse particle regions and the fine particle regions from the first detection image to obtain a second detection image to be classified.

[0121] The region division unit 400 is configured to calculate a particle density distribution feature of the second detection image, and divide a coarse particle region and a fine particle region of the second detection image according to the particle density distribution feature.

[0122] The coarse particle integration unit 500 is configured to integrate the coarse particle regions in the first detection image and the second detection image to obtain a target coarse particle fuel to be crushed.

[0123] In an embodiment, the feature extraction unit 100 is further configured to:

[0124] extract image features of the coarse particle region and image features of the fine particle region from the standard image, wherein the image features include pixel gray scale features, texture features and shape features;

[0125] perform feature dimension reduction on the image features of the coarse particle region and the image features of the fine particle region respectively by using a SAE algorithm, and perform fitting on the reduced features by using a gradient boosting tree algorithm;

[0126] screen the feature quantity with the largest correlation after fitting, and extract target image features of the coarse particle region and target image features of the fine particle region;

[0127] perform feature fusion on the target image features of the coarse particle region to obtain a first seed feature, and perform feature fusion on the target image features of the fine particle region to obtain a second seed feature.

[0128] In an embodiment, the region division unit 400 is further configured to:

[0129] calculate a pixel gray scale value of each pixel point in the second detection image as a density value at a corresponding position;

[0130] perform sliding window traversal on the second detection image, and calculate an average density in each window according to the density value:

[0131] ;

[0132] ;

[0133] wherein, denotes the average density in each window, denotes a top-left corner coordinate of the window, denotes a width and a height of the window respectively, denotes a density value of the image at the position, denotes a pixel gray scale value of the image at the position;

[0134] ​​The average density of all windows is compared with a preset threshold value, when the average density is greater than the preset threshold value, the area where the corresponding window is located is divided into a coarse particle region, when the average density is less than or equal to the preset threshold value, the area where the corresponding window is located is divided into a fine particle region.

[0135] It can be understood that the remote production monitoring system of the biomass fuel provided by the embodiment has functions or contains modules that can be used to execute the method described in the above method embodiment, and specific implementation can be referred to the description of the above method embodiment. For the sake of brevity, it will not be repeated here.

[0136] The application further provides an electronic device, including a processor and a memory, the memory is used for storing computer program code, the computer program code includes computer instructions, when the processor executes the computer instructions, the electronic device executes the method of any one of the above possible implementation manners.

[0137] The application further provides a computer readable storage medium, the computer readable storage medium stores a computer program, the computer program includes program instructions, the program instructions are executed by the processor of the electronic device, and the processor executes the method of any one of the above possible implementation manners.

[0138] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here. Those skilled in the art can also clearly understand that each embodiment of the application describes each with emphasis, for the convenience and brevity of description, the same or similar parts may not be described in different embodiments, therefore, the parts not described or not described in detail in an embodiment can refer to the description of other embodiments.

Claims

1. A method of remote production monitoring of a biomass fuel, characterized by, The method comprises: establishing a standard image database containing standard particles of different particle sizes, extracting first seed features of coarse particle regions and second seed features of fine particle regions from the standard images; collecting fuel particle images after being finely crushed by a crusher on a biomass fuel production line, performing grayscale processing on the fuel particle images to obtain first detection images; performing region growing segmentation on the first detection images using the first seed features and the second seed features respectively to obtain coarse particle regions and fine particle regions in the first detection images, and removing the coarse particle regions and the fine particle regions from the first detection images to obtain second detection images to be classified; calculating particle density distribution features of the second detection images, and dividing coarse particle regions and fine particle regions of the second detection images according to the particle density distribution features, comprising: calculating pixel grayscale values of each pixel point in the second detection images as density values of corresponding positions; traversing the second detection images using a sliding window, and calculating average densities in each window according to the density values: ； ； wherein, denotes the average density within each window, denotes the upper left corner coordinate of the window, are the width and height of the window, respectively, is the density value of the image at is the pixel gray value of the image at is the pixel gray value of the image at is the pixel gray value of the image at comparing the average densities of all windows with a preset threshold value, dividing regions where the corresponding windows are located into coarse particle regions when the average densities are greater than the preset threshold value, and dividing regions where the corresponding windows are located into fine particle regions when the average densities are less than or equal to the preset threshold value; integrating the coarse particle regions in the first detection images and the second detection images to obtain target coarse particle fuel to be finely crushed.

2. The method of remote production monitoring of a biomass fuel according to claim 1, characterized in that, The method comprises: extracting image features of coarse particle regions and image features of fine particle regions from the standard images, wherein the image features comprise pixel grayscale features, texture features and shape features; performing feature dimension reduction on the image features of the coarse particle regions and the image features of the fine particle regions respectively using an SAE algorithm, and fitting the reduced features using a gradient boosting tree algorithm; screening feature quantities with the largest correlation after fitting, and extracting target image features of the coarse particle regions and target image features of the fine particle regions; performing feature fusion on the target image features of the coarse particle regions to obtain the first seed features, and performing feature fusion on the target image features of the fine particle regions to obtain the second seed features.

3. A remote production monitoring system for biomass fuel, characterized by, The system comprises: a feature extraction unit configured to establish a standard image database containing standard particles of different particle sizes, and extract first seed features of coarse particle regions and second seed features of fine particle regions from the standard images; a grayscale processing unit configured to collect fuel particle images after being finely crushed by a crusher on a biomass fuel production line, and perform grayscale processing on the fuel particle images to obtain first detection images; an image segmentation unit configured to perform region growing segmentation on the first detection images using the first seed features and the second seed features respectively to obtain coarse particle regions and fine particle regions in the first detection images, and remove the coarse particle regions and the fine particle regions from the first detection images to obtain second detection images to be classified; a region division unit configured to calculate particle density distribution features of the second detection images, and divide coarse particle regions and fine particle regions of the second detection images according to the particle density distribution features, comprising: calculating pixel grayscale values of each pixel point in the second detection images as density values of corresponding positions; The second detection image is traversed by using a sliding window, and the average density in each window is calculated according to the density value: ； ； In the formula, This represents the average density within each window. This indicates the coordinates of the top-left corner of the window. These are the width and height of the window, respectively. For the image in Density value at that location, For the image in The pixel grayscale value at that location; The average densities of all the windows are compared with a preset threshold value, when the average density is greater than the preset threshold value, the region where the corresponding window is located is divided into a coarse particle region, when the average density is less than or equal to the preset threshold value, the region where the corresponding window is located is divided into a fine particle region; The coarse particle integration unit is configured to integrate the coarse particle regions in the first detection image and the second detection image to obtain the target coarse particle fuel to be finely crushed.

4. The remote production monitoring system of biomass fuel according to claim 3, wherein, The feature extraction unit is further configured to: extract image features of the coarse particle region and image features of the fine particle region from the standard image, the image features including pixel gray scale features, texture features and shape features; perform feature dimension reduction on the image features of the coarse particle region and the image features of the fine particle region respectively by using a SAE algorithm, and perform fitting on the reduced features by using a gradient boosting tree algorithm; screen the feature quantity with the largest correlation after fitting, extract target image features of the coarse particle region and target image features of the fine particle region; perform feature fusion on the target image features of the coarse particle region to obtain a first seed feature, and perform feature fusion on the target image features of the fine particle region to obtain a second seed feature.

5. An electronic device, comprising: comprise: a processor and a memory, the memory being configured to store computer program code, the computer program code comprising computer instructions, when the processor executes the computer instructions, the electronic device executes the biomass fuel remote production monitoring method according to any one of claims 1 to 2.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, the computer program comprises program instructions, when the program instructions are executed by the processor of the electronic device, the processor executes the biomass fuel remote production monitoring method according to any one of claims 1 to 2.

Images