Coal dirt band sorting method and system based on intelligent dry separation system

Abstract

The invention relates to a coal dirt band sorting method and system based on an intelligent dry separation system, and the method comprises the steps: carrying out the continuous image collection and image preprocessing of a material on the intelligent dry separation system, and obtaining a to-be-recognized image set; inputting the to-be-recognized image set into a three-classification model based on visual recognition, and outputting a recognition result; when the identification result is coal dirt band, acquiring an image sequence for the material, and calculating the estimated coal content of the material according to the image sequence; the estimated coal content is compared with a preset separation threshold value, and when the estimated coal content is larger than the separation threshold value, the intelligent dry separation system blows the materials into a coal bunker; and when the estimated coal content is smaller than the separation threshold value, the intelligent dry separation system blows the materials to a gangue bin. According to the method, the accuracy of identifying and sorting the coal dirt band is improved.

Description

Technical Field

[0001] This invention relates to the field of coal raw material identification and sorting technology; specifically, this invention relates to a coal and gangue sorting method and system based on an intelligent dry sorting system. Background Technology

[0002] In coal mining, besides pure lump coal and gangue, raw coal also contains a large amount of "coal interbedded with gangue," which are mixed lumps of coal and gangue that are closely coexisting and physically combined due to geological processes. Currently, machine vision-based sorting methods have become one of the mainstream technologies for sorting lump coal and gangue. These systems typically capture surface images of materials in motion using high-speed cameras and utilize trained visual algorithms (such as deep learning models) to identify materials as "coal" or "gangue" in real time based on their texture, color, grayscale, and other appearance characteristics. This information is then used to control actuators (such as high-pressure air valves) to sort the materials into the appropriate channels.

[0003] However, this sorting strategy, which relies on single, instantaneous surface image features, has certain drawbacks when dealing with "coal mixed with gangue." First, the results of this sorting strategy are highly dependent on the orientation of the material at the moment of capture, making its identification criteria somewhat biased and accidental. Different surfaces of coal mixed with gangue exhibit vastly different compositional characteristics. When the material moves on the conveyor belt, if the camera captures a surface with a high coal content, the system will misclassify it as coal; conversely, if it captures a surface with a high gangue content, it will be misclassified as gangue. Second, this sorting strategy has unavoidable negative impacts. Misclassifying coal mixed with gangue as coal leads to a large amount of gangue being mixed into the refined coal product, significantly reducing the calorific value of the final coal product and affecting product quality and selling price. Misclassifying coal mixed with gangue as gangue results in recycled coal being discarded as waste, wasting valuable coal resources, reducing yield recovery rates, and directly harming economic benefits.

[0004] Existing visual sorting methods primarily focus on improving the accuracy and speed of identifying "pure coal" and "pure gangue" through core algorithm optimization. Essentially, they remain binary classification models based on single observations, resulting in either coal or gangue. This approach lacks an effective definition, identification, and processing mechanism for the special mixture of "coal and gangue," and it cannot comprehensively analyze and quantitatively evaluate the composition of its multiple surfaces. This leads to the dual problems of decreased product quality after sorting and waste of coal resources. Summary of the Invention

[0005] In view of this, the present invention provides a coal-gangue separation method and system based on an intelligent dry separation system, thereby solving or at least alleviating one or more of the above-mentioned problems and other problems existing in the prior art.

[0006] To achieve the aforementioned objective, a first aspect of the present invention provides a coal-gangue separation method based on an intelligent dry separation system, characterized in that the method comprises:

[0007] Step 1: Continuously acquire images of the material on the intelligent dry separation system to obtain an original image set; perform image preprocessing on the original image set to obtain an image set to be identified.

[0008] Step 2: Input the set of images to be identified into a three-class classification model based on visual recognition, and output the recognition results, which include "coal", "gangue" or "coal mixed with gangue";

[0009] Step 3: When the identification result is "coal", the intelligent dry separation system sprays the material into the coal bunker; when the identification result is "gangue", the intelligent dry separation system sprays the material into the gangue bunker; when the identification result is "coal mixed with gangue", proceed to the next step.

[0010] Step 4: Collect image sequences of the material and calculate the estimated coal content of the material based on the image sequences;

[0011] Step 5: Compare the estimated coal content with the preset sorting threshold. When the estimated coal content is greater than the sorting threshold, the intelligent dry separation system sprays the material into the coal bunker; when the estimated coal content is less than the sorting threshold, the intelligent dry separation system sprays the material into the gangue bunker.

[0012] In the coal-gangue separation method described above, step 4 may optionally include:

[0013] Step 4-1: Track the movement trajectory of the material using a target tracking algorithm, and acquire images according to spatial or time intervals during the movement of the material to obtain the image sequence;

[0014] Step 4-2: Calculate the area ratio of coal region in each image of the image sequence using a surface composition analysis algorithm;

[0015] Step 4-3: Calculate the estimated coal content of the material by combining the coal area ratio of each image using a data fusion algorithm.

[0016] In the coal-gangue sorting method described above, optionally, the three-classification model is pre-trained, and the pre-trained dataset consists of labeled coal-gangue images with different morphological and surface features.

[0017] In the coal-gangue separation method described above, optionally, in step 3, when the identification result is "coal-gangue", the intelligent dry separation system creates a unique identification code for the material.

[0018] In the coal-gangue separation method described above, the target tracking algorithm may optionally be a Kalman filter algorithm.

[0019] In the coal-gangue separation method described above, optionally, when tracking the movement trajectory of the material using the Kalman filter algorithm, the material is defined as... State vector at time step ,

[0020]

[0021] in, Let x be the x-coordinate of the centroid position of the material. Let be the ordinate of the centroid position of the material. The width of the identification border of the material. The height of the identification border of the material. for The rate of change, for The rate of change, for The rate of change, for The rate of change,

[0022] In the prediction phase of the Kalman filter algorithm, the state prediction equation is:

[0023]

[0024] in, For the material in The prior state estimate at time t. Here is the state transition matrix. For the material in The prior state estimate at time t. for Time and The time interval between moments

[0025] The covariance prediction equation is

[0026]

[0027] in, for The prior estimate of covariance at time t, for The prior estimate of covariance at time t, State transition matrix The transpose of the matrix, The process noise covariance matrix is...

[0028] During the update phase of the Kalman filter algorithm, the observation vector for

[0029]

[0030] in, for The observed values, for The observed values, for The observed values, for The observed values,

[0031] Observation matrix for

[0032] The Kalman gain equation is

[0033] in, Here is the filter gain matrix. Observation matrix The transpose of the matrix, To observe the noise covariance,

[0034] The state update equation is

[0035] in, For the material in The posterior state estimate at time t.

[0036] The covariance update equation is

[0037] in, for The posterior estimate of the covariance at time t. It is an identity matrix.

[0038] In the coal and gangue separation method described above, optionally, in step 4-1, when the material's movement trajectory passes through image acquisition devices from different perspectives, or when the material's posture changes due to the movement of the intelligent dry separation system, images of the material are acquired to obtain the image sequence.

[0039] In the coal-gangue separation method described above, step 4-2 may optionally include:

[0040] Step 4-2-1: Define the image sequence as ,in To determine the number of valid images acquired, a semantic segmentation model is used to segment the image sequence. Each image in Perform pixel-level classification and output the classification results, which include "coal", "gangue" or "background";

[0041] Step 4-2-2: Calculate the value of each image based on the classification results. coal-producing areas account for a certain percentage of the total area. ,

[0042]

[0043] in, For image The classification result mentioned above is the number of pixels for "coal". For image The number of pixels classified as "gangue" in the text;

[0044] Step 4-2-3: Combine each image coal-producing areas account for a certain percentage of the total area. Obtain the data set of coal area proportion of the material. , .

[0045] In the coal-gangue separation method described above, optionally, in step 4-3, the estimated coal content of the material is calculated using a maximum value fusion algorithm or an average value fusion algorithm. ,

[0046] When using the maximum value fusion algorithm ;

[0047] When using the average fusion algorithm .

[0048] To achieve the aforementioned objectives, a second aspect of the present invention provides a coal and gangue separation system based on an intelligent dry separation system, which can implement the coal and gangue separation method described above. The system is characterized in that it includes a material distribution module, an identification module, a control module, and an execution module, wherein...

[0049] The fabric module is used to convey the materials, including coal, gangue, and coal mixed with gangue.

[0050] The identification module is used to perform image acquisition, preprocessing, and three-classification model identification on the material on the fabric module. When the identification result of the three-classification model is coal mixed with gangue, the identification module is used to acquire image sequences of the material, calculate the estimated coal content, and obtain the final identification result of the material according to a preset sorting threshold.

[0051] The control module is used to send an execution signal to the execution module based on the final identification result.

[0052] The execution module is used to spray the material according to the execution signal, so that the material falls into the corresponding compartment.

[0053] This invention presents a coal-gangue separation method based on an intelligent dry separation system. It classifies coal raw materials into three categories using a visual recognition model and estimates the coal content of materials identified as coal-gangue. This allows for the setting of separation thresholds based on actual requirements for coal calorific value, flexibly separating this type of mixed raw material. This method reliably distinguishes coal-gangue from pure coal and pure gangue, improving the accuracy of coal-gangue identification and separation. Furthermore, this method maximizes the recovery of coal-gangue with recycling value, improving resource utilization while effectively preventing the mixing of low-calorific-value materials, ensuring the stability of the quality and calorific value of the clean coal product. Simultaneously, this method empowers users to define their own separation standards, enabling them to flexibly set separation thresholds according to economic benefit objectives. This represents a leap from general separation to customized precision separation, enhancing the effectiveness and economic efficiency of the intelligent dry separation system.

[0054] In optional embodiments, this method utilizes target tracking algorithms, surface composition analysis algorithms, and data fusion algorithms to calculate the area ratio of coal regions in multiple surface images of coal interbedded coal from multiple perspectives, thereby deriving the estimated coal content of the interbedded coal. This estimation method, through comprehensive analysis and quantitative evaluation, reduces the interference of the non-uniform surface composition of the interbedded coal with the calculation results, reduces the randomness in calculating the coal content in the interbedded coal, and further improves the accuracy of sorting interbedded coal.

[0055] In a further optional embodiment, this method uses a Kalman filter algorithm and a semantic segmentation model to improve the stability and accuracy of coal interbedding and sorting.

[0056] The present invention further provides a coal-gangue separation system based on an intelligent dry separation system, which can realize the coal-gangue separation method described above, and therefore the coal-gangue separation system also has the above advantages. Attached Figure Description

[0057] The disclosure of this invention will become more apparent from the accompanying drawings. It should be understood that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings:

[0058] Figure 1 This is a schematic flowchart of an embodiment of the coal and gangue separation method according to the present invention.

[0059] Figure 2 This is a schematic diagram showing the surface composition characteristics of coal interbedded with gangue from different perspectives.

[0060] Figure 3 This is a schematic block diagram of an embodiment of the coal and gangue separation method according to the present invention. Detailed Implementation

[0061] Referring to the accompanying drawings and specific embodiments, the structure, composition, features, and advantages of the movable belt conveyor truss relative pose monitoring system and method of the present invention will be described below by way of example. However, all descriptions should not be construed as limiting the present invention in any way.

[0062] Furthermore, for any single technical feature described or implied in the embodiments mentioned herein, or any single technical feature shown or implied in the various figures, the present invention still allows for any combination or deletion of these technical features (or their equivalents) without any technical obstacle, and thus these further embodiments according to the present invention should also be considered within the scope of this description.

[0063] Figure 1 This is a schematic flowchart of an embodiment of the coal and gangue separation method according to the present invention.

[0064] from Figure 1 As can be seen, this embodiment uses the AIS intelligent dry separation system. The AIS intelligent dry separation system is an intelligent separation system based on visual recognition technology that pneumatically separates lumpy materials such as coal and ore. The separation method and system of this invention do not change the main hardware of the AIS intelligent dry separation system, but improve and optimize the algorithm model and introduce an adjustable separation threshold, thereby enhancing the adaptability and effectiveness of the existing AIS dry separation system to different separation requirements.

[0065] As shown in the figure, according to the coal and gangue separation method of the present invention, after the intelligent dry separation system is started and the belt conveyor begins to transport the material flow, the material flow is first identified. An industrial camera is used to continuously capture high-speed images of the material flow on the belt, obtaining the original image set of each material. The material flow on the belt includes not only pure coal and pure gangue, but also a large amount of coal gangue, that is, a mixed mass of coal and gangue that are closely coexisting and physically combined. As shown in the figure, coal, gangue, and coal gangue exhibit different surface characteristics.

[0066] In an optional embodiment, the original image set of these materials is preprocessed, including noise reduction and contrast enhancement, to obtain an image set that is easy for the model to recognize. Preprocessing methods can include grayscale conversion, histogram equalization, Gaussian filtering, wavelet denoising, etc., as long as they meet the standardized input requirements of the model used.

[0067] The pre-processed image set is input into the visual recognition model. Based on the material's texture, color, grayscale, and other appearance features in the image set, the material is identified as "coal," "gangue," or "coal mixed with gangue." Compared to existing binary classification models that only target coal and gangue, this method uses a three-class classification model that adds a third class for identifying coal mixed with gangue. This three-class classification model has been pre-trained using a training dataset containing a large number of labeled coal mixed with gangue images of different shapes and surface features. This allows the model to learn the unique mixed texture features of coal mixed with gangue, enabling reliable differentiation between coal mixed with gangue and pure coal or pure gangue.

[0068] After material flow identification, this method proceeds to the tracking stage. For materials identified as "coal mixed with gangue," an industrial camera is used to acquire images at different time points T1, T2, and T3, obtaining an image sequence. The image sequence can be in the form of continuous video frames. In other optional embodiments, different image acquisition devices can be used, as long as they meet the requirements of the coal raw material sorting face in terms of dust and explosion prevention.

[0069] Combination Figure 2 As can be seen, at these different time points, the material tumbles due to the movement of the conveyor belt, resulting in surface images being captured by the industrial camera from different angles. While the gangue exhibits similar surface characteristics from different angles, the coal-gangue mixture, as a mixed mass, often displays drastically different surface characteristics at different angles; for example, the proportion of coal area shown on each surface varies significantly.

[0070] Figure 1 and Figure 2 T1, T2, and T3 in the diagram are labels for different time points, without limiting the number of image acquisitions or the method of acquisition. In practical applications, different numbers of acquisitions can be performed according to time or spatial intervals, depending on the sorting requirements.

[0071] In an optional embodiment, when a material is identified as "coal mixed with gangue," the intelligent dry separation system immediately creates a unique identification code for it and initiates a target tracking algorithm for each different identification code to track the material's movement trajectory. This enables simultaneous tracking of multiple materials, ensuring the efficient operation of the intelligent dry separation system and improving the separation efficiency and accuracy of coal mixed with gangue.

[0072] Target tracking algorithms, such as Kalman filtering and particle filtering, can stably track materials based on their position, size, and motion characteristics within an image sequence. Using target tracking algorithms ensures accurate tracking of the material's trajectory, guaranteeing that even when briefly obscured by other materials or undergoing rotation, the intelligent dry sorting system can continuously lock onto the material and associate it with the same identification code. This is crucial for ensuring that industrial cameras can perform multiple, effective samplings of the same object.

[0073] Based on target tracking, the intelligent dry separation system will trigger the industrial cameras to acquire images multiple times during the material's transport process, taking advantage of the moment when the material passes by industrial cameras from different perspectives or changes in posture due to belt movement within the same perspective. This will allow the system to obtain image sequences of multiple different surfaces of the same coal gangue material.

[0074] like Figure 1 As shown, after completing target tracking and image sequence acquisition, the sorting method of the present invention performs surface analysis on the acquired image sequence. In an optional embodiment, a surface composition analysis algorithm is used to calculate the surface coal area ratio of each valid image in the image sequence, and then based on a data fusion algorithm, the estimated coal content of the material is obtained.

[0075] Finally, this method performs sorting decisions. In an optional embodiment, a sorting threshold is preset according to the user's needs. For example, the user sets the sorting threshold to 40%, which requires that the coal content of any material in the final coal product must not be less than 40%. The intelligent dry separation system compares the calculated estimated coal content with the sorting threshold. If the estimated coal content of the material is ≥40%, when the material moves to the position of the actuator, such as the injection valve, it is injected onto the lump coal conveyor belt and eventually falls into the lump coal bin; if the estimated coal content of the material is <40%, when the material moves to the position of the actuator, such as the injection valve, it is injected onto the gangue conveyor belt and eventually falls into the gangue bin.

[0076] Simultaneously, during the identification phase, materials identified as "coal" by the three-classification model are injected and fall into the lump coal bin, while materials identified as "gangue" are injected and fall into the gangue bin, thus achieving effective separation of three different types of materials: coal, gangue, and coal mixed with gangue. The actuator and method can be configured differently based on the structure and function of the intelligent dry separation system, and the separation threshold value can also be dynamically adjusted according to the actual calorific value requirements of production.

[0077] Figure 3 This is a schematic block diagram of an embodiment of the coal and gangue separation method according to the present invention.

[0078] In such Figure 3In the illustrated embodiment, raw coal stockpile material is fed into the AIS intelligent dry separation system via a belt conveyor. A preliminary screening is performed using a three-class classification model based on visual recognition, identifying the raw coal stockpile material as "coal," "gangue," or "coal mixed with gangue." For materials identified as "coal mixed with gangue," a Kalman filter algorithm is used for target tracking, and a semantic segmentation model is used for surface composition analysis. Finally, the intelligent dry separation system uses an actuator to spray these materials, causing them to fall into their respective compartments.

[0079] In an optional embodiment, when using the Kalman filter algorithm, the state vector is first defined. That is, the material in The state at any given moment: .in, The x-coordinate represents the centroid position of the material. The ordinate represents the centroid of the material. The width of the identification border for the material. The height of the material identification border. for The rate of change, for The rate of change, for The rate of change, for The rate of change.

[0080] In the prediction phase of the Kalman filter algorithm, the state prediction equation is:

[0081]

[0082] in, For materials in The prior state estimate at time t. Here is the state transition matrix. For materials in The prior state estimate at time t. for Time and The time interval between moments.

[0083] The covariance prediction equation is ,in, for The prior estimate of covariance at time t, for The prior estimate of covariance at time t, State transition matrix The transpose of the matrix, Let be the process noise covariance matrix.

[0084] During the update phase of the Kalman filter algorithm, the observation vector for ,in, for The observed values, for The observed values, for The observed values, for The observed values.

[0085] Observation matrix for .

[0086] The Kalman gain equation is ,in, Here is the filter gain matrix. Observation matrix The transpose of the matrix, To observe the noise covariance.

[0087] The state update equation is ,in, For materials in The posterior state estimate at time t.

[0088] The covariance update equation is ,in, for The posterior estimate of the covariance at time t. It is an identity matrix.

[0089] In an optional embodiment, when using a semantic segmentation model, the acquired image sequence is defined as... ,in This represents the number of valid images acquired. The image sequence contains, for example: Figure 3 As shown Figure 1 , Figure 2 , Figure 3 ...Figure N-1, Figure N. Based on the semantic segmentation model, for each image in the image sequence... Perform pixel classification and output the classification results, which may include "coal," "gangue," or "background." Then, perform region feature statistics on the pixel classification results to obtain the results for each image. coal-producing areas account for a certain percentage of the total area. This refers to the area ratio of the coal region to the gangue region on the surface of the material under different viewing angles.

[0090] In calculating the area proportion of coal-producing regions At that time, the number of pixels whose classification result was "coal" was counted. The number of pixels classified as "gangue" Ignore pixels that are classified as "background".

[0091]

[0092] Therefore, the intelligent dry separation system can generate a set of data on the proportion of coal area for the same coal-gangue mixture. , In an optional embodiment, the data set can be represented as {Surface 1: 70% coal, Surface 2: 25% coal...Surface N: 60% coal}.

[0093] In a further optional embodiment, based on a data fusion algorithm, the estimated coal content from multiple perspectives can be calculated using the quantitative results of these single-view surface composition analyses. Different data fusion algorithms can be selected based on actual working conditions and user needs. For example, when a conservative recycling strategy is applicable, a maximum value fusion algorithm can be used. In other words, if a surface area with a high proportion of coal is considered to have recycling value, then the material is considered to have recycling value. When applying a balanced evaluation strategy, an average value fusion algorithm can be used. This means that the judgment is based on the average level of the coal area proportion of multiple surfaces. In other optional embodiments, other algorithms such as weighted averages can also be used, as long as they can meet the actual industrial needs.

[0094] In addition, this invention also includes a coal-gangue separation system based on an intelligent dry separation system, used to implement the coal-gangue separation method described above. The coal-gangue separation system of this invention includes a material distribution module, an identification module, a control module, and an execution module.

[0095] The fabric conveying module is used to transport coal, gangue, and coal-gangue mixtures. Figure 1 and Figure 3 The conveyor belt shown is the material distribution module. The identification module is used for image acquisition, preprocessing, and three-classification model recognition of the material on the material distribution module. For materials identified as coal mixed with gangue by the three-classification model, the identification module is also used to acquire image sequences of the material, calculate the estimated coal content, and obtain the final identification result of the material based on a preset sorting threshold. The control module is used to send an execution signal to the execution module based on the final identification result. The execution module is used to spray the material according to the execution signal, so that the material falls into the corresponding compartment.

[0096] The technical scope of this invention is not limited to the contents of the above specification. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the scope of this invention.

Claims

1. A method for separating coal and gangue based on an intelligent dry separation system, characterized in that, The method includes: Step 1: Continuously acquire images of the material on the intelligent dry separation system to obtain an original image set; perform image preprocessing on the original image set to obtain an image set to be identified. Step 2: Input the set of images to be identified into a three-class classification model based on visual recognition, and output the recognition results, which include "coal", "gangue" or "coal mixed with gangue"; Step 3: When the identification result is "coal", the intelligent dry separation system sprays the material into the coal bunker; when the identification result is "gangue", the intelligent dry separation system sprays the material into the gangue bunker; when the identification result is "coal mixed with gangue", proceed to the next step. Step 4: Collect image sequences of the material and calculate the estimated coal content of the material based on the image sequences; Step 5: Compare the estimated coal content with the preset sorting threshold. When the estimated coal content is greater than the sorting threshold, the intelligent dry separation system sprays the material into the coal bunker; when the estimated coal content is less than the sorting threshold, the intelligent dry separation system sprays the material into the gangue bunker.

2. The coal-gangue separation method as described in claim 1, characterized in that, Step 4 includes: Step 4-1: Track the movement trajectory of the material using a target tracking algorithm, and acquire images according to spatial or time intervals during the movement of the material to obtain the image sequence; Step 4-2: Calculate the area ratio of coal region in each image of the image sequence using a surface composition analysis algorithm; Step 4-3: Calculate the estimated coal content of the material by combining the coal area ratio of each image using a data fusion algorithm.

3. The coal-gangue separation method as described in claim 1, characterized in that, The three-classification model is pre-trained, and the pre-trained dataset consists of labeled images of coal interbedded with different morphological and surface features.

4. The coal-gangue separation method as described in claim 1, characterized in that, In step 3, when the identification result is "coal mixed with gangue", the intelligent dry separation system creates a unique identification code for the material.

5. The coal-gangue separation method as described in claim 2, characterized in that, The target tracking algorithm is the Kalman filter algorithm.

6. The coal-gangue separation method as described in claim 5, characterized in that, When tracking the motion trajectory of the material using the Kalman filter algorithm, the material is defined as... State vector at time step , in, Let x be the x-coordinate of the centroid position of the material. Let be the ordinate of the centroid position of the material. The width of the identification border of the material. The height of the identification border of the material. for The rate of change, for The rate of change, for The rate of change, for The rate of change, In the prediction phase of the Kalman filter algorithm, the state prediction equation is: in, For the material in The prior state estimate at time t. Here is the state transition matrix. For the material in The prior state estimate at time t. for Time and The time interval between moments The covariance prediction equation is in, for The prior estimate of covariance at time t, for The prior estimate of covariance at time t, State transition matrix The transpose of the matrix, The process noise covariance matrix is... During the update phase of the Kalman filter algorithm, the observation vector for in, for The observed values, for The observed values, for The observed values, for The observed values, Observation matrix for The Kalman gain equation is in, Here is the filter gain matrix. Observation matrix The transpose of the matrix, To observe the noise covariance, The state update equation is in, For the material in The posterior state estimate at time t. The covariance update equation is in, for The posterior estimate of the covariance at time t. It is an identity matrix.

7. The coal-gangue separation method as described in claim 2, characterized in that, In step 4-1, when the material's movement trajectory passes through image acquisition devices at different angles, or when the material's posture changes due to the movement of the intelligent dry sorting system, images of the material are acquired to obtain the image sequence.

8. The coal-gangue separation method as described in claim 2, characterized in that, Step 4-2 includes: Step 4-2-1: Define the image sequence as ,in To determine the number of valid images acquired, a semantic segmentation model is used to segment the image sequence. Each image in Perform pixel-level classification and output the classification results, which include "coal", "gangue" or "background"; Step 4-2-2: Calculate the value of each image based on the classification results. coal-producing areas account for a certain percentage of the total area. , in, For image The classification result mentioned above is the number of pixels for "coal". For image The number of pixels classified as "gangue" in the text; Step 4-2-3: Combine each image coal-producing areas account for a certain percentage of the total area. Obtain the data set of coal area proportion of the material. , .

9. The coal-gangue separation method as described in claim 2, characterized in that, In step 4-3, the estimated coal content of the material is calculated using either the maximum value fusion algorithm or the average value fusion algorithm. , When using the maximum value fusion algorithm ; When using the average fusion algorithm .

10. A coal and gangue separation system based on an intelligent dry separation system, capable of implementing the coal and gangue separation method as described in claim 1, characterized in that, The coal and gangue separation system includes a material distribution module, an identification module, a control module, and an execution module, wherein... The fabric module is used to convey the materials, including coal, gangue, and coal mixed with gangue. The identification module is used to perform image acquisition, preprocessing, and three-classification model identification on the material on the fabric module. When the identification result of the three-classification model is coal mixed with gangue, the identification module is used to acquire image sequences of the material, calculate the estimated coal content, and obtain the final identification result of the material according to a preset sorting threshold. The control module is used to send an execution signal to the execution module based on the final identification result. The execution module is used to spray the material according to the execution signal, so that the material falls into the corresponding compartment.

Images