Coal gangue identification method and device based on density

Abstract

The invention provides a coal gangue recognition method and device based on density, and the method comprises the steps: measuring a 3D model of a to-be-measured target object through a 3D measurement sensor, estimating the volume of the to-be-measured target object and the mass of the to-be-measured target object measured by a pressure sensor based on the measured 3D model, and then calculating the density of the to-be-measured target object. And if the density of the to-be-detected target object is within the density range of the coal gangue, identifying that the to-be-detected target object is the coal gangue. The device is arranged on a gangue conveying belt and comprises a 3D measurement sensor, a distributed pressure sensor, a truss type two-shaft gangue sorting mechanism and a mechanical gripper. The coal gangue and the coal gangue mixture can be accurately recognized, the gangue sorting accuracy is improved, and resource waste and environmental pollution can be effectively reduced; personnel participation is reduced, and digitization, intelligentization and unmanned operation of the coal gangue recognition technology are promoted; the hardware structure is simple, cost is low, operation is easy, and application and popularization are facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of coal gangue sorting, in particular to a density-based coal gangue identification method and device. Background Art

[0002] Traditional gangue identification technologies include manual, vibration, ultrasonic, radiographic, and visual methods. These methods suffer from low recognition rates, environmental pollution, high costs, and human hazards. In recent years, solutions based on sensor recognition and mechanical sorting have developed rapidly, with X-ray and visual recognition being the most representative. However, larger targets, especially complex ones like gangue mixtures, are difficult to penetrate using X-rays, and visual identification is also difficult. Therefore, accurate gangue identification, as the prerequisite and foundation for intelligent sorting, is a key technology for achieving digital, intelligent, and unmanned coal production, and has become a difficult problem that needs to be solved urgently.

[0003] The specific gravity (density) of gangue varies depending on its composition, origin, and other factors. Generally, gangue with a high carbon content and complex mineral composition has a lower specific gravity, while gangue with a high concentration of sandstone, limestone, or other components has a higher specific gravity, typically between 1.8 and 2.5. The specific gravity of common coal ranges from 1.2 to 1.8. Therefore, density identification can be used in intelligent coal sorting. However, when gangue mixtures are transported on conveyor belts, detecting the density of the minerals becomes crucial. Summary of the Invention

[0004] Aiming at the current intelligent coal sorting, which utilizes density recognition to identify gangue, the present invention provides a density-based gangue identification method and device.

[0005] The technical solution for achieving the technical objectives of the present invention is: a density-based coal gangue identification method, comprising the step of measuring the volume of a target object; in the step of measuring the volume of the target object, the volume of the target object conveyed on the gangue conveyor belt is measured using a 3D measurement sensor using a line laser measurement principle; the 3D measurement sensor is installed above the middle section of the gangue conveyor belt; and the method comprises the following steps:

[0006] Step 11, determine the system coordinate system: the conveying direction of the gangue conveyor belt is the x-axis direction, the vertical direction is the z-axis direction, and the horizontal direction perpendicular to the x-axis is the y-axis direction;

[0007] Step 12: Determine the kth target to be measured [x k ,y k ,l k ,d k ] to measure volume; where x k and y kis the center coordinate of the circumscribed rectangle of the target to be measured, l k and d k are the side lengths of the circumscribed rectangle along the x-axis and y-axis, respectively, k = 1, 2, 3, ...;

[0008] Step 13: When the kth target to be measured moves along the x-axis to the position of the 3D measurement sensor (5), a start 3D measurement signal is sent at time t1, and a stop 3D measurement signal is sent at time t2 to obtain its 3D model; wherein t1 and t2 are calculated as follows:

[0009]

[0010] L1 is the installation distance between the projection position of the CCD image sensor 3 of the visual recognition unit onto the gangue conveyor belt 9 and the projection position of the 3D measurement sensor 5 onto the gangue conveyor belt 9;

[0011] Step 14: Cluster and segment the point cloud based on the Euclidean distance to obtain the three-dimensional point data P of the target to be measured. k ;

[0012] P k =arg min l∈{n＝1,...,N} d[X k ,cluster1]

[0013] in,

[0014] ·X k Indicates the center point of the kth target to be measured, which is obtained by triangulating the pixel coordinates;

[0015] cluster l is the center of the candidate point cloud block;

[0016] ·d[X k ,cluster1] is the center point X k Cluster with candidate point cloud blocks l The Euclidean distance of

[0017] N is the number of targets to be measured;

[0018] n is the number of candidate point cloud blocks, and n <N;

[0019] Step 15: Using the conveyor belt as the reference plane and the normal direction of the plane as the projection direction, transform the point cloud coordinate system to coincide with the z-axis and its direction; project all points along this direction onto the conveyor belt plane to obtain the plane point cloud P. k '; In the plane point cloud P k 'Perform Voronoi division to obtain a series of planar polygons, each of which contains a point; extract Pk The z coordinate of the corresponding point is used to construct a prism using the polygon where the projection point of the point is located. The volume of all prisms is calculated to obtain the volume V of the kth target object to be measured. k ;

[0020]

[0021] Furthermore, the above-mentioned density-based gangue identification method further includes the step of calculating a volume correction coefficient:

[0022] Step 16: Randomly select a quantity W and measure the volume V k At the same time, the true volume V is obtained by the drainage method. k '; Count the average difference between the two and calculate the volume correction coefficient s,

[0023]

[0024] Step 17: All measured volumes V k Multiply by the correction coefficient s to get the corrected volume V of the kth target k ',

[0025] V k '=V k ×s.

[0026] Furthermore, the density-based gangue identification method further includes a step of measuring the mass of the target object, in which a truss-type two-axis gangue sorting mechanism is installed above the terminal section of the gangue conveyor belt; the ends of the truss-type two-axis gangue sorting mechanism are connected in sequence to a pressure sensor and a robotic gripper; and includes:

[0027] Step 21: Detect the force on the sensor in the Z axis:

[0028]

[0029] Among them, M is the mass of the gangue sorting mechanical gripper; m k is the mass of coal gangue; the gravity constant is g = 9.8N / kg;

[0030] Step 22: Calculate the mass of the currently captured quasi-gangue target.

[0031]

[0032] Furthermore, the above-mentioned density-based gangue identification method further includes:

[0033] Step 3: Calculate the density of the target:

[0034] ρ k =mk / V k ;

[0035] Step 4: Compare the calculated target density with the current average density ρ of actual coal obtained by statistics; when ρ k When ≥1.2ρ, the target is determined to be coal gangue.

[0036] Furthermore, in the above-mentioned density-based gangue identification method: a visual recognition unit is installed above the starting section of the gangue conveyor belt movement; the visual recognition unit visually identifies the coal and gangue on the gangue conveyor belt to obtain two types of targets: coal and measurement targets.

[0037] The present invention also provides a density-based gangue identification device for identifying quasi-gangue transported on a gangue conveyor belt. A 3D measurement sensor using a line laser measurement principle is installed above the middle section of the gangue conveyor belt. A truss-type two-axis gangue sorting mechanism is installed above the terminal section of the gangue conveyor belt. The ends of the truss-type two-axis gangue sorting mechanism are sequentially connected to a pressure sensor and a robotic gripper. A controller is sequentially connected to a visual recognition unit, a 3D measurement sensor, a pressure sensor, and a robotic gripper.

[0038] The 3D measuring sensor detects the volume of the quasi-gangue on the gangue conveyor belt, the robotic claw grabs the quasi-gangue, and uses the pressure sensor to measure the weight of the quasi-gangue;

[0039] The controller calculates the density of the quasi-gangue according to the volume and weight of the quasi-gangue, and identifies whether the quasi-gangue is gangue according to the density.

[0040] Furthermore, in the above-mentioned density-based gangue identification device: the 3D measurement sensor measures the 3D model of the quasi-gangue, and estimates the volume of the quasi-gangue based on the measured 3D model.

[0041] Furthermore, in the above-mentioned density-based gangue identification device: the mechanical claw (8) grabs the quasi-gangue, and the pressure sensor records the magnitude of the vertical force to estimate the mass of the quasi-gangue.

[0042] Furthermore, the above-mentioned density-based gangue identification device also includes a visual recognition unit, which is arranged above the starting section of the gangue conveyor belt; the visual recognition unit visually identifies the coal and gangue on the gangue conveyor belt to obtain two types of targets, coal and quasi-gangue, including a CCD image sensor.

[0043] The beneficial effects of the present invention are:

[0044] (1) The present invention can accurately identify coal gangue and coal gangue mixtures, improve the accuracy of gangue sorting, and effectively reduce resource waste and environmental pollution.

[0045] (2) Reduce personnel involvement and promote the digitalization, intelligence and unmanned operation of gangue identification technology;

[0046] (3) The hardware structure is simple, the cost is low, and it is easy to operate, which is conducive to promotion and application.

[0047] The present invention will be described in more detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Attachment Figure 1 This is a structural diagram of a device for density-based gangue identification according to Example 1 of the present invention;

[0049] Attachment Figure 2 In the embodiment 1 of the present invention, the point cloud P is k 'Perform Voronoi division to obtain a series of plane polygon graphics;

[0050] Attachment Figure 3 Extraction of P in Example 1 of the present invention k The z-coordinate of the corresponding point is used to construct a prismatic figure using the polygon where the projection point of the point is located. DETAILED DESCRIPTION

[0051] Example 1 Figure 1 This embodiment relates to a density-based, multi-source sensor composite gangue identification device and technology. The hardware components include: a computer 1, an illumination source 2, a CCD image sensor 3, a programmable controller 4, a 3D measurement sensor 5, a truss-type two-axis gangue sorting mechanism 6, a distributed pressure sensor 7, a robotic gripper 8, a gangue conveyor belt 9, power cables, and signal cables. The illumination source 2 and the CCD image sensor 3 form a visual recognition unit, installed above the starting section of the gangue conveyor belt 9. The 3D measurement sensor 5, which utilizes a line laser measurement principle, is installed above the middle section of the gangue conveyor belt 9. The truss-type two-axis gangue sorting mechanism 6 is installed above the terminal section of the gangue conveyor belt 9. The ends of the truss-type two-axis gangue sorting mechanism 6 are connected in sequence to the distributed pressure sensor 7 and the robotic gripper 8.

[0052] The system coordinate system is defined as follows: the conveying direction of the gangue conveyor belt 9 is the x-axis, and the horizontal and vertical directions of the truss-type two-axis gangue sorting mechanism 6 are the y-axis and the z-axis respectively.

[0053] When the device is in working state, coal and gangue are placed on the gangue conveyor belt 9, and the CCD image sensor 3 is used for visual recognition to obtain two types of targets, coal and quasi-gangue; when the quasi-gangue target moves along the x-axis to the position of the 3D measurement sensor 5, the programmable controller 4 sends a signal to start the 3D measurement sensor 5 and start measuring its 3D model; when the quasi-gangue target moves out of the position of the 3D measurement sensor 4 along the x-axis, the programmable controller 4 sends a stop measurement signal and estimates the volume of the target based on the measured 3D model; when the quasi-gangue target moves out of the position of the 3D measurement sensor 4 along the x-axis, the programmable controller 4 sends a stop measurement signal and estimates the volume of the target based on the measured 3D model; When the target moves along the x-axis to the location of the truss-type two-axis gangue sorting mechanism 6, the programmable controller 4 sends a grab signal to control the robot gripper 8 to grab it. At the same time, the distributed pressure sensor 7 starts to record the force in the z-axis direction until the movement stops. The mass of the robot gripper 8 is used as a known constant, and the mass of the target is estimated based on the relationship between force and mass. The density of the target is calculated based on the volume and mass of the target. Based on the density difference between coal and gangue, it is determined whether the target is gangue and thrown into the corresponding target area to complete the accurate identification of gangue. The specific steps are as follows:

[0054] Step 1: Start the system, calibrate the visual recognition unit using the pinhole imaging model, and establish the gangue recognition system coordinate system. The known system structural parameters are as follows: in the conveying direction of the gangue conveyor belt 9, that is, the x-axis direction, the installation distance between the projection position of the visual recognition unit's CCD image sensor 3 on the gangue conveyor belt 9 and the projection position of the 3D measurement sensor 5 on the gangue conveyor belt 9 is L1; the sampling rate of the 3D measurement sensor 5 is w seconds / time; the movement speed of the gangue conveyor belt 9 is v; the weight of the robot gripper 8 is M kilograms;

[0055] Step 2: Place the unsorted coal on the conveyor belt 9. When the coal and gangue are moved to the position below the visual recognition unit along with the gangue conveyor belt 9, the CCD image sensor 3 captures the image. Through visual recognition, two types of targets, coal and quasi-gangue, are obtained, and the kth quasi-gangue target (x k ,y k ,l k ,d k ). Where x k and y k is the center coordinate of the circumscribed rectangle of coal gangue, l k and d k are the side lengths of the rectangle along the x-axis and y-axis respectively, k = 1, 2, 3,….

[0056] Step 3: When the kth quasi-gangue target moves along the x-axis to the location of the 3D measurement sensor 5, the programmable controller 4 sends a start 3D measurement signal at time t1 and a stop 3D measurement signal at time t2 to obtain its 3D model. t1 and t2 are calculated as follows:

[0057]

[0058] Step 4: Cluster and segment the point cloud based on Euclidean distance to obtain the three-dimensional point data P of the target gangue point k .

[0059] P k =arg min l∈{n＝1,...,N} d[X k ,cluster1]

[0060] in,

[0061] ·X k represents the center point of the kth coal gangue to be identified, which is obtained by triangulating the pixel coordinates;

[0062] cluster l is the center of the candidate point cloud block;

[0063] ·d[X k ,cluster1] is the center point X k Cluster with candidate point cloud blocks l The Euclidean distance of

[0064] N is the number of gangue to be identified.

[0065] n is the number of candidate point cloud blocks, and n <N;

[0066] Step 5: Using the conveyor belt as the reference plane and the normal direction of the plane as the projection direction, transform the point cloud coordinate system to coincide with the z-axis and its direction; project all points along this direction onto the conveyor belt plane to obtain the plane point cloud P k '; In the plane point cloud P k 'Perform Voronoi division to obtain a series of planar polygons, each of which contains a point; extract P k The z coordinate of the corresponding point is used to construct a prism using the polygon where the projection point of the point is located. The volume of all prisms is calculated to obtain the volume V of the k-th gangue. k ;

[0067]

[0068] in,

[0069] i is the point P in the set k The serial number of the midpoint;

[0070] ·z i is the z coordinate of the point

[0071] ·S i is the area of the prism base.

[0072] Step 6: Since the point cloud data only covers the upper surface of the gangue, the volume data calculated using the prism will inevitably have a large error. Therefore, randomly select a number W of gangue with a particle size distribution of 50-400mm and use the above method to obtain its measured volume V k At the same time, the true volume V is obtained by the drainage method. k '. Statistical average difference between the two, and calculate the volume correction factor s,

[0073]

[0074] In actual sorting, all the measured volumes V k Multiply by the correction coefficient s to get the corrected volume V of the kth target k ',

[0075] V k '=V k ×s

[0076] Step 7: The gangue sorting mechanical gripper device includes a six-dimensional force sensor 7. When the k-th quasi-gangue target moves along the conveyor belt to the truss position, the mechanical gripper 8 grabs it. At the same time, the sensor 7 begins to record the force in the z-axis direction. According to physical knowledge, the force on the sensor during the uniform motion stage is equal to the gravity of the load.

[0077]

[0078] Among them, m k The mass of coal gangue is standard; the gravity constant is g = 9.8N / kg.

[0079] Step 8: According to formula (4), the mass of the currently captured quasi-gangue target can be calculated.

[0080]

[0081] The density of the target is,

[0082] ρ k =m k / V k

[0083] Step 9: Compare the calculated density with the average density ρ of the actual coal obtained through statistics. k When ≥1.2ρ, the target is judged to be coal gangue, and the robot gripper 8 is released while transporting the A area along the y-axis. k When the value is less than 1.2ρ, the target is determined to be non-gangue and is transported to area B along the y-axis to release the robot gripper 8.

[0084] Step 10: Reset the system device, clear the current calculation data, and prepare to sort new coal gangue.

Claims

1. A density-based gangue identification method, comprising the step of measuring the volume of a target object; characterized in that: In the step of measuring the volume of the target object, the volume of the target object transported on the gangue conveyor belt (9) is measured using a 3D measurement sensor (5) based on a line laser measurement principle; the 3D measurement sensor (5) is installed above the middle section of the gangue conveyor belt (9); and the steps include: Step 11, determine the system coordinate system: the conveying direction of the gangue conveyor belt (9) is the x-axis direction, the vertical direction is the z-axis direction, and the horizontal direction perpendicular to the x-axis is the y-axis direction; Step 12: Determine the kth target to be measured [x k ,y k ,l k ,d k ] to measure volume; where x k and y k is the center coordinate of the circumscribed rectangle of the target to be measured, l k and d k are the side lengths of the circumscribed rectangle along the x-axis and y-axis, respectively, k = 1, 2, 3, ...; Step 13: When the kth target to be measured moves along the x-axis to the position of the 3D measurement sensor (5), a start 3D measurement signal is sent at time t1, and a stop 3D measurement signal is sent at time t2 to obtain its 3D model; wherein t1 and t2 are calculated as follows: L1 is the installation distance between the projection position of the CCD image sensor 3 of the visual recognition unit onto the gangue conveyor belt 9 and the projection position of the 3D measurement sensor 5 onto the gangue conveyor belt 9; Step 14: Cluster and segment the point cloud based on the Euclidean distance to obtain the three-dimensional point data P of the target to be measured. k ; P k =arg min l∈{n＝1,...,N} d[X k ,cluster1] in, ·X k represents the center point of the kth target to be measured, which is obtained by triangulating the pixel coordinates; cluster l is the center of the candidate point cloud block; ·d[X k ,cluster1] is the center point X k Cluster with candidate point cloud blocks l The Euclidean distance of N is the number of targets to be measured; n is the number of candidate point cloud blocks, and n <N; Step 15: Using the conveyor belt as the reference plane and the normal direction of the plane as the projection direction, transform the point cloud coordinate system to coincide with the z-axis and its direction; project all points along this direction onto the conveyor belt plane to obtain the plane point cloud P. k '; In the plane point cloud P k 'Perform Voronoi division to obtain a series of planar polygons, each of which contains a point; extract P k The z coordinate of the corresponding point is used to construct a prism using the polygon where the projection point of the point is located. The volume of all prisms is calculated to obtain the volume V of the kth target object to be measured. k ; 2. The density-based gangue identification method according to claim 1, characterized in that: It also includes the steps for calculating the volume correction factor: Step 16: Randomly select a quantity W and measure the volume V k At the same time, the true volume V is obtained by the drainage method. k '; Count the average difference between the two and calculate the volume correction coefficient s, Step 17: Set all measured volumes V k Multiply by the correction coefficient s to get the corrected volume V of the kth target k ', V k '=V k ×s。 3. The density-based gangue identification method according to claim 1 or 2, characterized in that: The method further includes a step of measuring the mass of the target object, in which a truss-type two-axis gangue sorting mechanism (6) is installed above the terminal section of the gangue conveyor belt (9); the ends of the truss-type two-axis gangue sorting mechanism (6) are connected in sequence to a pressure sensor (7) and a mechanical gripper (8); include: Step 21: Detect the force on the sensor in the Z axis: Among them, M is the mass of the gangue sorting mechanical gripper; m k is the mass of coal gangue; the gravity constant is g = 9.8N / kg; Step 22: Calculate the mass of the currently captured quasi-gangue target.

4. The density-based gangue identification method according to claim 3, characterized in that: Also includes: Step 3: Calculate the density of the target: r k =m k / V k ; Step 4: Compare the calculated target density with the current average density ρ of actual coal obtained by statistics; when ρ k When ≥1.2ρ, the target is determined to be coal gangue.

5. The density-based gangue identification method according to claim 4, characterized in that: A visual recognition unit is installed above the starting section of the gangue conveyor belt (9); the visual recognition unit visually recognizes the coal and gangue on the gangue conveyor belt (9) to obtain two types of targets: coal and line measurement targets.

6. A device for identifying coal gangue based on density, for identifying quasi-coal gangue transported on a gangue conveyor belt (9); characterized in that: A 3D measurement sensor (5) using a line laser measurement principle is installed above the middle section of a gangue conveyor belt (9); a truss-type two-axis gangue sorting mechanism (6) is installed above the terminal section of the gangue conveyor belt (9); the ends of the truss-type two-axis gangue sorting mechanism (6) are connected in sequence to a pressure sensor (7) and a robotic gripper (8); and a controller (4) is connected in sequence to a visual recognition unit, the 3D measurement sensor (5), the pressure sensor (7), and the robotic gripper (8); The 3D measurement sensor (5) detects the volume of the quasi-gangue on the gangue conveyor belt (9), the mechanical claw (8) grabs the quasi-gangue, and uses the pressure sensor (7) to measure the weight of the quasi-gangue; The controller (4) calculates the density of the quasi-gangue according to the volume and weight of the quasi-gangue, and identifies whether the quasi-gangue is gangue according to the density.

7. The density-based gangue identification device according to claim 6, characterized in that: The 3D measurement sensor (5) measures the 3D model of the quasi-gangue and estimates the volume of the quasi-gangue based on the measured 3D model.

8. The density-based gangue identification device according to claim 6, characterized in that: The mechanical claw (8) grabs the quasi-gangue, and the pressure sensor (7) records the magnitude of the vertical force to estimate the mass of the quasi-gangue.

9. The density-based gangue identification device according to claim 6, characterized in that: The invention also includes a visual recognition unit, which is arranged above the starting section of the gangue conveyor belt (9); the visual recognition unit visually recognizes the coal and gangue on the gangue conveyor belt (9) to obtain two types of targets, namely coal and quasi-gangue, and includes a CCD image sensor (3).

Images