CN116017160A - An Intelligent Mineral Processing Method Based on Image Processing System - Google Patents

Abstract

The invention discloses an intelligent beneficiation method based on an image processing system, which belongs to the technical field of intelligent beneficiation, and comprises an image server, a patrol robot, a DTU, a PLC control box, an electric push rod assembly, a switch and an AP module; image data of the cradle collected by the inspection robot are transmitted to an image server through WIFI, the image server adopts a deep learning algorithm to mark the boundary between ore grains and impurities, and the image processing algorithm: the marking of the mine line is finished based on the deep learning method, so that the accuracy is high and the effect is good; modular system architecture: the system structure adopts a modularized design, so that the reconfigurability is strong, and the expansibility and compatibility are good; advanced and economical efficiency: mature and mainstream equipment is adopted to construct a system, and the system construction fully utilizes the current latest technologies such as video and audio, data, network and the like, and fully considers the continuous change of requirements and technologies. The system has high overall configuration performance, reasonable price and lower construction cost and investment.

Description

An Intelligent Mineral Processing Method Based on Image Processing System

technical field

The invention relates to the technical field of intelligent mineral processing, in particular to an intelligent mineral processing method based on an image processing system.

Background technique

The traditional tin ore shaking table needs to manually move the tray to pick up the ore after separation. Due to the long time of separation and the large number of shaking tables, a lot of labor and time costs are required. In order to respond to the national call, it is necessary to replace the backward manual method with an advanced intelligent method. Based on this, the present invention designs an intelligent mineral processing method based on an image processing system to solve the above problems.

Contents of the invention

The object of the present invention is to provide an intelligent mineral processing method based on an image processing system to solve the problems raised in the above-mentioned background technology.

In order to achieve the above object, the present invention provides the following technical solutions: an intelligent mineral processing method based on an image processing system, including an image server, an inspection robot, a DTU, a PLC control box, an electric push rod assembly, a switch and an AP module;

The inspection robot collects the image data of the shaker and transmits it to the image server through WIFI. The image server uses a deep learning algorithm to mark the boundary line between mineral particles and impurities, and then outputs the corresponding control signal according to the position of the boundary line. The control signal is forwarded by the DTU module To the PLC control box, the PLC control box controls the electric push rod assembly to automatically track the position of the mine line to complete the mine receiving task;

Mine access control strategy method:

1) Establish a coordinate system,

Take the intersection point of the boundary line on the water outlet side of the shaking table and the boundary line on the ore-connecting side of the shaking table as the coordinate origin O, and establish a coordinate system parallel to the boundary line on the ore-connecting side and the reverse flow direction as the positive direction;

2) image recognition data,

The image algorithm calibrates the boundary line according to different screening accuracy, and calculates the positions X1, X2, X3 of the intersection points of each tin mine boundary line and the coordinate system; the image algorithm will also try to detect the position X0 of the center line of the pallet on the coordinate system, which is In the first process, the target detection frame can be used to select the position of the pallet, or the midline position can be artificially marked on the pallet during design to facilitate image recognition;

3) Mobile control algorithm,

1. After each startup, the pallet returns to the coordinate origin 0 position;

2. The server collects the values of the boundary coordinates X1_n, X2_n, X3_n and X0_n of the nth time period; at the same time, the server should also record the values of the last boundary coordinates X1_n-1, X2_n-1, X3_n-1;

3. Select the coordinate Xi_n of the boundary line according to the requirements of the working conditions, i is the index of the boundary line, and n is the index of the time period. If the value of X0_n can be obtained, the moving distance of the controller is d_n=Xi_n-X0_n. If X0_n cannot be obtained, then The moving distance of the controller d_n=Xi_n-Xi_n-1;

4. The server issues control commands,

After the server calculates d_n, it can send it to the DTU, and add a custom script in the server to realize the conversion of "moving d_n" into a control command that the actuator controller can recognize; the custom script supports Python, JavaScript, Java, Rust , C/C++, Lua, Groovy and other languages, support secondary development, provide API interface, high compatibility and high scalability.

After receiving the control command from the server, the DTU directly forwards it to the mine receiving equipment controller to realize automatic and intelligent mine receiving operation.

Preferably, a gantry frame is set up above the shaker group for hoisting the track, and triangular support columns are added on both sides of the aisle to improve the structural stability of the gantry frame and avoid guide rails.

Playing area: Set the camera to stay in front of each shaker for 10 seconds to collect image information. The 24 shakers need to stay for 240 seconds in total; the moving path is: left end → right end → left end; the total travel distance is 72 meters, set the movement The camera runs at a speed of 5 m/min and takes approximately 20 minutes to complete a cycle.

Preferably, the next field area: set the camera to stay in front of each shaker for 10 seconds to collect image information, and the 22 shakers need to stay for 220 seconds in total; the moving path is: left end → right end → left end; the total stroke is 72 meters, Set the mobile camera to run at a speed of 5 m/min. It takes about 20 minutes to complete a cycle, and the shaker is connected to the cement wall.

Preferably, a layer of U-shaped support is installed on the cement fence, and it is stuck on the cement wall. Two rows of mounting holes are set on the U-shaped support, corresponding to the installation size of the guide rail of the electric cylinder; the slider of the electric cylinder and the splitter plate Connection, the electric cylinder slider drives the splitter plate to move left and right; the splitter plate is made of sheet metal.

Preferably, after the electric cylinder is powered on, it will feed back the position information of the current slider to the controller, and the camera will take pictures of the shaker, and after image processing, the position information of the water flow dividing line will be obtained, and the server will send the position information of the water flow dividing line to Electric cylinder controller, the controller sends position information instructions to the electric cylinder, the slider of the electric cylinder moves, and drives the diverter plate to move to the designated position.

Preferably, the establishment of the image recognition model requires a large amount of image data information, and the establishment of the basic model requires at least 50,000 pictures; in order to identify as much information as possible, not only the customer is required to provide a large amount of on-site image information, but also real-time processing at the processing site. Collect images for training; this system provides a special training mode. In the training mode, the inspection robot inspects and takes photos of each shaker in the Yunnan Tin Mine on the circular track, and sends the location information and image information to the server monitoring and management software. Platform, the monitoring management software calls the image model processing software to train the image model.

Preferably, after the monitoring and management software platform receives the picture taken by the inspection robot, the image labeling software will mark the image, and it is necessary to manually check whether the mark is accurate. If the mark is correct, the picture will enter the training library. If the mark is wrong, it needs to After manual correction, the pictures are then entered into the training library.

Preferably, the image model processing software regularly uses the training library to train the image model to enhance the recognition accuracy of the model.

Preferably, the identification process of the image acquisition is:

(1) The camera module installed on the inspection robot performs 30s video capture of the target area. When the light intensity at the collection position is set to be less than a certain value, the light next to the camera is automatically turned on to illuminate the shooting position to avoid a certain part of the position. The problem of the angle of light caused the pictures taken to be unclear, which is not convenient for the follow-up staff to observe the pictures, which affects the results of subsequent judgments. In the process of shooting small parts, the comparison size of the camera can be automatically enlarged after the system judgment, which is very important for the shooting. It can be used for direct observation without the need for staff to perform other operations on the picture, and it is more practical to carry out faster and accelerated work;

(2) transmit the collected target area image data and positioning information data to the server;

(3) The server receives and stores the target area image data and positioning information data, shooting time, and image shooting location data;

(4) The server marks the image data of the target area.

Compared with prior art, the beneficial effect of the present invention is:

Image processing algorithm: The method based on deep learning completes the labeling of mine lines, with high accuracy and good effect.

Modular system structure: The system structure adopts a modular design, which has strong reconfigurability, good scalability and compatibility.

Advancement and economy: mature and mainstream equipment is used to build the system, and the system construction makes full use of the latest audio-visual, data, network and other technologies, fully taking into account the continuous changes in demand and technology. The overall configuration performance of the system is high, the price is reasonable, and the construction cost and investment are low.

Of course, any product implementing the present invention does not necessarily need to achieve all the above-mentioned advantages at the same time.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that are required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. Those of ordinary skill in the art can also obtain other drawings based on these drawings without any creative effort.

Fig. 1 is a schematic diagram of the overall structure of the present invention;

Fig. 2 is a schematic diagram of the gantry frame in the field area of the present invention;

Fig. 3 is a schematic diagram of the gantry in the lower field area of the present invention;

Fig. 4 is the second schematic diagram of the gantry frame in the lower field area of the present invention;

Fig. 5 is a schematic diagram of a tripod of the present invention;

Fig. 6 is the connection schematic diagram of shaking bed and cement wall of the present invention;

Fig. 7 is the connection schematic diagram of cement wall and U-shaped support of the present invention;

Fig. 8 is a schematic diagram of a pallet for selecting a coordinate system in the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Please refer to Figures 1 to 8, the present invention provides a technical solution for an intelligent mineral processing system based on an image processing system: including an image server, a patrol robot, a DTU, a PLC control box, an electric push rod assembly, a switch and an AP module;

The inspection robot collects the image data of the shaker and transmits it to the image server through WIFI. The image server uses a deep learning algorithm to mark the boundary line between mineral particles and impurities, and then outputs the corresponding control signal according to the position of the boundary line. The control signal is forwarded by the DTU module To the PLC control box, the PLC control box controls the electric push rod assembly to automatically track the position of the mine line to complete the mine receiving task;

Mine access control strategy method:

1) Establish a coordinate system,

Take the intersection point of the boundary line on the water outlet side of the shaking table and the boundary line on the ore-connecting side of the shaking table as the coordinate origin O, and establish a coordinate system parallel to the boundary line on the ore-connecting side and the reverse flow direction as the positive direction;

2) image recognition data,

The image algorithm calibrates the boundary line according to different screening accuracy, and calculates the positions X1, X2, X3 of the intersection points of each tin mine boundary line and the coordinate system; the image algorithm will also try to detect the position X0 of the center line of the pallet on the coordinate system, which is In the first process, the target detection frame can be used to select the position of the pallet, or the midline position can be artificially marked on the pallet during design to facilitate image recognition;

3) Mobile control algorithm,

1. After each startup, the pallet returns to the coordinate origin 0 position;

2. The server collects the values of the boundary coordinates X1_n, X2_n, X3_n and X0_n of the nth time period; at the same time, the server should also record the values of the last boundary coordinates X1_n-1, X2_n-1, X3_n-1;

3. Select the coordinate Xi_n of the boundary line according to the requirements of the working conditions, i is the index of the boundary line, and n is the index of the time period. If the value of X0_n can be obtained, the moving distance of the controller is d_n=Xi_n-X0_n. If X0_n cannot be obtained, then The moving distance of the controller d_n=Xi_n-Xi_n-1;

4. The server issues control commands,

After the server calculates d_n, it can send it to the DTU, and add a custom script in the server to realize the conversion of "moving d_n" into a control command that the actuator controller can recognize; the custom script supports Python, JavaScript, Java, Rust , C/C++, Lua, Groovy and other languages, support secondary development, provide API interface, high compatibility and high scalability.

After receiving the control command from the server, the DTU directly forwards it to the mine receiving equipment controller to realize automatic and intelligent mine receiving operation.

Preferably, a gantry frame is set up above the shaker group for hoisting the track, and triangular support columns are added on both sides of the aisle to improve the structural stability of the gantry frame and avoid guide rails.

Playing area: Set the camera to stay in front of each shaker for 10 seconds to collect image information. The 24 shakers need to stay for 240 seconds in total; the moving path is: left end → right end → left end; the total travel distance is 72 meters, set the movement The running speed of the camera is 5 m/min. It takes about 20 minutes to complete a cycle. The next area: set the camera to stay in front of each shaker for 10 seconds to collect image information. The 22 shakers need to stay for 220 seconds in total; move The path is: left end → right end → left end; the total travel distance is 72 meters, and the moving camera is set at a speed of 5 m/min. It takes about 20 minutes to complete a cycle. The shaker is connected to the cement wall, and a U-shaped layer is installed on the cement fence. The bracket is stuck on the cement wall, and two rows of mounting holes are set on the U-shaped bracket, corresponding to the installation size of the guide rail of the electric cylinder; the slider of the electric cylinder is connected with the manifold, and the slider of the electric cylinder drives the manifold to move left and right The diverter plate is made of sheet metal. After the electric cylinder is powered on, it will feed back the position information of the current slider to the controller. The camera takes pictures of the shaking table. After image processing, the position information of the water flow boundary is obtained, and the server divides the water flow. Line position information is sent to the electric cylinder controller, the controller sends position information instructions to the electric cylinder, the slider of the electric cylinder moves, and drives the diverter plate to move to the designated position. The establishment of the image recognition model requires a large amount of image data information , the establishment of the basic model requires at least 50,000 pictures; in order to identify as much information as possible, not only does the customer need to provide a large amount of on-site image information, but also needs to collect images in real time at the processing site for training; this system provides a special training mode. In this mode, the inspection robot inspects and takes pictures of each shaker of Yunxi Mine on the circular track, sends the location information and image information to the server monitoring and management software platform, and the monitoring management software calls the image model processing software to train the image model. After the monitoring and management software platform receives the pictures taken by the inspection robot, the image labeling software will mark on the image, and it is necessary to manually check whether the label is accurate. If the label is correct, the picture will enter the training library. If the label is wrong, it needs to be corrected manually. , the picture enters the training library again, and the image model processing software regularly uses the training library to train the image model and strengthen the recognition accuracy of the model. The recognition process of the image acquisition is:

(1) The camera module installed on the inspection robot collects the target area for 30s. When the light intensity at the collection position is set to be less than a certain value, the light next to the camera is automatically turned on to illuminate the shooting position, avoiding the angle of a certain part of the position. The problem of light makes the pictures taken unclear, which is not convenient for the follow-up staff to observe the pictures, and affects the results of follow-up judgments;

(2) transmit the collected target area image data and positioning information data to the server;

(3) The server receives and stores the target area image data and positioning information data, shooting time, and image shooting location data;

(4) The server marks the image data of the target area.

The product model provided by the present invention is only for the use of this technical solution based on the structural characteristics of the product, and its product will be adjusted and transformed after purchase to make it more matching and in line with the technical solution of the present invention, which is one of the technical solutions The technical solution for the best application, the model of its product can be replaced and transformed according to its required technical parameters, which is well known to those skilled in the art, therefore, those skilled in the art can clearly pass through the provided by the present invention The technical solution obtains the corresponding use effect.

In the description of this specification, descriptions with reference to the terms "one embodiment", "example", "specific example" and the like mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment of the present invention. In an embodiment or example. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

The preferred embodiments of the invention disclosed above are only to help illustrate the invention. The preferred embodiments are not exhaustive in all detail, nor are the inventions limited to specific embodiments described. Obviously, many modifications and variations can be made based on the contents of this specification. This description selects and specifically describes these embodiments in order to better explain the principle and practical application of the present invention, so that those skilled in the art can well understand and utilize the present invention. The invention is to be limited only by the claims, along with their full scope and equivalents.

Claims (9)

1. An intelligent mineral processing method based on an image processing system, characterized in that it includes an image server, an inspection robot, a DTU, a PLC control box, an electric push rod assembly, a switch and an AP module; The inspection robot collects the image data of the shaker and transmits it to the image server through WIFI. The image server uses a deep learning algorithm to mark the boundary line between mineral particles and impurities, and then outputs the corresponding control signal according to the position of the boundary line. The control signal is forwarded by the DTU module To the PLC control box, the PLC control box controls the electric push rod assembly to automatically track the position of the mine line to complete the mine receiving task; Mine access control strategy method: 1) Establish a coordinate system, Take the intersection point of the boundary line on the water outlet side of the shaking table and the boundary line on the ore-connecting side of the shaking table as the coordinate origin O, and establish a coordinate system parallel to the boundary line on the ore-connecting side and the reverse flow direction as the positive direction; 2) image recognition data, The image algorithm calibrates the boundary line according to different screening accuracy, and calculates the positions X1, X2, X3 of the intersection points of each tin mine boundary line and the coordinate system; the image algorithm will also try to detect the position X0 of the center line of the pallet on the coordinate system, which is In the first process, the target detection frame can be used to select the position of the pallet, or the midline position can be artificially marked on the pallet during design to facilitate image recognition; 3) Mobile control algorithm, 1. After each startup, the pallet returns to the coordinate origin 0 position; 2. The server collects the values of the boundary coordinates X1_n, X2_n, X3_n and X0_n of the nth time period; at the same time, the server should also record the values of the last boundary coordinates X1_n-1, X2_n-1, X3_n-1; 3. Select the coordinate Xi_n of the boundary line according to the requirements of the working conditions, i is the index of the boundary line, and n is the index of the time period. If the value of X0_n can be obtained, the moving distance of the controller is d_n=Xi_n-X0_n. If X0_n cannot be obtained, then The moving distance of the controller d_n=Xi_n-Xi_n-1; 4. The server issues control commands, After the server calculates d_n, it can send it to the DTU, and add a custom script in the server to realize the conversion of "moving d_n" into a control command that the actuator controller can recognize; the custom script supports Python, JavaScript, Java, Rust , C/C++, Lua, Groovy and other languages. 2. An intelligent mineral processing method based on an image processing system according to claim 1, characterized in that: a gantry frame is set up above the shaker group for hoisting the track, and triangular support columns are added on both sides of the aisle. Playing area: Set the camera to stay in front of each shaker for 10 seconds to collect image information. The 24 shakers need to stay for 240 seconds in total; the moving path is: left end → right end → left end; the total travel distance is 72 meters, set the movement The camera runs at a speed of 5 m/min and takes approximately 20 minutes to complete a cycle. 3. A kind of intelligent mineral processing method based on image processing system according to claim 2, characterized in that: the next field area: the camera is set to stay in front of each shaker for 10 seconds for collecting image information, 22 shakers The bed needs to stay for a total of 220 seconds; the moving path is: left end → right end → left end; the total travel distance is 72 meters, and the moving camera is set at a speed of 5 m/min. It takes about 20 minutes to complete a cycle. The shaker is connected to the cement wall. 4. A kind of intelligent mineral processing method based on image processing system according to claim 3, it is characterized in that: add a layer of U-shaped support on the cement fence, be stuck on the cement wall, set two rows on the U-shaped support The installation hole corresponds to the installation size of the guide rail of the electric cylinder; the slider of the electric cylinder is connected with the manifold, and the slider of the electric cylinder drives the manifold to move left and right; the manifold is made of sheet metal. 5. The intelligent mineral processing method based on an image processing system according to claim 4, characterized in that: after the electric cylinder is powered on, it will feed back the position information of the current slider to the controller, and the camera will take pictures of the shaker , after image processing, the position information of the water flow dividing line is obtained, the server sends the position information of the water flow dividing line to the electric cylinder controller, the controller sends the position information command to the electric cylinder, the slider of the electric cylinder moves, and drives the diverter plate to move to the specified location. 6. a kind of intelligent mineral processing method based on image processing system according to claim 1, is characterized in that: the establishment of described image recognition model needs a large amount of image data information, and basic model needs 50000 pictures at least; In this mode, the inspection robot inspects and takes photos of each cloud tin mine shaking table on the circular track, and sends the location information and image information to the server monitoring management software platform, and the monitoring management software calls the image model processing software to train the image model. 7. The intelligent mineral processing method based on an image processing system according to claim 6, characterized in that: after the monitoring and management software platform receives the picture taken by the inspection robot, the image labeling software will mark the image , It is necessary to manually check whether the label is accurate. If the label is correct, the picture will enter the training database. If the label is wrong, it needs to be corrected manually before the picture enters the training database. 8. The intelligent mineral processing method based on an image processing system according to claim 7, characterized in that: said image model processing software regularly uses a training library to train the image model to enhance the recognition accuracy of the model. 9. A kind of intelligent mineral processing method based on image processing system according to claim 6, characterized in that: the identification process of the image acquisition is: (1) The camera module installed on the inspection robot performs 30s camera capture on the target area, and when the light intensity at the collection position is set to be less than a certain value, the light next to the camera is automatically turned on to illuminate the shooting position; (2) transmit the collected target area image data and positioning information data to the server; (3) The server receives and stores the target area image data and positioning information data, shooting time, and image shooting location data; (4) The server marks the image data of the target area.

Images