Intelligent tobacco impurity removal method, storage medium and system

Abstract

The invention discloses an intelligent tobacco impurity removal method, a storage medium and a system. A material monolayer processing device, an optical imaging unit, a detection processing unit, a time sequence and rejection control unit, an electrical control unit and a blowing electromagnetic valve are adopted to form the tobacco impurity removal method based on the intelligent tobacco impurity removal system of an image. According to the invention, an image recognition technology combining a color search algorithm and an AI intelligent algorithm is used, the types, distribution conditions and the like of sundries in the material can be counted, and according to different weights, forms and motion tracks of the sundries, the track change and floating of the sundries in the flying process are reduced through the algorithm, so that the accuracy of the sundries is improved. Richer and more accurate sundry movement data are provided for execution of subsequent sundry removal actions, and data support is provided for quality control analysis; and sundries in the fragments can be accurately removed, so that the purity of the tobacco leaves is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco leaf debris identification and removal, and in particular to an intelligent tobacco debris removal method, storage medium and system. Background Art

[0002] At present, the mainstream detection technology adopted by the photoelectric impurity removal equipment used in the tobacco processing process is the color lookup table (CLUT). This detection method has the problem of poor recognition effect on some light impurities, resulting in poor prevention and control effect on actual impurities in tobacco. In addition, there are many types of impurities in tobacco. This method cannot identify the types of impurities in tobacco, cannot estimate the movement state of impurities, and cannot count the distribution of impurities. In order to further improve the level of impurity prevention and control in the tobacco industry, and to track the information of impurities in tobacco, the impurities in tobacco can be more scientifically prevented and controlled. The present invention proposes a tobacco impurity removal method that combines a color lookup table to detect impurities with a deep convolutional neural network (DCNN) technology, which provides a new solution for quality safety and quality control in tobacco processing and other light material processing industries. Summary of the invention

[0003] In order to solve the problems existing in the above-mentioned prior art, the present invention provides an intelligent tobacco impurity removal method, storage medium and system, which solve the problems that the recognition effect of tobacco impurities is poor, resulting in unsatisfactory impurity prevention and control effect, the impurity type cannot be identified in the tobacco, the movement state of the impurities cannot be estimated, the distribution of the impurities cannot be statistically analyzed, and the tobacco quality cannot be traced and controlled.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] Firstly, the implementation of the method of the present invention is divided into the following functionalities: a material single-layer processing area, a picture taking area, an image processing area and a rejection area are sequentially arranged along the conveying direction, the material single-layer processing area realizes uniform spreading processing of tobacco materials, the picture taking area is provided with an imaging unit, the image processing area is provided with a detection and processing unit, and the rejection area is provided with a timing and rejection control unit.

[0006] The method comprises the following steps:

[0007] S1, passing the tobacco material through the monolayer treatment area for material monolayer treatment;

[0008] S2, continuously acquire images in the acquisition area using a line scan camera at a fixed line frequency and frame rate;

[0009] S3 uses the AI-based target detection model and color search algorithm to complete image recognition processing on the collected images, including:

[0010] S3.1, AI object detection model

[0011] Use a deep learning-based target detection model to obtain the location and category information of debris;

[0012] S3.2, Color Lookup Table Algorithm

[0013] Using a color lookup table algorithm, the colors occupied by tobacco in the RGB color space are marked by manual or automatic learning, and the unmarked colors are characterized as debris to form a color lookup table;

[0014] S3.3, Detection Result Fusion

[0015] The results of the target detection model and the color lookup table algorithm recognition are combined and processed to obtain the corresponding rejection grid matrix (after the material speed, field of view size, number of spray points and arrangement method information are determined, each frame of the image is divided according to the corresponding number of spray solenoid valves and line scan camera cycle to obtain the rejection grid matrix);

[0016] S4. Finally, the timing and rejection control unit will realize the timing synchronization control of the three actions of image acquisition, detection and processing, and rejection to ensure the consistency between the relationships between the execution actions, and then accurately reject the moving debris according to the rejection grid matrix.

[0017] The S5 rejection valve is installed above the belt outlet in the rejection area. The discrimination result is sent to the rejection valve through the timing and rejection control unit. The rejection valve releases compressed air to remove debris from the material. In order to prevent some special targets (such as blocky and lightweight targets, etc.) from rolling and other motion posture changes on the high-speed conveyor belt, which causes position deviation when they move to the rejection point, thereby affecting the rejection accuracy. According to the detected debris target category and position, the posture and running trajectory estimation algorithm is used to estimate the running posture and trajectory of special debris targets. After identifying the category of debris according to the target detection algorithm, the movement trajectory of the debris after being thrown into the air by the high-speed belt conveyor is calculated.

[0018] The tobacco or non-tobacco materials in the material are thrown out horizontally at a constant speed v0 at the end of the conveyor belt, and the thrown materials move in a projectile motion. That is, the material is thrown horizontally at an initial speed v0, and moves along a curve under the action of gravity. In the horizontal motion direction, the motion trajectory equation is:

[0019] Horizontal direction:

[0020] x(t)=v0*cos(θ)*t

[0021] Vertical direction:

[0022]

[0023] Where θ is the projection angle.

[0024] The material is parabolized at an initial velocity v0, and moves along a curved trajectory under the action of gravity. Under ideal physical conditions (excluding air resistance), the mass of an object will not affect its trajectory after projection. All objects will move along the same trajectory at the same initial velocity and projection angle, regardless of their mass. This is because in free fall, the acceleration of an object is determined only by gravity and has nothing to do with the mass of the object. However, in the production of actual materials, since tobacco is relatively light and thin, the accuracy of the debris removal action in tobacco materials is closely related to the motion state of the material. Therefore, after identifying the type of debris, the trajectory of the debris must be accurately inferred to bring better removal effects. Therefore, it is necessary to fully consider that if objects of different masses are thrown under the same conditions, the theoretical trajectory is the same, but when considering air resistance, the effect of air resistance on the trajectory is related to the mass, shape, speed, air density, etc. of the object. In general, air resistance will have a greater impact on light objects, causing them to decelerate faster, resulting in a different trajectory from that of heavy objects. Light objects will be more affected, resulting in changes in trajectory, and heavy materials will be less affected by air resistance.

[0025] Furthermore, a simplified model of air resistance is derived: for low-speed motion, it can be assumed that air resistance is proportional to speed, that is:

[0026] F drag =-b*v

[0027] Among them, b is a coefficient related to the shape of the object and the air density, and v is the speed.

[0028] The equation of motion is:

[0029]

[0030] Based on the above motion equation, the motion difference states of materials of different weights can be calculated, so that different objects can be removed more accurately.

[0031] The parabola slope of heavy objects (such as metal and stone) is large, while the parabola slope of light objects (such as feathers, plastic filaments and weeds) is small. The corresponding parameters in the debris rejection control matrix are calculated based on the parabola and slope parameters of the identified object. The position of the debris reaching the rejection point is estimated and corrected based on the corresponding rejection parameters. Then, the rejection grid matrix corresponding to the target detection is generated. The debris in the material is rejected according to the rejection grid matrix. At the same time, statistical analysis is performed on the target category and quantity of the debris, and the basic data required for quality control analysis is collected.

[0032] S6 qualified materials will not be sprayed, but will be thrown by the conveyor belt to the next level of material conveying equipment.

[0033] S7 counts factors such as production material batches, origins, and debris identification information, summarizes information within a period of time or multiple periods of time to generate a variety of statistical reports, and then analyzes the statistical reports. Ultimately, the conclusions of the reports are fed back to the tobacco processing debris prevention and control process to optimize production control and improve tobacco processing quality.

[0034] A computer-readable storage medium stores a computer program, which is executed by a processor to implement the steps of the intelligent tobacco impurity removal method of the present invention.

[0035] An intelligent tobacco impurity removal system includes a material single-layer processing area, a picture collection area, an image processing area and a rejection area which are arranged in sequence along the conveying direction. The material single-layer processing area realizes uniform spreading processing of tobacco materials, the picture collection area is provided with an imaging unit, the image processing area is provided with a detection processing unit, and the rejection area is provided with a timing and rejection control unit.

[0036] The imaging unit includes a camera, a lens and a light source, and the detection processing unit includes an AI image processor; the timing and rejection control unit is used to complete the detection processing, the timing synchronization control of the rejection action and the rejection valve action control, and the timing and rejection control unit includes a rejection unit, which is composed of a plurality of spraying solenoid valves and an air supply pipeline arranged in a row, and is used to achieve accurate rejection of debris by controlling the compressed air spray; and also includes a computer-readable storage medium described in the present invention.

[0037] Beneficial effects of the present invention:

[0038] I. The present invention can estimate the motion state of the identified debris, optimize the rejection signal according to the motion state of the debris, and thus improve the rejection accuracy;

[0039] II. Based on AI technology, the present invention can count the types and distribution of impurities in materials, providing data support for quality control analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the division of impurity removal functions of the present invention.

[0041] Figure 2 It is a detection and rejection workflow diagram of the present invention.

[0042] Figure 3 Schematic diagram of system deployment in an embodiment.

[0043] In the figure: 1-vibration feeding device, 2-high-speed belt conveyor, 3-optical imaging unit, 4-rejection unit, 5-frame, 6-detection and processing unit, 7-timing and rejection control unit. DETAILED DESCRIPTION

[0044] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.

[0045] like Figure 3 As shown, the system deployment based on the method of the present invention includes:

[0046] The imaging unit is composed of a pattern-mounted camera, lens, and light source. First, the image of the material on the conveyor belt is collected through the imaging unit. The detection and processing unit 6 is mainly composed of an AI image processor, an AI target detection model, and a color lookup table algorithm. The collected image uses a target detection model based on AI technology and a color lookup table algorithm to complete the detection of debris in the material. The timing and rejection control unit 7 is used for image acquisition, detection processing, and timing synchronization control between rejection actions and rejection valve action control. The rejection unit 4 is composed of multiple spray solenoid valves and air supply pipelines to control compressed air spraying to achieve accurate rejection of debris.

[0047] S1 First, the tobacco material is diluted by the vibrating feeding device 1. After the diluted material falls into the high-speed belt conveyor 2, since the conveying speed of the high-speed belt is much greater than the conveying speed of the vibrating feeding device 1, a better single-layer dilution effect of the tobacco can be obtained by differential speed.

[0048] The materials after S2 treatment enter Figure 1 The sampling area shown in the figure is Figure 3 The high-speed belt conveyor 2 shown in the figure conveys the tobacco to the corresponding position of the optical imaging unit 3, and takes real-time photos through the linear array camera in the optical imaging unit 3. In combination with the conveying state of the tobacco and the material characteristics, a linear light source is set up at the upper end of the material surface to illuminate the material, and then the line scanning camera scans and photographs the real-time produced material for image acquisition.

[0049] S3 uses the collected images to complete the detection of debris in materials by integrating the target detection model based on AI technology and the color search algorithm. The specific workflow is as follows: Figure 2 The detection processing workflow shown. The image data after taking the photo is transmitted to the detection processing unit 6, which is equipped with the processing hardware and software of the deep learning model algorithm. In the detection processing unit 6, the above-described deep learning target detection model and color lookup table algorithm are used to distinguish the debris. After the result is distinguished, the two detection results are fused and transmitted to the timing and rejection control unit 7. Among them:

[0050] a) AI target detection model

[0051] The location and category information of debris can be obtained by using a target detection model based on deep learning. In order to prevent some special targets (such as blocky and lightweight targets, etc.) from rolling on high-speed conveyor belts and other motion posture changes, which will cause position deviation when they move to the rejection point, thereby affecting the rejection accuracy. According to the detected debris target category and position, the posture and running trajectory estimation method is used to estimate the running posture and trajectory of special debris targets, and then the position of the debris reaching the rejection point is corrected according to the estimation results. Then the rejection grid matrix corresponding to the target detection is generated. At the same time, statistical analysis is performed on the target category and quantity of the debris, and the basic data required for quality control analysis is collected.

[0052] b) Color Lookup Table Algorithm

[0053] A color lookup table algorithm is used to mark the colors occupied by tobacco leaves in the RGB color space through manual or automatic learning, and unmarked colors are characterized as debris to form a color lookup table.

[0054] When identifying debris, after the entire image is searched, the search results are mapped to the elimination grid matrix, and the spray solenoid valve is controlled to spray the debris according to the results in each grid matrix.

[0055] c) Detection result fusion

[0056] The rejection grid matrix obtained by the target detection model and the color lookup table algorithm is processed to obtain the final rejection grid matrix.

[0057] After data fusion of the detection results, the coordinate position and type information of each debris will be obtained. Due to the different weights, shapes, and movement trajectories of different debris, this method is to reduce the change and floating of the trajectory of debris during flight, and the different throwing trajectories of different debris. Before sending the debris removal signal, the AI ​​image processor will calculate the movement trajectory of the debris flight and estimate the movement posture based on the coordinate position and type information of the debris, providing more abundant and accurate debris movement data for the execution of subsequent debris removal actions.

[0058] S4 uses a line scan camera to continuously collect images at a fixed line frequency and frame rate. When the material speed, field of view size, number of spray points and arrangement method information are determined, each frame of the image can be divided according to the corresponding number of spray solenoid valves and the line scan camera cycle to obtain a rejection grid matrix. The rejection grid matrix is ​​divided into the number of columns of the matrix by the number of rejection device units in the width direction of each frame of the image, and the pixels collected in the rejection control period are divided into the number of rows of the matrix by the pixels of each frame of the image height as the step length. A matrix is ​​constructed with the number of rows and columns to control the position and time of the rejection unit action; after the image is detected and processed, the rejection grid matrix of the target object to be rejected is obtained according to the shape, contour and other characteristics of the target object. After obtaining the rejection grid matrix, the timing and rejection control unit will realize the timing synchronization control of the three parts of image acquisition, detection processing and rejection to ensure the consistency between the execution action relationships, and then accurately remove the debris in motion. The timing and rejection control unit 7 sends the corresponding rejection signal in the rejection matrix to the rejection unit 4.

[0059] S5 Attachment Figure 3 The spray solenoid valve in the rejection unit 4 performs the spraying action according to the timing required by the system. After the spray solenoid valve is actuated, compressed air is blown out. The rejection valve is installed above the belt outlet in the rejection area. The compressed air blows the debris at the corresponding position into the debris collection box.

[0060] S6 qualified materials will not be sprayed, but will be thrown by the conveyor belt to the next level of material conveying equipment.

[0061] S7 counts factors such as production material batches, origins, and debris identification information, summarizes information within a period of time or multiple periods of time to generate a variety of statistical reports, and then analyzes the statistical reports. Ultimately, the conclusions of the reports are fed back to the tobacco processing debris prevention and control process to optimize production control and improve tobacco processing quality.

Claims

1. An intelligent tobacco impurity removal method, characterized in that: A material monolayer processing area, a picture collection area, an image processing area and a rejection area are sequentially arranged along the material conveying direction. The material monolayer processing area realizes uniform spreading of tobacco materials. The picture collection area is provided with an imaging unit. The image processing area is provided with a detection processing unit. The rejection area is provided with a timing and rejection control unit. The method comprises the following steps: S1, passing the tobacco material through the monolayer treatment area for material monolayer treatment; S2, continuously acquire images in the acquisition area using a line scan camera at a fixed line frequency and frame rate; S3 uses the AI-based target detection model and color search algorithm to complete image recognition processing on the collected images, including: S3.1, AI object detection model Use a deep learning-based target detection model to obtain the location and category information of debris; S3.2, Color Lookup Table Algorithm Using a color lookup table algorithm, the colors occupied by tobacco in the RGB color space are marked by manual or automatic learning, and the unmarked colors are characterized as debris to form a color lookup table; S3.3, the results of the target detection model and the color lookup table algorithm recognition are combined and processed to obtain the corresponding elimination grid matrix, including: Use a line scan camera to continuously collect images at a fixed line frequency and frame rate. When the material speed, field of view size, number of spray points and arrangement method information are determined, divide each frame of the image according to the corresponding number of spray solenoid valves and the line scan camera cycle to obtain a rejection grid matrix; S4. Finally, the timing and rejection control unit will realize the timing synchronization control of the three actions of image acquisition, detection and processing, and rejection to ensure the consistency between the relationships between the execution actions, and then accurately reject the moving debris according to the rejection grid matrix.

2. The intelligent tobacco impurity removal method according to claim 1, characterized in that: The step S3.3 also includes: Processing the rejection grid matrix obtained by the target detection model and the color lookup table algorithm to obtain the final rejection grid matrix; After data fusion of the detection results, the coordinate position and type information of each debris will be obtained. Due to the different weights, shapes, and motion trajectories of different debris, the algorithm is used to reduce the changes in the trajectory of debris during flight, floating, and the different throwing trajectories of different debris. Before sending the debris removal signal, the motion trajectory of the debris flight will be calculated and the motion posture will be estimated based on the coordinate position and type information of the debris, providing accurate debris movement data for the execution of subsequent debris removal actions.

3. The intelligent tobacco impurity removal method according to claim 2, characterized in that: According to the detected debris target category and position, the posture and trajectory estimation method is used to estimate the running posture and trajectory of special debris targets; after identifying the category of the debris according to the target detection algorithm, the movement trajectory of the debris after being thrown into the air by the high-speed belt conveyor is calculated.

4. The intelligent tobacco impurity removal method according to claim 3, characterized in that: When objects of different masses are thrown under the same conditions, the effect of air resistance on their trajectory is related to the mass, shape, speed and air density of the object. The simplified model of air resistance is: F drag =-b*v Among them, b is a coefficient related to the shape of the object and the air density, and v is the speed; The equation of motion is: According to the above motion equation, the motion difference states of materials of different weights can be calculated, so that different debris can be removed more accurately.

5. The intelligent tobacco impurity removal method according to claim 4, characterized in that: The parabola slope of the motion of heavy objects is large, while the parabola slope of the motion of light objects is small. The corresponding parameters in the debris rejection control matrix are calculated based on the motion parabola and slope parameters of the identified object, and then the position of the debris reaching the rejection point is estimated and corrected based on the corresponding rejection parameters.

6. The intelligent tobacco impurity removal method according to claim 1, characterized in that: The culling grid matrix is ​​a matrix that is formed by dividing each frame of image into the number of columns of the matrix at equal intervals according to the number of culling device units in the width direction of the conveying device, and dividing each frame of image height pixels into the number of rows of the matrix at equal intervals according to the pixels collected during the culling control period as the step length, and constructing a matrix with the number of rows and columns to control the corresponding position and time of the culling unit action; After the image is inspected and processed, a rejection grid matrix of the target object to be rejected is obtained according to the shape and contour characteristics of the target object, and the debris in the material is rejected according to the rejection grid matrix.

7. The intelligent tobacco impurity removal method according to any one of claims 1 to 6, characterized in that: It also includes statistics on production material batches, origins and debris identification information, summarizing information within a period of time or multiple periods of time to generate a variety of statistical reports, and then analyzing the statistical reports. Ultimately, the conclusions of the reports are fed back to the tobacco processing debris prevention and control process to optimize production control and improve tobacco processing quality.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the steps of the intelligent tobacco impurity removal method according to any one of claims 1 to 7.

9. An intelligent tobacco impurity removal system, characterized in that: The method comprises a material monolayer processing area, a picture collection area, an image processing area and a rejection area which are sequentially arranged along the conveying direction, wherein the material monolayer processing area realizes uniform spreading of tobacco materials, the picture collection area is provided with an imaging unit, the image processing area is provided with a detection processing unit, and the rejection area is provided with a timing and rejection control unit; The imaging unit includes a camera, a lens and a light source, and the detection processing unit includes an AI image processor; the timing and rejection control unit is used to complete the detection processing, the timing synchronization control of the rejection action and the rejection valve action control, and the timing and rejection control unit includes a rejection unit, which is composed of a plurality of spraying solenoid valves and an air supply pipeline arranged in a row, and is used to achieve accurate rejection of debris by controlling the compressed air spray; Also included is the computer-readable storage medium of claim 8.

Images