Flue-cured tobacco leaf quality evaluation method and device based on machine vision and near infrared

Abstract

The invention discloses a flue-cured tobacco leaf quality evaluation method and device based on machine vision and near infrared, and the evaluation method comprises the steps: analyzing the appearance indexes of tobacco leaves, determining the physical indexes of the tobacco leaves, measuring the chemical components of the tobacco leaves, and analyzing the images of the tobacco leaves through the machine vision to obtain the appearance indexes of the tobacco leaves; according to the method, physical indexes such as the thickness and the tensile strength of tobacco leaves are measured, various chemical components of the tobacco leaves are measured through near infrared rays, and finally evaluation of the tobacco leaf mass fraction, the industrial usability and the substitutability is obtained based on various index values of the tobacco leaves and a tobacco leaf mass weight model. The evaluation device comprises a tobacco leaf transportation area, an appearance and physical measurement area, a chemical component measurement area, a grading area and a main body structure.

Description

Technical Field

[0001] The present invention relates to the field of tobacco leaf evaluation, and in particular to a method and device for evaluating the quality of flue-cured tobacco leaves based on machine vision and near infrared. Background Art

[0002] When tobacco leaves are purchased, they are usually graded according to their appearance and other qualities. Different grades of tobacco leaves have different prices and different industrial availability. In addition to their high appearance and physical quality, high-grade tobacco leaves generally have more coordinated chemical compositions and better sensory evaluation. In addition, the leaf group formula of a brand of cigarettes is often composed of tobacco raw materials from multiple origins and different grades. When one raw material is in short supply, tobacco leaves of similar quality will be used as substitutes. At present, most of the grading in production is done manually, mainly through visual observation and empirical judgment, which has problems such as time-consuming, laborious, and subject to significant subjective influence; the methods and devices proposed in the field of tobacco leaf grading research are mostly single machine vision for feature and color judgment, and rarely involve physical quality and chemical composition detection. There are few methods for quality judgment after measuring multiple indicators in front-line production. Therefore, exploring and improving the evaluation method of tobacco leaf quality grade will help improve the industrial availability and substitutability of raw materials, so as to improve the quality of cigarette products and the stability of leaf group formula.

[0003] Machine vision uses an image capture device to shoot an object, analyzes the image, and extracts features from the image using models such as convolutional neural networks, allowing machines to replace human eyes in making measurements and judgments. It has been widely used in the fields of autonomous driving, face recognition, and product quality grade classification. Near-infrared spectroscopy uses a near-infrared light source to illuminate the sample, analyze the effective information carried by the object based on the transmitted and reflected light, and achieve accurate and rapid measurement of component content. It is now widely used in the fields of material composition analysis, food and agricultural product testing, and non-destructive testing. In the past, most of the patents in this research field were to generate evaluation values ​​of tobacco leaves only according to the chromaticity distribution, length, width and thickness of tobacco leaves through unified image processing and algorithm filtering (CN114757936A, CN115690030A, CN117990646A, CN103752531A), and there were problems such as missing supporting devices (CN117990646A, CN118464900A), fuzzy functional components of supporting devices or their functions could not correspond to the methods in detail (CN114757936A, CN115690030A); or the methods proposed in the patents were only to identify cracks, damages and other parts of tobacco leaves to obtain evaluation values ​​of tobacco leaves, but could not extract and analyze the color and damage distribution of specific parts of tobacco leaves separately (CN118464900A). The methods used could not enrich the model with the amount of recognition and did not have growth potential.

[0004] In addition, the previous methods in this research field mostly used different indicators to obtain evaluation indexes (CN118464900A), and obtained tobacco leaf quality evaluation by summing up the indices. Because different indicators have different weights for tobacco leaf quality evaluation, the actual impact of each indicator on tobacco leaf quality cannot be well reflected by the evaluation index alone; or by using the near-infrared spectrum image of tobacco leaves for similarity comparison (CN117990646A), when each indicator has a certain difference, the classification error tolerance is low, and tobacco leaves with large differences in the content of a single specific indicator are easily classified as unsuitable grades; or according to the numerical values ​​of various indicators of tobacco leaves, the perceptron predicts the grade (CN114757936A), and the tobacco leaf classification grade is determined only based on the similarity difference with the specific grade, so that some quality indicators of tobacco leaves may be different from the definition of their standard grades. Moreover, most of these patents are for single-grade classification based on the appearance quality or chemical composition of tobacco leaves, and it is rare to see methods that add the physical quality of tobacco leaves as the basis for classification. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a method and device for evaluating the quality of flue-cured tobacco leaves based on machine vision and near infrared.

[0006] The technical solution adopted by the present invention to solve the technical problem is: Flue-cured tobacco leaf quality evaluation method based on machine vision and near infrared, including Analysis of Appearance Indexes of S1 Tobacco Leaves The tobacco leaves are moved to the label holder along with the tobacco leaf conveyor belt, and the main vein base of the tobacco leaves is clamped to fix it. The mechanical arm in the physical area of ​​the appearance clamps a label from the label placement box, fixes the label to the base of the tobacco leaves through the label codec and swipes the serial number. After the label is fixed, the label is loosened to make the tobacco leaves fall off. The tobacco leaves continue to move along the tobacco leaf conveyor belt, and are moved to the center of the top of the photography table in the order of stacking. The tobacco leaves are spread out flat, and images of the tobacco leaves are collected. The image features of the tobacco leaves are extracted through the convolutional neural network model, and various appearance indicators of the tobacco leaves are analyzed. After acquiring the tobacco leaf image, S11 identifies the outline of the tobacco leaf, removes the background, and confirms the position of the characteristic part according to the outline shape and color distribution information of the tobacco leaf; S12 compares the features in the convolutional neural network feature library and classifies the features into types of damage or maturity such as variegated, damaged, and diseased; S13 corrodes the image of the characteristic part, selects the characteristic area and calculates the number of pixels, and further obtains the area of ​​the characteristic area: Feature area = number of pixels in the feature area * image shooting range area / total number of pixels in the image; S14 Different characteristic types have different effects on tobacco leaf quality, and the same characteristic type has different effects on different parts of tobacco leaves. Therefore, the location and area of ​​the characteristic area are used to evaluate the appearance index. Obtain data on the characteristic types, locations, and areas of tobacco leaves, and train the characteristic images obtained in this evaluation to enrich the model and improve recognition accuracy; Determination of physical indexes of S2 tobacco leaves The tobacco strip cutting knife and the tobacco circular punching knife are moved along the cutting knife moving track to the location of the tobacco leaves, and the cutting knife part is extended to the surface of the tobacco leaves to cut the tobacco leaves into tobacco strips with a fixed width and a number of tobacco sheets with the same diameter; The cut tobacco rod is placed on the tobacco rod tension measuring device by the mechanical arm in the appearance physical area. The two fixing clamps of the tobacco rod tension measuring device clamp the two sides of the tobacco rod respectively. With the help of the mechanical arm, the tobacco rod is slightly straightened. Then, the two fixing clamps are spread out to both sides until the tobacco rod breaks. The elongation and tension data of the tobacco rod are recorded at this time. The cut tobacco slices are placed on the tobacco slice thickness measuring device by the mechanical arm in the appearance physical area. The hydraulic telescopic rod of the tobacco slice thickness measuring device is lowered to contact the tobacco leaves, and the thickness of the tobacco slices at this time is recorded. The pressure is continuously increased, and the thickness change of the tobacco slices with the change of pressure is recorded. After the tobacco rods and tobacco leaves are measured, the robot arm in the appearance physics area cleans the measured samples and moves the tobacco leaves to the tobacco leaf conveyor belt; Measurement of chemical composition of S3 tobacco leaves The tobacco leaves are moved to the chemical composition measurement area along the tobacco leaf conveyor belt. The mechanical arm in the chemical composition area moves the tobacco leaves and places them on the top of the tobacco leaf placement table. The near-infrared transmitter moves along its moving track to the measurement points such as the sides of the main vein in the leaf, the leaf tip, the leaf base, and the leaf edges on both sides. After reaching each point, the near-infrared transmitter descends under the action of the lifting rod and contacts the surface of the tobacco leaves. Then, the near-infrared transmitter module emits near-infrared rays to the tobacco leaves, and the near-infrared receiver module receives the reflected rays. The received data is transmitted to the near-infrared device, and then it rises along the lifting rod and moves to the next point. The mechanical arm in the chemical composition area moves the tobacco leaves back to the tobacco leaf conveyor belt, and the near-infrared device reads the reflected rays of different wavelengths to obtain the main chemical composition content of the tobacco leaves; Quality evaluation of S4 tobacco leaves After obtaining the appearance indicators, physical indicators, and chemical composition data of the tobacco leaves, the scores of each indicator of the tobacco leaves are obtained according to the grading standards of each indicator, and the obtained scores are substituted into the indicator weight model to calculate the quality evaluation score of the tobacco leaves.

[0007] As a further improvement, the evaluation method further includes S5 Alternative Calculation After completing the quality evaluation of tobacco leaves, the industrial feasibility of the different tobacco leaves measured this time will be evaluated and cluster analysis will be performed to show the strength of substitutability between tobacco leaves of different grades.

[0008] As a further improvement, the evaluation method further includes S6 tobacco leaf grade classification After completing the quality evaluation of the tobacco leaves, the tobacco leaves are graded according to the quality evaluation scores. The tobacco leaves are clamped by the grading area robot arm and moved along the grading area moving track to the corresponding grading area for stacking.

[0009] The present invention also adopts the following technical solutions to solve the technical problems: Flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared, including The main structure and the components arranged thereon The tobacco transportation area includes a tobacco conveyor belt, one side of which is equipped with a label holder, the label holder is provided with a movable track, a main vein fixing clamp and a label codec are installed in the label holder, the main vein fixing clamp can move horizontally in the movable track, and the label codec can move horizontally and vertically in the movable track; the middle and downstream of the tobacco conveyor belt are respectively equipped with an appearance physical area robot arm and a chemical composition area robot arm; The appearance and physical measurement area includes a photographic table, one side of which is equipped with an industrial camera, and the other side is equipped with a cutter moving track, in which two independent transmission chains are arranged, and the two transmission chains are respectively connected to the tobacco leaf strip cutting knife and the tobacco leaf circular punching knife, and the appearance and physical measurement area is also equipped with a tobacco leaf thickness measuring device and a tobacco rod tension measuring device; The chemical composition measurement area includes a tobacco leaf placement platform, two movable tracks are installed in parallel on both sides of the tobacco leaf placement platform, a metal rod is spanned between the movable tracks, a built-in bridge-type movable track, and the three movable tracks together constitute a near-infrared movable track, a near-infrared emitter is installed in the bridge-type movable track, and a near-infrared device is installed in the vacant area of ​​the chemical composition measurement area as a reflected light different band information analysis device; The grading area includes a grading area moving track, which is perpendicular to the tobacco conveyor belt. A grading area robot arm is installed in the grading area moving track, and its movable range covers the entire grading area. A plurality of grading boxes are arranged on one side of the grading area moving track, and a tobacco storage box is installed next to each grading box for storing tobacco leaves after single batch grading.

[0010] Furthermore, the tobacco conveyor belt structure is a roller conveyor belt with a leather surface material.

[0011] Furthermore, the label holder structure is a raised column shape.

[0012] Furthermore, the main vein fixing clamp structure is a mechanical clamp to clamp the tobacco stem on the side, and the tag codec has built-in NFC and an electromagnet for data import and tag codec switch.

[0013] Furthermore, the mechanical arm in the appearance and physical area has the same structure as the mechanical arm in the chemical composition area, and both have a rotating base connected to the top of the main structure, which can drive the mechanical arm to rotate. A multi-joint movable arm is installed on the rotating base, and a mechanical gripper is installed on the top of the movable arm. Its movable range covers the adjacent sections of the tobacco conveyor belt, and covers all positions of the appearance and physical measurement area and the chemical composition measurement area respectively.

[0014] Furthermore, fixed baffles are installed on the side and downstream end of the tobacco conveyor belt to prevent the device from being damaged by collision and tobacco powder from falling into the device.

[0015] Furthermore, a label placement box is installed on one side of the appearance physical area robot arm, and the structure is a rectangular box with a missing top surface, which is used for placing and storing labels.

[0016] Furthermore, the appearance and physical measurement area is located upstream of the tobacco leaf transportation area.

[0017] Furthermore, the photographic table structure is a rectangular prism protrusion made of frosted material.

[0018] Furthermore, the industrial camera structure is an inverted "L" shape, with a camera pointing vertically downward installed at the top end, and its photographic range covers the entire top of the photographic table.

[0019] Furthermore, the tobacco strip cutting knife is a circular roller with a certain spacing, which is installed together on the same roller, and the roller is connected to a horizontal telescopic rod, and the horizontal telescopic rod is installed on a movable base, and the movable base is connected to the inside of the cutting knife movable track; the tobacco circular punching knife is a circular ring-shaped twisting knife, which is installed on a horizontal telescopic rod, and the horizontal telescopic rod is installed on a movable base, and the movable base is also connected to the inside of the cutting knife movable track.

[0020] Furthermore, the tobacco leaf thickness measuring device includes a cylindrical base, a measuring rod is installed on one side of the cylindrical base, and a hydraulic column with a downward direction and adjustable pressure is provided on the top of the measuring rod, which is opposite to the center of the cylindrical base and can record the moving distance during the extension and retraction process; the structure of the tobacco rod tension measuring device is a raised prism, and a moving track is installed on one side of the prism. Two mechanical clamps with the same structure are installed in the moving track, which can open and merge along the track and have force and distance measurement functions.

[0021] Further, the chemical composition measurement area is located downstream of the tobacco leaf transportation area.

[0022] Furthermore, the near-infrared transmitter is a probe composed of a near-infrared transmitter and a receiver, and has a vertical lifting rod at the top, which is connected to the bridge-type moving track and can drive the near-infrared transmitter to perform lifting movement in the vertical direction.

[0023] Furthermore, the grade classification area is located at the end of the tobacco leaf transportation area.

[0024] Furthermore, the grade classification box is a semi-enclosed box body and is arranged adjacent to each other, and the tobacco storage box is a rectangular box body with a missing top surface.

[0025] The beneficial effects brought by the present invention are: The present invention introduces machine vision to shoot tobacco leaves, and then uses a convolutional neural network model to analyze the image features of tobacco leaves, so as to achieve rapid and accurate measurement of indicators such as maturity, damage, color, oil content and leaf area of ​​tobacco leaves, thereby making up for the problems that previously existed in the field of tobacco leaf grading research, such as fuzzy extraction of tobacco leaf features, inability to directly identify and classify tobacco leaf tissues in specific areas separately, and inability to perform specific divisions according to feature conditions; The present invention measures the content of the main internal chemical components of tobacco leaves by introducing near-infrared spectroscopy measurement, and obtains the coordination of the chemical components of tobacco leaves by substituting the contents of indicators such as sugar, nitrogen, nicotine, potassium, and chlorine into a tobacco leaf quality index weight model; at the same time, through computer programming control technology, full automatic control is achieved in the tobacco leaf quality measurement process, and each quality index data of tobacco leaves is automatically analyzed, so as to achieve the purpose of high process efficiency and reduce human errors; The present invention also introduces the analysis of the thickness, tensile strength and elasticity of tobacco leaves and a matching cutting and measuring device to improve the quality index of tobacco leaves and enhance the accuracy of quality judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments: Figure 1 This is a top view of the structure of a flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared according to the present invention; Figure 2 This is an isometric view of the structure of a flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared according to the present invention; Figure 3 This is an isometric view of the appearance and physical measurement area of ​​a flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared according to the present invention; Figure 4 This is a structural isometric view of the chemical composition measurement area of ​​a flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared according to the present invention; Figure 5It is a schematic structural diagram of a near-infrared emitter probe in a chemical composition measurement area of ​​a flue-cured tobacco leaf quality evaluation device based on machine vision and near-infrared according to the present invention; Figure 6 A schematic diagram of a tobacco leaf appearance index analysis and evaluation method of a flue-cured tobacco leaf quality evaluation method based on machine vision and near infrared according to the present invention; Figure 7 This is a diagram showing an evaluation test of a method and device for evaluating flue-cured tobacco leaf quality based on machine vision and near infrared according to the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Example 1 This embodiment relates to a flue-cured tobacco leaf quality evaluation method based on machine vision and near infrared. Before the evaluation method is implemented, the tobacco leaves to be tested are selected and the device parameters are set: According to local tobacco leaf purchasing technical specifications or experimental requirements, the tobacco leaves that need to be evaluated will be untied, flattened and other pre-processed, and the tobacco leaves infected with infectious diseases will be removed. The remaining tobacco leaves will be neatly stacked in order and placed at the starting end of the tobacco leaf conveyor belt.

[0029] Turn on the power of the device at the control panel 52, and selectively set the tobacco leaf indicators that need to be collected this time (including: tobacco leaf appearance indicators: maturity, oil content, color, damage, leaf area; tobacco leaf physical indicators: tension resistance, elasticity, identity and tissue structure; tobacco leaf chemical indicators: sugar, nitrogen, nicotine, potassium, chlorine; the determination of physical indicators requires cutting of the tobacco leaves, which will cause damage, and each indicator can be selected to be turned on or off separately). After determining the indicators, start the device.

[0030] The flue-cured tobacco leaf quality evaluation method comprises Analysis of Appearance Indexes of S1 Tobacco Leaves The tobacco leaves are moved to the label holder 12 along with the tobacco leaf conveyor belt 11, and the main vein fixing clamp 13 clamps the main vein base of the tobacco leaves to fix it. The appearance physical area mechanical arm 15 clamps a label from the label placement box 18 and transfers it to the label codec 14. The label is fixed to the base of the tobacco leaves and the serial number is brushed in through the label codec 14. After the label is fixed, the main vein fixing clamp 13 is loosened to make the tobacco leaves fall off. The tobacco leaves continue to move along the tobacco leaf conveyor belt 11, and the mechanical arm 15 of the appearance physical area moves the tobacco leaves to the top center position of the photographic table 21 in the stacking order, and spreads the tobacco leaves flat. The industrial camera 22 collects the image of the tobacco leaves and transmits it to the built-in computer. The image features of the tobacco leaves are extracted through the convolutional neural network model, and various appearance indicators of the tobacco leaves are analyzed; After acquiring the tobacco leaf image, S11 identifies the outline of the tobacco leaf, removes the background, and confirms the position of the characteristic part according to the outline shape and color distribution information of the tobacco leaf; S12 compares the features in the convolutional neural network feature library and classifies the features into types of damage or maturity such as variegated, damaged, and diseased; S13 corrodes the image of the characteristic part, and selects the characteristic area to calculate the number of pixels. Since the method and its supporting device have a constant shooting range and the number of pixels of the shot image, the area of ​​the characteristic area can be further obtained: Feature area = number of pixels in the feature area * image shooting range area / total number of pixels in the image; S14 Different types of features have different effects on tobacco leaf quality. For example, the same area of ​​damage has a greater impact than variegated color, and the same feature has different effects on different parts of the tobacco leaf. For example, the damage in the leaf core area has a greater impact than the damage on the edge. Therefore, the position and area of ​​the feature area are used to make a more accurate judgment on tobacco leaf quality and industrial availability. Reference Figure 6 According to the above steps, based on the tobacco leaf shooting image, the tobacco leaf surface is identified and the type of the characteristic area (such as damage, variegated color, diseased spot, etc.) is determined; after the characteristic type is identified, the area of ​​the corresponding characteristic part is calculated, and the impact of the bad area is weighted and evaluated according to the area size and the position of the characteristic part; Obtain data on the characteristic types, locations, and areas of tobacco leaves, and train the characteristic images obtained in this evaluation to enrich the model and improve recognition accuracy; If overfitting or underfitting occurs, the training samples and parameters can be changed. For example, the final loss function usually controls the overfitting of weights by increasing the L2 norm, and controls the intensity of the overfitting effect by the parameter λ (weight decay).

[0031] In addition, this embodiment adopts a typical convolutional neural network model, which is mainly composed of an input layer, a convolutional layer, a downsampling layer (pooling layer), a fully connected layer, and an output layer. The input of the convolutional neural network is usually the original image X, which is passed alternately through multiple convolutional layers and downsampling layers. The convolutional neural network relies on a fully connected network to classify the extracted features and obtain a probability distribution Y based on the input.

[0032] Determination of physical indexes of S2 tobacco leaves If the measurement of physical indicators is turned on, the tobacco strip cutting knife 24 and the tobacco circular punching knife 25 are moved along the cutting knife moving track 23 to the position of the tobacco leaves, and the cutting knife part is extended to the surface of the tobacco leaves to cut the tobacco leaves into tobacco strips of fixed width and a number of tobacco sheets with the same diameter; in this process, the camera image of the industrial camera can be analyzed in real time by the built-in computer to ensure that the position is accurate during cutting and does not contain adverse effects such as veins and burrs; The cut tobacco rod is placed on the tobacco rod tension measuring device 27 by the mechanical arm 15 in the appearance physical area. The two fixing clamps of the tobacco rod tension measuring device 27 clamp the two sides of the tobacco rod respectively. With the help of the mechanical arm, the tobacco rod is slightly straightened. Then, the two fixing clamps are spread out to both sides until the tobacco rod breaks. The elongation and tension data of the tobacco rod are recorded at this time. The cut tobacco slices are placed at the position of the tobacco slice thickness measuring device 26 by the mechanical arm 15 in the appearance physical area. The hydraulic telescopic rod of the tobacco slice thickness measuring device 26 is lowered to contact the tobacco leaves, and the thickness of the tobacco slices at this time is recorded. The telescopic rod continuously increases the pressure, and the thickness change of the tobacco slices under the pressure change is recorded; After the tobacco rods and tobacco leaves are measured, the mechanical arm 15 in the appearance physics area cleans the measured samples and moves the tobacco leaves to the tobacco leaf conveyor belt 11; Measurement of chemical composition of S3 tobacco leaves The tobacco leaves are moved to the chemical composition measurement area 3 along the tobacco leaf conveyor belt 11. The chemical composition area mechanical arm 16 moves the tobacco leaves and places them on the top of the tobacco leaf placement platform 31. The near-infrared transmitter 33 moves along its moving track 32 to the measurement points such as the two sides of the main vein in the leaf, the leaf tip, the leaf base, and the leaf edges on both sides. After reaching the corresponding point, the near-infrared transmitter 33 descends under the action of the lifting rod and contacts the surface of the tobacco leaves. Then, the near-infrared transmitter module emits near-infrared rays to the tobacco leaves, and the near-infrared receiver module receives the reflected rays. The received data is transmitted to the near-infrared device 34, and then it rises along the lifting rod and moves to the next point. The chemical composition area robot arm 16 moves the tobacco leaves back to the tobacco leaf conveyor belt 11, and the near infrared device 34 reads the reflected rays of different wavelengths to obtain the main chemical composition content of the tobacco leaves; Quality evaluation of S4 tobacco leaves After obtaining the appearance indicators, physical indicators, and chemical composition data of the tobacco leaves, the scores of each data of the tobacco leaves are obtained according to the established grading-scoring standards for each indicator, and the obtained scores are substituted into the indicator weight model to calculate the quality evaluation score of the tobacco leaves.

[0033] S5 Alternative Calculation After completing the quality evaluation of tobacco leaves, the different tobacco leaves measured this time will be evaluated for industrial feasibility and cluster analysis will be performed to show the strength of substitutability between tobacco leaves of different grades.

[0034] S6 tobacco leaf grade classification After the quality evaluation of the tobacco leaves is completed, the tobacco leaves are graded according to the quality evaluation scores, and the tobacco leaves are clamped by the grading area robot 42 and moved along the grading area moving track 41 to the corresponding grading area for stacking.

[0035] Of course, before the evaluation device is started, the purchasing staff or the experimental staff can input the tobacco leaf data required for calculating the substitutability into the control console 52. When the device obtains tobacco leaves with similar quality evaluation, the tobacco leaves are placed separately through the grade classification area.

[0036] It should be noted that the order of the steps in this embodiment is only for the purpose of clear description, and does not mean that the steps must be executed in a strict order. The execution order can be flexibly adjusted according to the selectively set tobacco leaf collection indicators.

[0037] Example 2 This embodiment relates to a flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared to achieve the determination of various quality indicators of tobacco leaves. Figure 1~Figure 5 , the evaluation device comprises: Tobacco transportation area 1, appearance and physical measurement area 2, chemical composition measurement area 3, grade classification area 4, main structure 5. In the above structure, the main structure 5 is divided into a device table 51 and a control console 52, wherein the device table 51 is a box-shaped structure, and the other functional components are installed on the top surface of the device table 51. The built-in computer is connected to each component and the control console 52. The control console 52 is a touch monitor with a data interface to realize data transmission, serving as an interactive platform between the device and the operator.

[0038] The tobacco transport area 1 includes a tobacco conveyor belt 11, which is a roller conveyor belt made of leather. A label holder 12 is installed on one side of the belt. The label holder 12 is a raised columnar structure with a movable track, in which a main vein fixing clamp 13 and a label codec 14 are installed. The main vein fixing clamp 13 can move horizontally in the movable track, and the label codec 14 can move horizontally and vertically to avoid being affected by the tobacco stem when handing over the label. The main vein fixing clamp 13 is a mechanical clamp with a horizontal 90° bend to clamp the tobacco stem on its side. The label codec 14 is a common mechanical clamp design with built-in NFC and electromagnets for data import and label codec switches.

[0039] The mechanical arm 15 of the appearance physics area and the mechanical arm 16 of the chemical composition area are installed in the middle and downstream of the tobacco conveyor belt 11. Except for the installation position, the other structures of the two are the same. There are two copies of them. Both have a rotating base connected to the top of the device table 51, which can drive the mechanical arm to rotate. A multi-joint movable arm is installed on the rotating base, and a mechanical gripper is installed on the top of the movable arm. Its range of movement covers the adjacent sections of the tobacco conveyor belt 11, and covers the entire position of the appearance and physical measurement area 2 and the chemical composition measurement area 3 respectively. Fixed baffles 17 are installed on the side and downstream end of the tobacco conveyor belt 11 to prevent collision damage between motorized devices, tobacco powder falling into the device, etc. A label placement box 18 is installed on one side of the mechanical arm 15 of the appearance physics area. The structure is a rectangular box with a missing top surface, which is used for label placement and storage.

[0040] The appearance and physical measurement area 2 is located upstream of the tobacco transportation area 1, and includes a photographic table 21, which is a rectangular prism protrusion, and is made of matte and cut-resistant material. An industrial camera 22 is installed on one side of the photographic table 21, and the structure is an inverted "L" shape. A camera is installed at the top end facing vertically downward, and the photographic range covers the entire top of the photographic table 21. A cutter moving track 23 is installed on the other side of the photographic table 21, which has a stepping motor as a motor, and there are two independent transmission chains in the track as a transmission. The two transmission chains are respectively connected to the tobacco strip cutting knife 24 and the tobacco circular punching knife 25. The main structure of the tobacco strip cutting knife 24 is a circular roller with a certain spacing, which is installed on the same roller shaft, and the roller shaft is connected to the horizontal telescopic rod, which is installed on the mobile base, and the mobile base is connected to the inside of the cutter moving track 23; the main structure of the tobacco circular punching knife 25 is a ring-shaped twisting knife, which is installed on the horizontal telescopic rod, which is installed on the mobile base, and the mobile base is also connected to the inside of the cutter moving track 23.

[0041] In the vacant area of ​​the appearance and physical measurement area 2, a tobacco leaf thickness measuring device 26 is installed, which has a cylindrical base structure, and a measuring rod is installed on one side of the base. A hydraulic column with a downward direction and adjustable pressure is provided on the top of the measuring rod, and its axis is directly opposite to the center of the cylindrical base. The moving distance can be recorded during the extension and retraction process; in addition, a tobacco rod tension measuring device 27 is installed, which has a raised prism structure, and a moving track is installed on one side of the prism. Two mechanical grippers with the same structure are installed in the moving track, which can open and merge along the track and have force and distance measurement functions.

[0042] The chemical composition measurement area 3 is located downstream of the tobacco transport area 1, and includes a tobacco placement platform 31, which is a raised rectangular prism. Two movable tracks are installed in parallel on both sides of the prism. A metal rod body is crossed between the tracks, and a bridge-type movable track is built in, which is perpendicular to the two movable tracks. The three tracks together constitute a near-infrared movable track 32. A near-infrared transmitter 33 is installed in the bridge-type movable track, which can move precisely along the track on the horizontal plane. The near-infrared transmitter 33 is a probe composed of a near-infrared transmitter and a receiver, with a vertical lifting rod at the top, connected to the track, which can drive the near-infrared transmitter 33 to move up and down in the vertical direction. A near-infrared device 34 is installed in the vacant area of ​​the chemical composition measurement area 3 as a device for analyzing information of different bands of reflected light.

[0043] The grading area 4 is located at the end of the tobacco transport area 1, and has a grading area moving track 41, which is perpendicular to the tobacco conveyor belt 11. A grading area mechanical arm 42 is installed in the grading area moving track 41, which has a mobile base. The structure on the base is the same as that of other mechanical arms, and its range of movement covers the entire grading area 4. A grading box 43 is installed on one side of the grading area moving track 41. The structure is a semi-enclosed box body, multiple copies are arranged adjacently, and a tobacco storage box 44 is installed next to each box body. The structure is a rectangular box body with a missing top surface, which is used for storing tobacco leaves after a single batch of grading.

[0044] Example 3 The method of the present invention involves measuring different quality indicators of tobacco leaves to obtain the score of each indicator, and substituting the score of each indicator into the indicator weight model to calculate the quality score of the tobacco leaves. This embodiment aims to illustrate the scores and weights of appearance indicators and chemical indicators (only as an example basis to fully reflect the model calculation process of the present invention).

[0045] It is worth noting that the names, scores and weights of specific indicators need to be changed accordingly in combination with the tobacco grading technical specifications of the producing areas and this method. The examples do not represent the final results.

[0046] Table 1 shows the grading and scoring standards of appearance indicators. The image features of tobacco leaves are extracted and analyzed by the convolutional neural network model to obtain the data of various appearance indicators of tobacco leaves, such as the characteristic type, position and area of ​​tobacco leaves. The scores of various indicators can be obtained by combining Table 1.

[0047] Table 1

[0048] Table 2 exemplifies feasible calculation methods for weight coefficients of appearance indicators and chemical composition indicators (the calculation method is only an illustration, not the final weight coefficient value, and the specific method needs to be selected according to the characteristics of the sample material).

[0049] Table 2

[0050] Example 4 Evaluation test of flue-cured tobacco leaf quality based on the tobacco leaf quality evaluation method and device of the present invention: Reference Figure 7 As shown, four tobacco leaves were numbered and visually analyzed. The results showed that the color of the three channels of (RGB) values ​​of the four tobacco leaves were all in the lemon yellow group range, the maturity was mature, and the specifications were as follows: Among them, the No. 1 tobacco leaves are intact and are of B2F grade; the No. 2 tobacco leaves have a smaller damaged area and the damaged position is at the edge of the leaf tip, with a specification of B3F grade, which meets the grade requirements; the No. 3 tobacco leaves have a smaller damaged area and the damaged position is at the edge of the leaf tip, with a specification of C2F grade, which meets the grade requirements; the No. 4 tobacco leaves have a larger damaged area and the damaged position is at the edge of the leaf tip and the center of the upper part of the leaf, with a specification of C3F grade. As it does not meet the grade requirements, it is downgraded to C4F grade.

[0051] After visual judgment, the order of the four tobacco leaves was shuffled, and relevant grading experts were hired to conduct blind inspection. After manual judgment, the evaluation grade was consistent with the result obtained by this method.

[0052] After completing the visual judgment, the thickness and tensile strength of the sample tobacco leaves are shown in Table 3.

[0053] Table 3 serial number Tensile strength (N) Organizational structure evaluation Organizational structure score Thickness (μm) Identity evaluation Identity score 1 1.63 Slightly dense 3 125 Slightly thick 3 2 1.14 Still loose 7 94 Slightly thin 3 3 1.4 Moderate 5 94 Slightly thin 3 4 1.35 Moderate 6 107 Moderate 5 After obtaining the tensile strength and thickness data of tobacco leaves of each number, the organizational structure and identity of tobacco leaves of each number are evaluated based on the relative scores of the grades, and scores are given based on the actual measured values.

[0054] As shown in Table 4, the scores of different chemical indicators of tobacco leaves with different numbers are as follows based on the content of components.

[0055] Table 4

[0056] According to the weight coefficient calculation method described in Example 3, the characteristics of the tobacco leaves to be evaluated, such as the origin, part, variety and formulation requirements, are comprehensively considered to obtain a weight model suitable for tobacco leaves with different characteristics. The scores of the quality indicators such as the appearance and chemistry of the tobacco leaves are substituted into the corresponding indicator weight model for assignment and calculation. The evaluation scores of the tobacco leaves of each number are: No. 1: 76.5, No. 2: 70.7, No. 3: 88.4, No. 4: 82.6.

[0057] By comparing the quality scores of No. 1 and No. 2, and No. 3 and No. 4, it can be seen that the quality scores obtained by this method are consistent with the differences in tobacco leaves of different grades at the same part. By comparing the differences in appearance and chemical indicators between No. 1 and No. 3, and No. 2 and No. 4, it can be seen that the differences in various quality indicators measured by this method are consistent with the performance of tobacco leaves of different parts at the same grade.

[0058] This test example further verified that this method and device evaluated the quality of tobacco leaves by utilizing the weights of different indicators, thus solving the problem that the identity and organizational structure of tobacco leaves could not be measured based on visual methods alone. The results were accurate and the construction difficulty was relatively low, thus filling a gap in the research field of tobacco quality evaluation.

[0059] It should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A flue-cured tobacco leaf quality evaluation method based on machine vision and near infrared, characterized in that: include Analysis of Appearance Indexes of S1 Tobacco Leaves The tobacco leaves are moved to the label holder along with the tobacco leaf conveyor belt, and the main vein base of the tobacco leaves is clamped to fix it. The mechanical arm in the physical area of ​​the appearance clamps the label in the label placement box, fixes the label to the base of the tobacco leaves through the label codec and swipes the serial number. After the label is fixed, the label is loosened to make the tobacco leaves fall off. The tobacco leaves continue to move along the tobacco leaf conveyor belt, and are moved to the center of the top of the photography table in the order of stacking. The tobacco leaves are spread out flat, and images of the tobacco leaves are collected. The image features of the tobacco leaves are extracted through the convolutional neural network model, and various appearance indicators of the tobacco leaves are analyzed. Obtain the characteristic type, location and area data of tobacco leaves, and train the characteristic images obtained in this evaluation; Determination of physical indexes of S2 tobacco leaves The tobacco strip cutting knife and the tobacco circular punching knife are moved along the cutting knife moving track to the location of the tobacco leaves, and the cutting knife part is extended to the surface of the tobacco leaves to cut the tobacco leaves into tobacco strips with a fixed width and a number of tobacco sheets with the same diameter; The cut tobacco rod is placed on the tobacco rod tension measuring device by the mechanical arm in the appearance physical area. The two fixing clamps of the tobacco rod tension measuring device clamp the two sides of the tobacco rod respectively. With the help of the mechanical arm, the tobacco rod is slightly straightened. Then, the two fixing clamps are spread out to both sides until the tobacco rod breaks. The elongation and tension data of the tobacco rod are recorded at this time. The cut tobacco slices are placed on the tobacco slice thickness measuring device by the mechanical arm in the appearance physical area. The hydraulic telescopic rod of the tobacco slice thickness measuring device is lowered to contact the tobacco leaves, and the thickness of the tobacco slices at this time is recorded. The pressure is continuously increased, and the thickness change of the tobacco slices with the change of pressure is recorded. Measurement of chemical composition of S3 tobacco leaves The tobacco leaves are moved to the chemical composition measurement area along the tobacco leaf conveyor belt. The mechanical arm in the chemical composition area moves the tobacco leaves and places them on the top of the tobacco leaf placement table. The near-infrared transmitter moves along its moving track to the measurement points such as the two sides of the main vein in the leaf, the leaf tip, the leaf base, and the leaf edges on both sides. After reaching each point, the near-infrared transmitter descends to contact the surface of the tobacco leaves, and then emits near-infrared rays to the tobacco leaves through the near-infrared transmitting module, and receives the reflected rays with the near-infrared receiving module, and transmits the received data to the near-infrared device. Then the near-infrared transmitter rises and moves to the next point. The mechanical arm in the chemical composition area moves the tobacco leaves back to the tobacco leaf conveyor belt, and the near-infrared device reads the reflected rays of different wavelengths to obtain the main chemical composition content of the tobacco leaves; Quality evaluation of S4 tobacco leaves After obtaining the appearance indicators, physical indicators, and chemical composition data of the tobacco leaves, the scores of each indicator of the tobacco leaves are obtained according to the grading standards of each indicator, and the obtained scores are substituted into the indicator weight model to calculate the quality evaluation score of the tobacco leaves.

2. The method for evaluating tobacco leaf quality according to claim 1, characterized in that: The evaluation method also includes S5 Alternative Calculation After completing the quality evaluation of tobacco leaves, the industrial feasibility of the different tobacco leaves measured this time will be evaluated and cluster analysis will be performed to show the strength of substitutability between tobacco leaves of different grades.

3. The method for evaluating tobacco leaf quality according to claim 1, characterized in that: The evaluation method also includes S6 tobacco leaf grade classification After completing the quality evaluation of the tobacco leaves, the tobacco leaves are graded according to the quality evaluation scores. The tobacco leaves are clamped by the grading area robot arm and moved along the grading area moving track to the corresponding grading area for stacking.

4. The method for evaluating tobacco leaf quality according to claim 1, characterized in that: The S1 includes After acquiring the tobacco leaf image, S11 identifies the outline of the tobacco leaf, removes the background, and confirms the position of the characteristic part according to the outline shape and color distribution information of the tobacco leaf; S12 compares the features in the convolutional neural network feature library and classifies the features into types of damage or maturity such as variegated, damaged, and diseased; S13 corrodes the image of the characteristic part, selects the characteristic area and calculates the number of pixels, and further obtains the area of ​​the characteristic area: Feature area = number of pixels in the feature area * image shooting range area / total number of pixels in the image; S14 Different characteristic types have different degrees of influence on tobacco leaf quality, and the same characteristic types have different effects in different parts of the tobacco leaves. Therefore, the location and area of ​​the characteristic area are used to evaluate the appearance index.

5. A flue-cured tobacco leaf quality evaluation device based on machine vision and near infrared, characterized in that: Including the main structure and the The tobacco transport area includes a tobacco conveyor belt, one side of which is equipped with a label holder, the label holder is provided with a movable track, a main vein fixing clamp and a label codec are installed therein, the main vein fixing clamp can move horizontally in the movable track, the label codec can move horizontally and vertically in the movable track, and the middle and downstream of the tobacco conveyor belt are respectively equipped with an appearance physical area robot arm and a chemical composition area robot arm; The appearance and physical measurement area includes a photographic table, one side of which is equipped with an industrial camera, and the other side is equipped with a cutter moving track, in which two independent transmission chains are arranged, and the two transmission chains are respectively connected to the tobacco leaf strip cutting knife and the tobacco leaf circular punching knife, and the appearance and physical measurement area is also equipped with a tobacco leaf thickness measuring device and a tobacco rod tension measuring device; The chemical composition measurement area includes a tobacco leaf placement platform, two movable tracks are installed in parallel on both sides of the tobacco leaf placement platform, a metal rod body and a built-in bridge-type movable track are spanned between the movable tracks, and the three movable tracks together constitute a near-infrared movable track, a near-infrared emitter is installed in the bridge-type movable track, and a near-infrared device is installed in the vacant area of ​​the chemical composition measurement area as a reflected light different band information analysis device; The grading area includes a grading area moving track, which is perpendicular to the tobacco conveyor belt. A grading area robot arm is installed in the grading area moving track. A plurality of grading boxes are arranged on one side of the grading area moving track, and a tobacco storage box is installed next to each grading box.

6. The tobacco leaf quality evaluation device according to claim 5, characterized in that: The main vein fixing clamp structure is a mechanical clamp to clamp the tobacco stem on the side. The tag codec has built-in NFC and an electromagnet for data import and tag codec switch.

7. The tobacco leaf quality evaluation device according to claim 5, characterized in that: The mechanical arm in the appearance and physical area has the same structure as the mechanical arm in the chemical composition area, and both have a rotating base connected to the top of the main structure, which can drive the mechanical arm to rotate. A multi-joint movable arm is installed on the rotating base, and a mechanical gripper is installed on the top of the movable arm. Its movable range covers the adjacent sections of the tobacco conveyor belt, and covers all positions of the appearance and physical measurement area and the chemical composition measurement area respectively.

8. The tobacco leaf quality evaluation device according to claim 5, characterized in that: The tobacco strip cutting knife is a circular roller with a certain spacing, which is installed together on the same roller. The roller is connected to a horizontal telescopic rod, and the horizontal telescopic rod is installed on a movable base, and the movable base is connected to the inside of the cutting knife movable track; the tobacco circular punching knife is a circular ring-shaped twisting knife, which is installed on a horizontal telescopic rod, and the horizontal telescopic rod is installed on a movable base, and the movable base is also connected to the inside of the cutting knife movable track.

9. The tobacco leaf quality evaluation device according to claim 5, characterized in that: The tobacco leaf thickness measuring device comprises a cylindrical base, a measuring rod is installed on one side of the cylindrical base, a hydraulic column with a downward direction and adjustable pressure is provided on the top of the measuring rod, which is directly opposite to the center of the cylindrical base and can record the moving distance during the extension and retraction process; the structure of the tobacco rod tension measuring device is a raised prism, a moving track is installed on one side of the prism, and two mechanical clamps with the same structure are installed in the moving track, which can open and merge along the track and have the functions of force measurement and distance measurement.

10. The tobacco leaf quality evaluation device according to claim 5, characterized in that: The near-infrared transmitter is a probe composed of a near-infrared transmitter and a receiver, and has a vertical lifting rod at the top, which is connected to the bridge-type moving track and can drive the near-infrared transmitter to perform lifting movement in the vertical direction.

Images