Tobacco leaf sorting quality intelligent detection device and detection method
By combining the feeding and conveying module, the multimodal vision acquisition component, and the chemical composition detection component, the problems of low efficiency and poor consistency caused by manual reliance in existing tobacco leaf sorting are solved, and automated and accurate sorting and quality traceability of tobacco leaves are realized.
Patent Information
- Application Number
- CN202511155864.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-18
- Publication Date
- 2025-11-14
AI Technical Summary
The existing tobacco leaf sorting process relies heavily on manual labor, resulting in low sorting efficiency, difficulty in ensuring consistency, and difficulty in obtaining objective and quantitative quality data, making it impossible to achieve accurate grading based on intrinsic quality.
The system employs a feeding and conveying module, a multimodal vision acquisition component, and a chemical composition detection component, combined with neural network algorithms and a self-learning model, to achieve automated sorting of tobacco leaves. The feeding and conveying module ensures stable transport of tobacco leaves through tobacco frame racks and flattening rollers. The multimodal vision acquisition component acquires multi-dimensional appearance features, the chemical composition detection component analyzes internal components in real time, and the sorting module quickly sorts the leaves to the corresponding grade based on the detection results.
It has achieved automation and precise grading of tobacco leaf sorting, improved sorting efficiency and consistency, avoided the inefficiency and errors of manual sorting, and provided a basis for quality traceability and production optimization.
Smart Images
Figure CN120940252A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of tobacco leaf sorting, and in particular to an intelligent detection device and method for tobacco leaf sorting quality. Background Technology
[0002] Tobacco leaf sorting is a key pre-processing step in the tobacco processing industry. It is usually carried out after the tobacco leaves are purchased. The sorting process involves systematically grading the appearance quality (such as part, color, and maturity) and internal quality of the tobacco leaves according to national standards. Different quality tobacco leaves are divided into several grades, such as the main group (normal tobacco leaves) and the secondary group (low-quality tobacco leaves). The sorting process includes three stages: initial sorting, re-sorting, and final sorting.
[0003] In the current tobacco processing sector, particularly in the tobacco leaf sorting stage, China generally employs a traditional process heavily reliant on manual labor. The core of this model involves a large number of leaf sorters manually sorting the tobacco leaves based on experience, followed by random quality checks by quality inspectors. This manual-intensive approach permeates the entire process of sorting, inspection, and recording. However, because the entire process relies excessively on subjective judgment and physical labor, sorting efficiency and consistency are difficult to guarantee, and objective, quantifiable quality data is hard to obtain.
[0004] Meanwhile, manual leaf-by-leaf inspection is inefficient and extremely labor-intensive, with prolonged work easily leading to worker fatigue and consequently causing missed inspections or misclassifications. Existing processes struggle to perform online analysis of tobacco leaf parameters such as sugar-alkali ratio and total nitrogen, making accurate grading based on intrinsic quality impossible. Summary of the Invention
[0005] To reduce the manual labor intensity of quality inspectors and improve the consistency and reliability of testing, this application provides an intelligent detection device and method for tobacco leaf sorting quality.
[0006] This application provides an intelligent detection device and method for tobacco leaf sorting quality, which adopts the following technical solution:
[0007] A smart detection device and method for tobacco leaf sorting quality, comprising:
[0008] The feeding and conveying module includes a tobacco frame rack, a feeding belt, and a flattening roller. The feeding and conveying module is used for the stable conveying and flattening of tobacco leaves.
[0009] The detection module includes a multimodal visual acquisition component and a chemical composition detection component, which is used to acquire the appearance characteristics and internal composition data of tobacco leaves;
[0010] The sorting module sorts tobacco leaves to the corresponding grade collection area based on the detection results obtained from the detection module;
[0011] A control system for controlling and / or adjusting the operation of the feeding and conveying module, the detection module, and the sorting module.
[0012] By adopting the above technical solutions, the tobacco frame rack in the feeding and conveying module provides a stable temporary storage space for the tobacco leaves to be tested. The feeding belt, in conjunction with the flattening roller, ensures the smooth and continuous conveying and uniform spreading of the tobacco leaves, effectively avoiding blind spots in the detection caused by wrinkles and stacking of the tobacco leaves. This provides the subsequent detection module with fully unfolded and unobstructed samples for testing. The multimodal vision acquisition component in the detection module comprehensively covers the multi-dimensional appearance characteristics of the tobacco leaves, such as color, texture, thickness, and surface micro-details, through a combination of reflected light, transmitted light, and local high-definition imaging. At the same time, the chemical composition detection component analyzes the internal component data such as nicotine and total sugar in real time. Based on the accurate data output by the detection module, the sorting module quickly and accurately sorts the tobacco leaves to the corresponding grade collection area, avoiding the inefficiency and errors of manual sorting. The control system coordinates the operation rhythm of the feeding, conveying, detection, and sorting modules to ensure the automated control of the entire process.
[0013] Preferably, the multimodal vision acquisition component includes at least one front-facing camera and one back-facing camera, and the multimodal vision acquisition component is used to perform reflected light imaging, reflected local high-definition imaging, and transmitted light imaging on tobacco leaves.
[0014] By employing the above technical solutions, the front camera collects reflected light images from above the tobacco leaf, acquiring macroscopic surface features such as the overall color and texture of the leaf. The back camera, in conjunction with transmitted light imaging from below, simulates thickness characteristics perceived by human touch (such as leaf looseness and tissue compactness) through the projection information after light penetrates the tobacco leaf. The high-resolution local reflection imaging, through the high-magnification local focusing of the front camera, magnifies details of microscopic areas such as tobacco leaf wrinkles and spots, clearly presenting features that are difficult to observe with the naked eye, such as cell structure and leaf vein distribution. The transmitted light imaging, in conjunction with the back camera, supplements implicit features such as the uniformity of internal tissue and water content of the tobacco leaf through the difference in light transmittance.
[0015] Preferably, in the reflected light imaging, a first reflected camera and a first LED linear array light source are used. The first reflected camera and the first LED linear array light source are located on the same side of the sorting and conveying line. The reflected light imaging also includes a first light source, which is a light source with multiple angles and multiple focal planes.
[0016] By adopting the above technical solution, the first reflection camera and the first LED linear array light source are located on the same side of the sorting and conveying line, so that the light emitted by the light source enters the camera directly after being reflected by the surface of the tobacco leaf, shortening the optical path, reducing ambient light interference, and improving the acquisition efficiency of reflected light signals. The multi-angle and multi-focus surface combined light source integrates sub-light source modules with different illumination angles and different focusing depths, and adjusts the incident angle and focusing position of the light for local surfaces at different heights of the tobacco leaf, ensuring that all areas of the tobacco leaf can be uniformly illuminated, avoiding the local shadows or overexposure problems caused by the undulations of the tobacco leaf surface due to traditional single-angle light sources. At the same time, the multi-focus surface design of the combined light source can dynamically match the optimal focusing depth of different areas according to the deformation of the tobacco leaf caused by speed changes during the conveying process, so that the camera can clearly capture the texture, veins, damage and other detailed features of the tobacco leaf.
[0017] Preferably, in the reflective local high-definition imaging, a second reflective camera and a second LED linear array light source are also used. The second reflective camera and the second LED linear array light source are located on the same side of the sorting conveyor line, and the magnification of the second reflective camera is greater than 0.2.
[0018] By adopting the above technical solution, the second reflection camera and the second LED linear array light source are located on the same side of the sorting and conveying line, so that the light from the light source enters the camera directly after being reflected from the surface of the tobacco leaf, shortening the optical path and reducing ambient light scattering interference, thereby ensuring the signal-to-noise ratio of the imaging in the local area; the magnification of the second reflection camera is greater than 0.2, which can capture close-up images of local areas that are easily blurred in global imaging, such as the wrinkles and depressions of the tobacco leaf, the branching points of the leaf veins, and minor damage, and clearly observe the microscopic features of the tobacco leaf; combined with the directional supplementary lighting of the second LED linear array light source, the contrast of the local area is further enhanced, avoiding shadow occlusion caused by wrinkles, so that the camera can accurately capture the texture direction and cell arrangement uniformity of the tobacco leaf surface.
[0019] Preferably, in the transmitted light imaging, a first transmission camera and a first water-cooled LED linear array light source are used. The first water-cooled LED linear array light source is located below the sorting conveyor line, and the first transmission camera is located above the sorting conveyor line.
[0020] By adopting the above technical solution, the first water-cooled LED linear array light source is located below the sorting conveyor line. The circulating cooling system avoids light intensity attenuation or fluctuation caused by heat generation during long-term operation, ensuring the stability of transmitted light intensity. The light source and camera are placed on both sides of the conveyor line, so that the light penetrates the tobacco leaf in a direction perpendicular to the conveyor line, avoiding light scattering or perspective distortion caused by oblique angle incidence. The weaker the transmitted light intensity, the thicker the tobacco leaf; the stronger the transmitted light intensity, the thinner the tobacco leaf. The first linear array light source covers the entire surface of the tobacco leaf, avoiding local missed detection due to incomplete light source coverage.
[0021] Preferably, there are two sets of the first water-cooled LED linear array light source, the two sets of the first water-cooled LED linear array light source are symmetrically installed, and the focusing lines of the two sets of the first water-cooled LED linear array light source coincide.
[0022] By adopting the above technical solution, two sets of symmetrically installed water-cooled LED linear array light sources emit light synchronously from symmetrical positions below the sorting conveyor line, avoiding insufficient or overexposed lighting in local areas of the tobacco leaves due to angular deviation of a single light source; the overlapping focal lines of the two light sources further enhance the concentration of light, improving the penetration efficiency of transmitted light by converging the dispersed light energy onto the same vertical path, avoiding the attenuation of transmitted light intensity due to light divergence, thereby clearly capturing the implicit characteristics of the internal fiber structure and moisture content distribution of the tobacco leaves.
[0023] Preferably, the sorting module includes a high-speed pneumatic jetting mechanism or a transverse lever-type sorting mechanism, as well as multi-level discharge bins set according to grade.
[0024] By adopting the above technical solutions, the high-speed pneumatic blowing mechanism precisely pushes the tobacco leaves to the corresponding dropping bin with high-speed airflow, avoiding damage to the tobacco leaves due to contact and achieving rapid sorting; the horizontal lever-type sorting mechanism pushes the tobacco leaves to the target area through an electric lever with adjustable stroke; the multi-level dropping bins set according to grade correspond one-to-one with the sorting mechanism, and each bin is equipped with a weighing sensor and a full bin alarm device at the bottom, which not only achieves physical isolation of tobacco leaves of different grades, but also triggers a bin-changing prompt when the weight of tobacco leaves in the bin is reached in real time, ensuring the continuity of the sorting process.
[0025] A method for intelligent detection of tobacco leaf sorting quality includes the following steps:
[0026] S1. The tobacco leaf samples are manually placed on the feed inlet platform of the feeding and conveying module, so that the tobacco leaves are laid flat and transported one by one to the testing area;
[0027] S2. The multimodal vision acquisition component acquires images of tobacco leaves;
[0028] S3. The chemical composition detection component simultaneously collects chemical composition data of tobacco leaves;
[0029] S4. The data processing system uses a neural network algorithm to fuse multimodal image information and chemical composition data, and combines it with a self-learning model to grade tobacco leaves;
[0030] S5. The sorting module sorts the tobacco leaves to the corresponding grade multi-level discharge bins by high-speed pneumatic blowing or horizontal lever according to the grading results.
[0031] S6. The data management system of the control system organizes the test data and generates quality inspection reports.
[0032] By adopting the above technical solution, tobacco leaves are manually laid out and transported one by one to the testing area. The stable transmission of the feeding and conveying module avoids the stacking or wrinkling of tobacco leaves. The multimodal vision acquisition component simultaneously acquires panoramic images of reflected light, transmitted light thickness images, and local high-definition detail images of tobacco leaves, comprehensively covering multi-dimensional information on the appearance of tobacco leaves. The simultaneous acquisition of chemical composition data supplements the key indicators of the internal quality of tobacco leaves. The neural network algorithm performs fusion analysis of multimodal images and chemical composition data, and dynamically optimizes feature weights through a self-learning model to improve the accuracy of grading. The high-speed pneumatic blowing or lateral lever sorting mechanism realizes rapid sorting of tobacco leaves based on accurate grading results. The physical isolation design of the multi-level feeding bins avoids mixing of grades and ensures the purity of tobacco leaves of each grade after sorting. The data management system organizes the test data and generates reports, providing a basis for quality traceability and production optimization.
[0033] Preferably, in step S2, the color, texture, and detailed features of both sides of the tobacco leaf need to be collected.
[0034] By adopting the above technical solution, key information such as color uniformity and texture direction of the tobacco leaves can be obtained simultaneously through image acquisition from both sides, avoiding misjudgment of grade due to incomplete information from one side; for the acquisition of texture and detail features, dual-side acquisition combined with multimodal imaging further eliminates blind spots.
[0035] Preferably, in step S6, after generating the quality inspection report, the control system optimizes the grading model through a self-learning algorithm.
[0036] By adopting the above technical solution, the quality inspection report records the grading results and test data. For newly emerging tobacco varieties or special grades, if their appearance / chemical characteristics are not covered by the initial model, the algorithm can extract their unique features through cluster analysis and incorporate them into the model's discriminative feature library, thus expanding the model's applicability.
[0037] In summary, this application includes at least one of the following beneficial technical effects:
[0038] 1. The tobacco frame rack in the feeding and conveying module provides a stable temporary storage space for the tobacco leaves to be tested. The feeding belt, in conjunction with the flattening roller, ensures smooth and continuous conveying and uniform spreading of the tobacco leaves, effectively avoiding blind spots in the detection caused by wrinkles and stacking. This provides fully unfolded and unobstructed samples for subsequent detection modules. The multimodal vision acquisition component in the detection module comprehensively covers the color, texture, thickness, and surface micro-details of the tobacco leaves through a combination of reflected light, transmitted light, and local high-definition imaging. At the same time, the chemical composition detection component (such as a near-infrared NIR detector) analyzes the intrinsic component data such as nicotine and total sugar in real time. Based on the accurate data output by the detection module, the sorting module quickly and accurately sorts the tobacco leaves to the corresponding grade collection area, avoiding the inefficiency and errors of manual sorting. The control system coordinates the operation rhythm of the feeding, conveying, detection, and sorting modules to ensure automated control of the entire process.
[0039] 2. The front camera captures reflected light images from above the tobacco leaf, obtaining macroscopic surface features such as the overall color and texture of the leaf. The back camera, in conjunction with transmitted light imaging from below, simulates thickness characteristics perceived by touch (such as leaf looseness and tissue compactness) through the projection information after light passes through the tobacco leaf. High-definition local reflection imaging, through high-magnification local focusing (magnification > 0.2) of the front camera, magnifies details of microscopic areas such as wrinkles and spots in the tobacco leaf, clearly presenting features that are difficult to observe with the naked eye, such as cell structure and leaf vein distribution. Transmitted light imaging, in conjunction with the back camera, supplements implicit features such as the uniformity of internal tissue and water content of the tobacco leaf by using differences in light transmittance.
[0040] 3. Tobacco leaves are manually laid out and transported one by one to the testing area. The stable transmission of the feeding and conveying module avoids the stacking or wrinkling of the tobacco leaves. The multimodal vision acquisition component simultaneously acquires panoramic images of reflected light, transmitted light thickness images, and local high-definition detail images of the tobacco leaves, comprehensively covering multi-dimensional information on the appearance of the tobacco leaves. The simultaneous acquisition of chemical composition data supplements the key indicators of the internal quality of the tobacco leaves. The neural network algorithm performs fusion analysis of multimodal images and chemical composition data, and dynamically optimizes feature weights through a self-learning model to improve the accuracy of grading. The high-speed pneumatic blowing or lateral lever sorting mechanism achieves rapid sorting of tobacco leaves based on the accurate grading results. The physical isolation design of the multi-stage material hopper avoids mixing of grades and ensures the purity of tobacco leaves of each grade after sorting. The data management system organizes the test data and generates reports, providing a basis for quality traceability and production optimization. Attached Figure Description
[0041] Figure 1 This is a front view of the intelligent detection device for tobacco leaf sorting quality in the embodiments of this application;
[0042] Figure 2 This is a top view showing the structure of the intelligent detection device for tobacco leaf sorting quality;
[0043] Figure 3 This is a flowchart illustrating an intelligent detection method for tobacco leaf sorting quality.
[0044] Explanation of reference numerals in the attached drawings: 1. Feeding and conveying module; 11. Smoke frame storage rack; 12. Feeding belt; 13. Flattening roller; 2. Detection module; 21. Multimodal vision acquisition component; 211. First reflection camera; 212. First LED linear array light source; 213. First light source; 214. Second reflection camera; 215. Second LED linear array light source; 216. First transmission camera; 217. First water-cooled LED linear array light source; 22. Chemical composition detection component; 3. Sorting module; 31. High-speed pneumatic jetting mechanism; 32. Material discharge bin; 4. Control system. Detailed Implementation
[0045] The following is in conjunction with the appendix Figure 1-3 This application will be described in further detail.
[0046] This application discloses an intelligent detection device and method for tobacco leaf sorting quality. (Refer to...) Figure 1 and Figure 2 The intelligent detection device for tobacco leaf sorting quality includes a feeding and conveying module 1, a detection module 2, a sorting module 3, and a control system 4. The feeding and conveying module 1 includes a tobacco frame rack 11, a feeding belt 12, and a flattening roller 13. The feeding and conveying module 1 is used for the stable conveying and flattening of tobacco leaves. The tobacco frame rack 11 in the feeding and conveying module 1 provides a stable temporary storage space for the tobacco leaves to be detected. The feeding belt 12, together with the flattening roller 13, realizes the stable and continuous conveying and uniform flattening of tobacco leaves, effectively avoiding detection blind spots caused by wrinkles and stacking of tobacco leaves.
[0047] The detection module 2 includes a multimodal visual acquisition component 21 and a chemical composition detection component 22. The detection module 2 is used to acquire the appearance characteristics and internal composition data of tobacco leaves. The multimodal visual acquisition component 21 in the detection module 2 comprehensively covers the multi-dimensional appearance characteristics of tobacco leaves, such as color, texture, thickness, and surface micro-details, through a combination of reflected light, transmitted light, and local high-definition imaging. Simultaneously, the chemical composition detection component 22 analyzes internal composition data such as nicotine and total sugar in real time. In an optional embodiment, the chemical composition detection component 22 is a near-infrared (NIR) detector.
[0048] In an optional embodiment, the multimodal vision acquisition component 21 includes at least one front-facing camera and one back-facing camera. The multimodal vision acquisition component 21 is used to perform reflected light imaging, reflected local high-definition imaging, and transmitted light imaging on tobacco leaves. The front-facing camera acquires reflected light images from above the tobacco leaves, obtaining macroscopic surface features such as the overall color and texture of the leaves. The back-facing camera, working in conjunction with transmitted light imaging from below, simulates the thickness characteristics perceived by human touch through the projection information after light penetrates the tobacco leaves. In an optional embodiment, the thickness characteristics include leaf looseness and tissue compactness. Reflected local high-definition imaging uses high-magnification local focusing by the front-facing camera, where the magnification is >0.2 to achieve a microscope effect, magnifying details in microscopic areas such as wrinkles and spots on the tobacco leaves, clearly presenting features difficult to observe with the naked eye, such as cell structure and leaf vein distribution. Transmitted light imaging, in conjunction with the back-facing camera, supplements implicit features such as the uniformity of internal tissue and water content of the tobacco leaves through differences in light transmittance.
[0049] In a preferred embodiment, the reflected light imaging employs a first reflecting camera 211 and a first LED linear array light source 212, both located on the same side of the sorting conveyor line. The reflected light imaging also includes a first light source 213, which is a multi-angle, multi-focal surface combination light source. The fact that the first reflecting camera 211 and the first LED linear array light source 212 are located on the same side of the sorting conveyor line allows the light emitted by the light source to directly enter the camera after reflection from the tobacco leaf surface, shortening the optical path, reducing ambient light interference, and improving the efficiency of reflected light signal acquisition.
[0050] The multi-angle, multi-focus surface combined light source integrates sub-light source modules with different illumination angles and focusing depths. It adjusts the incident angle and focusing position of light for local surfaces at different heights of the tobacco leaf, ensuring that all areas of the tobacco leaf are evenly illuminated. This avoids the local shadows or overexposure problems caused by the uneven surface of the tobacco leaf due to traditional single-angle light sources. At the same time, the multi-focus surface design of the combined light source can dynamically match the optimal focusing depth for different areas according to the deformation of the tobacco leaf caused by changes in speed during transportation. This allows the camera to clearly capture the texture, veins, damage and other detailed features of the tobacco leaf.
[0051] In an optional embodiment, the high-definition imaging of the reflected area also employs a second reflective camera 214 and a second LED linear array light source 215. The second reflective camera 214 and the second LED linear array light source 215 are located on the same side of the sorting conveyor line, and the magnification of the second reflective camera 214 is greater than 0.2. The fact that the second reflective camera 214 and the second LED linear array light source 215 are located on the same side of the sorting conveyor line allows the light from the light source to directly enter the camera after being reflected from the tobacco leaf surface, shortening the optical path and reducing ambient light scattering interference, thereby ensuring the signal-to-noise ratio of the local area. The magnification of the second reflective camera 214, greater than 0.2, allows for close-up acquisition of local areas that are easily blurred in global imaging, such as tobacco leaf wrinkles, vein branching points, and minor damage, clearly observing the microscopic features of the tobacco leaf. Combined with the directional supplementary lighting of the second LED linear array light source 215, the contrast of the local area is further enhanced, avoiding shadow occlusion caused by wrinkles, enabling the camera to accurately capture the texture direction and cell uniformity of the tobacco leaf surface.
[0052] In an optional embodiment, the transmitted light imaging employs a first transmitted light camera 216 and a first water-cooled LED linear array light source 217. The first water-cooled LED linear array light source 217 is located below the sorting conveyor line, while the first transmitted light camera 216 is located above the sorting conveyor line. The first water-cooled LED linear array light source 217, located below the sorting conveyor line, utilizes a circulating cooling system to prevent light intensity attenuation or fluctuations caused by prolonged operation and heat generation, ensuring the stability of the transmitted light intensity. The light source and camera are positioned on opposite sides of the conveyor line, allowing light to penetrate the tobacco leaves perpendicular to the conveyor line, avoiding light scattering or perspective distortion caused by oblique angle incidence. Weaker transmitted light intensity indicates thicker tobacco leaves, while stronger transmitted light intensity indicates thinner tobacco leaves. The first linear array light source covers the entire surface of the tobacco leaves, preventing localized missed detections due to incomplete light source coverage.
[0053] In a preferred embodiment, two sets of first water-cooled LED linear array light sources 217 are provided, and the two sets of first water-cooled LED linear array light sources 217 are symmetrically installed, with their focal lines coinciding. The two symmetrically installed sets of water-cooled LED linear array light sources emit light synchronously from symmetrical positions below the sorting conveyor line, avoiding insufficient or overexposed lighting in local areas of the tobacco leaves due to angular deviation of a single light source; the design of the overlapping focal lines of the two light sources further enhances the concentration of light, improving the penetration efficiency of transmitted light by converging the dispersed light energy onto the same vertical path, avoiding the attenuation of transmitted light intensity due to light divergence, thereby clearly capturing the implicit characteristics of the internal fiber structure and moisture content distribution of the tobacco leaves.
[0054] The sorting module 3 sorts the tobacco leaves to the corresponding grade collection area based on the detection results obtained from the detection module 2. In an optional embodiment, the sorting module 3 includes a high-speed pneumatic blowing mechanism 31 or a transverse lever-type sorting mechanism, and multi-level discharge bins 32 arranged according to grade. The high-speed pneumatic blowing mechanism 31 uses high-speed airflow to precisely push the tobacco leaves to the corresponding discharge bins 32, avoiding damage to the tobacco leaves due to contact and achieving rapid sorting; the transverse lever-type sorting mechanism pushes the tobacco leaves to the target area through an electric lever with adjustable stroke; the multi-level discharge bins 32 arranged according to grade correspond one-to-one with the sorting mechanism, and each bin is equipped with a weighing sensor and a full bin alarm device at the bottom, which not only achieves physical isolation of tobacco leaves of different grades, but also triggers a bin-changing prompt when the weight of the tobacco leaves in the bin is reached in real time, ensuring the continuity of the sorting process. The control system 4 is used to control and / or adjust the operation of the feeding and conveying module 1, the detection module 2 and the sorting module 3. By coordinating the operating rhythm of each module, the feeding, conveying, detection and sorting are ensured to achieve automated control of the entire process.
[0055] Reference Figure 3 The intelligent detection method for tobacco leaf sorting quality includes the following steps:
[0056] S1. The tobacco leaf samples are manually placed on the feed inlet platform of the feeding and conveying module 1, so that the tobacco leaves are laid flat and transported one by one to the testing area;
[0057] S2. The multimodal vision acquisition component 21 acquires images of tobacco leaves;
[0058] S3. The chemical composition detection component 22 simultaneously collects chemical composition data of tobacco leaves;
[0059] S4. The data processing system uses a neural network algorithm to fuse multimodal image information and chemical composition data, and combines it with a self-learning model to grade tobacco leaves;
[0060] S5. Sorting module 3 sorts tobacco leaves to the corresponding grade multi-level discharge bin 32 by high-speed pneumatic blowing or horizontal lever according to the grading results.
[0061] S6. The data management system of control system 4 organizes the test data and generates quality inspection reports.
[0062] The tobacco leaves are manually laid out and transported one by one to the testing area. The stable transmission of the feeding and conveying module 1 avoids the stacking or wrinkling of the tobacco leaves. The multimodal vision acquisition component 21 simultaneously acquires panoramic images of reflected light, transmitted light thickness images, and local high-definition detail images of the tobacco leaves, comprehensively covering multi-dimensional information on the appearance of the tobacco leaves. The simultaneous acquisition of chemical composition data supplements the key indicators of the internal quality of the tobacco leaves. The neural network algorithm performs fusion analysis of multimodal images and chemical composition data, and dynamically optimizes feature weights through a self-learning model to improve the accuracy of grading. The high-speed pneumatic blowing or lateral lever sorting mechanism realizes rapid sorting of tobacco leaves based on the accurate grading results. The physical isolation design of the multi-level feeding bin 32 avoids mixing of grades and ensures the purity of tobacco leaves of each grade after sorting. The data management system organizes the test data and generates reports, providing a basis for quality traceability and production optimization.
[0063] In step S2, the color, texture, and detail features of both sides of the tobacco leaves need to be collected. Through image acquisition from both sides, key information such as color uniformity and texture direction on both sides of the tobacco leaves is simultaneously obtained to avoid misclassification due to incomplete information from one side. For the acquisition of texture and detail features, dual-side acquisition combined with multimodal imaging further eliminates blind spots. In step S6, after generating the quality inspection report, the control system 4 optimizes the grading model through a self-learning algorithm. The quality inspection report records the grading results and test data. For newly emerging tobacco varieties or special grades, if their appearance / chemical characteristics are not covered by the initial model, the algorithm can extract their unique features through cluster analysis and incorporate them into the model's discriminative feature library, expanding the model's applicability.
[0064] The implementation principle of this application embodiment is as follows: The tobacco frame rack 11 in the feeding and conveying module 1 provides a stable temporary storage space for the tobacco leaves to be tested. The feeding belt 12, together with the flattening roller 13, realizes the smooth and continuous conveying and uniform spreading of the tobacco leaves, effectively avoiding the blind spots in the detection caused by the wrinkles and stacking of the tobacco leaves, and providing the subsequent detection module 2 with fully unfolded and unobstructed samples to be tested; The multimodal vision acquisition component 21 in the detection module 2 comprehensively covers the multi-dimensional appearance features of the tobacco leaves, such as color, texture, thickness, and surface micro-details, through the combination of reflected light, transmitted light and local high-definition imaging. At the same time, the chemical composition detection component 22 (such as a near-infrared NIR detector) analyzes the internal component data such as nicotine and total sugar in real time; Based on the accurate data output by the detection module 2, the sorting module 3 quickly and accurately sorts the tobacco leaves to the corresponding grade collection area, avoiding the inefficiency and error of manual sorting; The control system 4 ensures the automated control of the entire process by coordinating the operation rhythm of the feeding, conveying, detection and sorting modules.
[0065] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. An intelligent detection device for tobacco leaf sorting quality, characterized in that, include: The feeding and conveying module (1) includes a tobacco frame rack (11), a feeding belt (12) and a flattening roller (13). The feeding and conveying module (1) is used for the stable conveying and flattening of tobacco leaves. The detection module (2) includes a multimodal visual acquisition component (21) and a chemical composition detection component (22). The detection module (2) is used to acquire the appearance characteristics and internal composition data of tobacco leaves. The sorting module (3) sorts the tobacco leaves to the corresponding grade collection area according to the detection results obtained from the detection module (2); The control system (4) is used to control and / or regulate the operation of the feeding and conveying module (1), the detection module (2) and the sorting module (3).
2. The intelligent detection device for tobacco leaf sorting quality according to claim 1, characterized in that, The multimodal vision acquisition component (21) includes at least one front-facing camera and one back-facing camera. The multimodal vision acquisition component (21) is used to perform reflected light imaging, reflected local high-definition imaging, and transmitted light imaging on tobacco leaves.
3. The intelligent detection device for tobacco leaf sorting quality according to claim 2, characterized in that, In the reflected light imaging, a first reflected camera (211) and a first LED linear array light source (212) are used. The first reflected camera (211) and the first LED linear array light source (212) are located on the same side of the sorting and conveying line. The reflected light imaging also includes a first light source (213), which is a light source with multiple angles and multiple focal planes.
4. The intelligent detection device for tobacco leaf sorting quality according to claim 2, characterized in that, In the aforementioned high-definition local reflection imaging, a second reflection camera (214) and a second LED linear array light source (215) are also used. The second reflection camera (214) and the second LED linear array light source (215) are located on the same side of the sorting and conveying line, and the magnification of the second reflection camera (214) is greater than 0.
2.
5. The intelligent detection device for tobacco leaf sorting quality according to claim 2, characterized in that, In the transmitted light imaging, a first transmission camera (216) and a first water-cooled LED linear array light source (217) are used. The first water-cooled LED linear array light source (217) is located below the sorting conveyor line, and the first transmission camera (216) is located above the sorting conveyor line.
6. The intelligent detection device for tobacco leaf sorting quality according to claim 5, characterized in that, The first water-cooled LED linear array light source (217) is provided in two sets, and the two sets of the first water-cooled LED linear array light source (217) are symmetrically installed, and the focusing lines of the two sets of the first water-cooled LED linear array light source (217) coincide.
7. The intelligent detection device for tobacco leaf sorting quality according to claim 1, characterized in that, The sorting module (3) includes a high-speed pneumatic jetting mechanism (31) or a transverse lever sorting mechanism, as well as a multi-level discharge bin (32) set according to grade.
8. A method for intelligent detection of tobacco leaf sorting quality, comprising using the intelligent detection device for tobacco leaf sorting quality as described in any one of claims 1-7, characterized in that, Including the following steps: S1. Manually place the tobacco leaf samples on the feed inlet platform of the feeding and conveying module (1) so that the tobacco leaves are laid flat and transported one by one to the testing area; S2. The multimodal vision acquisition component (21) acquires images of tobacco leaves; S3. The chemical composition detection component (22) synchronously collects chemical composition data of tobacco leaves; S4. The data processing system uses a neural network algorithm to fuse multimodal image information and chemical composition data, and combines it with a self-learning model to grade tobacco leaves; S5. Sorting module (3) sorts tobacco leaves to the corresponding grade multi-level discharge bin (32) by high-speed pneumatic blowing or horizontal lever according to the grading results; S6. The data management system of the control system (4) organizes the test data and generates quality inspection reports.
9. The intelligent detection method for tobacco leaf sorting quality according to claim 8, characterized in that, In step S2, the color, texture, and detailed features of both sides of the tobacco leaves need to be collected.
10. The intelligent detection device for tobacco leaf sorting quality according to claim 8, characterized in that, In step S6, after generating the quality inspection report, the control system (4) optimizes the grading model through a self-learning algorithm.