Three-eyed vision coordinated tobacco length and volume direct detection device and method

Abstract

This invention belongs to the field of tobacco shred detection technology, specifically relating to a device and method for direct detection of tobacco shred length and volume using a three-view vision-coordinated system. The detection device includes a tobacco shred metering and discharging unit, a tobacco shred uniform conveying unit, a three-view vision acquisition unit, a cleaning unit, and a computer control and calculation output unit. These units are mechanically connected sequentially according to the tobacco shred conveying direction, and all electronic control components are electrically connected to the computer. This device features a three-view vision synchronous acquisition structure, with three sets of CCD cameras whose fields of view completely overlap. LED light sources and cameras are arranged coaxially, and the side CCD camera is tilted towards the center of the detection section, eliminating the inherent blind spots of single-view acquisition and solving the problem of incomplete contour extraction caused by bent or non-flat tobacco shreds. This lays the hardware foundation for multi-dimensional feature extraction and three-dimensional measurement of tobacco shreds.

Description

Technical Field

[0001] This invention belongs to the field of tobacco shred detection technology, specifically relating to a device and method for direct detection of tobacco shred length and volume using a three-eye vision-assisted approach. Background Technology

[0002] Tobacco shred size (including length, width, and particle size distribution) is a core characteristic parameter of cigarette quality, and its rationality directly determines the stability of cigarette combustion, smoke release, and smoking taste. Current mainstream methods for detecting tobacco shred size are mainly divided into two categories: traditional sieving methods and conventional machine vision inspection methods. Specific technical solutions are as follows:

[0003] Traditional sieving method: This method uses a vibrating sieve machine specifically designed for the tobacco industry. The equipment consists of a support frame, a vibration drive module, multi-layer grading screens, a material collection module, and a shock absorption system. During testing, tobacco leaves are placed in stacked screens with different aperture sizes. The vibration drive module causes the screens to vibrate, causing the tobacco leaves to be graded and retained according to size. The cumulative mass of tobacco leaves retained by each layer of screens is used as the ratio of the total mass to the total mass as a quantitative evaluation index of the tobacco leaf's size structure. Volume measurement can only be indirectly calculated using the "mass / density" formula, and there is no direct detection method.

[0004] Conventional machine vision inspection methods involve placing a single light source and industrial camera at a single angle directly above the tobacco conveyor belt. Basic algorithms such as binarization segmentation and line fitting are used to extract the outline of the tobacco shreds and calculate their length. The volume of the tobacco shreds is calculated using traditional density conversion methods, without direct representation through visual features. The equipment uses a simple feeding chute and non-directional collection and discharge. The conveyor belt speed is fixed, and there are no self-cleaning or anti-sticking structures. Furthermore, no frame rate or supplementary lighting optimizations have been implemented specifically for tobacco inspection scenarios.

[0005] Traditional sieving methods suffer from significant limitations due to their inherent technical design. They can only screen tobacco leaves by size, lacking any design for identifying leaf morphology. Therefore, they cannot capture morphological features such as leaf curvature and curl, resulting in a single detection dimension. Core detection steps in traditional sieving methods, such as feeding, weighing, and data calculation, all rely on manual operation. Individual differences in human operation inevitably lead to deviations in test results, and multiple manual steps significantly reduce detection efficiency. Furthermore, traditional sieving methods lack direct volume measurement design, relying solely on density conversion to obtain tobacco leaf volume data. However, tobacco leaf moisture content and density dynamically change with environment and grade, thus affecting density parameters and leading to low accuracy in volume calculation.

[0006] Conventional machine vision inspection methods suffer from limitations in both accuracy and practicality due to design constraints in their acquisition methods and volume detection techniques. Employing a monocular, single-angle image acquisition approach, they have inherent blind spots. When tobacco leaves are bent or not laid flat, incomplete extraction of the leaf outline is easily achieved, leading to inaccuracies in length and width calculations. Furthermore, conventional machine vision inspection solutions lack a visual volume detection module, relying instead on traditional density conversion methods. This adds extra steps to density measurement and formula conversion, and because tobacco leaf density fluctuates due to environmental and grade factors, volume calculation errors are significant, making it impossible to obtain accurate tobacco leaf volume data. Summary of the Invention

[0007] The purpose of this invention is to provide a tri-vision collaborative device and method for directly detecting the length and volume of tobacco shreds. Addressing the shortcomings of existing technologies for tobacco shred length and volume detection in the tobacco industry, this invention utilizes an integrated metering-transportation-acquisition-control device structure. A tri-vision acquisition system is constructed in the visible light detection area, consisting of one CCD positioned directly above, two CCDs positioned to the sides, and three sets of matching LED light sources. Combined with a suitable tobacco shred metering, feeding, and cleaning structure, the system uses a computer to achieve multi-component parameter linkage control and image processing calculations, ultimately achieving accurate measurement of tobacco shred length and direct visual representation of volume. This solution significantly reduces measurement errors caused by noise interference from tiny tobacco shreds and tobacco leaf shrinkage through multi-view synchronous acquisition and targeted algorithm optimization. The detection accuracy and efficiency are improved compared to existing technologies, and the device structure and detection method form a complete technical solution.

[0008] To achieve the above objectives, this application employs the following technical solution:

[0009] The three-vision collaborative direct detection device for tobacco length and volume includes a tobacco metering and discharging unit, a tobacco uniform conveying unit, a three-vision acquisition unit, a cleaning unit, and a computer control and calculation output unit. The above units are mechanically connected in sequence according to the tobacco conveying direction, and all electronic control components are electrically connected to the computer.

[0010] Furthermore, the tobacco shred calculation and discharge unit includes a sample library, a weighing sensor, a vacuum port, and a positive pressure port. The weighing sensor is symmetrically arranged at the bottom support of the sample library. The discharge end of the sample library has an air path interface along the direction of tobacco shred falling. The interface is sealed and connected to the vacuum port and the positive pressure port respectively, and the air path interface is directly opposite the tobacco shred discharge channel. The electronic control and air path regulation terminals of the sample library, the weighing sensor, and the positive and negative pressure ports are all electrically connected to the computer via data cables.

[0011] Furthermore, the discharge channel of the sample library has a tapered structure, which seamlessly connects with the feed end of the conveyor belt of the subsequent material homogenization and transmission unit; the vacuum port is connected to a negative pressure air source, and the positive pressure port is connected to a positive pressure air source.

[0012] Furthermore, the tobacco shred uniform conveying unit includes a uniform rake and a conveyor belt. The uniform rake is rotatably connected above the feed end of the conveyor belt and corresponds to the discharge end of the sample storage.

[0013] Furthermore, the tri-lens vision acquisition unit includes three CCD cameras and three sets of LED light sources. The three sets of LED light sources correspond one-to-one with the three CCD cameras, are arranged coaxially, and are all fixed on the periphery of the conveyor belt detection section. One CCD camera and its matching LED light source are installed perpendicular to the conveyor belt surface, with the lens facing the center of the detection section. The other two CCD cameras and their matching LED light sources are symmetrically arranged on both sides of the conveyor belt, with their lenses tilted at 30-60 degrees and pointing towards the center of the detection section. The shooting fields of the three CCD cameras completely overlap.

[0014] Furthermore, the cleaning unit is a cleaning roller, which is symmetrically arranged at the discharge end and closely fits the surface of the conveyor belt, rotating synchronously with the conveyor belt.

[0015] Furthermore, the computer-controlled computing output unit includes control software, a database, and a parameter output module.

[0016] The method for direct detection of tobacco length and volume using a tri-vision cooperative system, employing any of the aforementioned tri-vision cooperative devices for direct detection of tobacco length and volume, comprises the following specific steps:

[0017] S1. System parameter initialization and camera joint calibration;

[0018] S2. Simultaneous acquisition of tobacco shreds and tri-lens images;

[0019] S3. Image preprocessing and contour extraction;

[0020] S4. Calculation of actual tobacco length;

[0021] S5, trinocular stereo matching and 3D point cloud reconstruction;

[0022] S6. Point cloud post-processing;

[0023] S7. Calculation of tobacco volume using the analyte method;

[0024] S8. Output and storage of test results.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. Design a three-view vision synchronous acquisition structure, with the fields of view of the three CCD cameras completely overlapping, the LED light source and the camera arranged coaxially, and the side CCD camera tilted to point to the center of the detection section, eliminating the inherent blind zone of single-view acquisition, solving the problem of incomplete contour extraction caused by the bending of tobacco and non-flat placement, and laying the hardware foundation for multi-dimensional feature extraction and three-dimensional measurement of tobacco.

[0027] 2. A multi-view fusion algorithm for accurate measurement of tobacco length is proposed. Based on the bending morphology of tobacco, the algorithm completes the contour geometric posture correction through a rotation matrix. It combines equidistant slice sampling, polynomial curve fitting and calculus arc length integration to calculate the actual curve length of a single view. Then, the three-view results are weighted and fused according to the advantages of view detection, which breaks through the measurement error limitation of the traditional straight line fitting method and greatly improves the accuracy of tobacco length detection.

[0028] 3. A method for directly calculating tobacco volume based on multi-view visual features is developed. Core two-dimensional parameters such as the projected area above the tobacco and the height or width of the projected area to the side are extracted from the tri-view visual acquisition. A simplified three-dimensional geometric model is constructed according to the tobacco type to directly calculate the tobacco volume. This completely eliminates the indirect conversion method of "mass / density" in the existing technology and gets rid of the influence of density parameter fluctuations on volume detection.

[0029] 4. The geometric posture correction method of tobacco shreds is adopted. The tilted tobacco shreds on the conveyor belt are uniformly adjusted to a horizontal posture by rotating the matrix. This establishes a unified geometric benchmark for the three-view measurement of tobacco shred length and volume, and eliminates the detection deviation caused by the tobacco shred placement angle.

[0030] 5. A linkage control mechanism for detection parameters is proposed, which sets the conveyor belt running speed in a positive correlation with the CCD camera acquisition frame rate and the LED light source brightness. The computer fine-tunes the camera frame rate according to the real-time speed of the conveyor belt to prevent image ghosting and feature loss and ensure the clarity of the acquired image.

[0031] 6. Configure an anti-sticking and self-cleaning structure adapted to tobacco shred detection. Use an anti-static smooth conveyor belt to avoid tobacco shred sticking. Set a cleaning roller that rotates synchronously with the belt at the discharge end to remove tobacco shred residue and debris from the belt surface in real time, ensuring a clean background in the detection area and eliminating measurement errors caused by background interference.

[0032] 7. Design a positive and negative pressure auxiliary air path and a closed-loop control structure for the tobacco shred discharge, which achieves unblocked tobacco shred discharge through vacuum adsorption and positive pressure push. Combined with the weight data from the weighing sensor, the air path pressure is adjusted in real time to achieve quantitative tobacco shred discharge and avoid material accumulation. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the detection device of the present invention.

[0034] Figure 2 This is a block diagram of the detection device of the present invention.

[0035] Figure 3 This is a flowchart of the visual inspection process of the present invention.

[0036] Explanation of reference numerals in the attached figures:

[0037] 1. Weighing sensor; 2. Sample storage chamber; 3. Vacuum port; 4. Positive pressure port; 5. Material distribution rake; 6. Cleaning roller. Detailed Implementation

[0038] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings. The following embodiments are merely exemplary and can only be used to explain and illustrate the technical solution of the present invention, and should not be construed as limiting the technical solution of the present invention.

[0039] like Figure 1 and Figure 2 As shown, Figure 1 The middle arrow points to the direction of tobacco conveying. This application's three-view vision-coordinated direct detection device for tobacco length and volume includes a tobacco metering and discharging unit, a tobacco uniform conveying unit, a three-view vision acquisition unit, a cleaning unit, and a computer control and calculation output unit. These units are mechanically connected sequentially according to the tobacco conveying direction, and all electrical control components are electrically connected to the computer. The computer uniformly completes parameter control, data reception, processing calculations, and result output. The device has a compact overall structure and is suitable for online real-time tobacco detection requirements.

[0040] 1. Tobacco Metering and Discharge Unit:

[0041] This unit is the core execution unit at the feeding end of the device, which realizes the functions of quantitative storage, real-time accurate measurement, and continuous and stable discharge of tobacco. The core components include sample chamber 2, weighing sensor 1, vacuum port 3, and positive pressure port 4. Each component is mechanically connected according to functional requirements and forms an electrical linkage control with the computer: the weighing sensor is symmetrically set at the bottom support of the sample chamber, providing stable mechanical support for the sample chamber while realizing real-time measurement of the weight of tobacco; the discharge end of the sample chamber has a gas path interface along the direction of tobacco falling, and the interface is sealed to the vacuum port and positive pressure port respectively, and the gas path interface is directly opposite the tobacco discharge channel to ensure that the aerodynamic force acts directly on the discharged tobacco; the electrical control and gas path regulation terminals of the sample chamber, weighing sensor, and positive and negative pressure ports are all electrically connected to the computer via data cables, and the computer realizes unified parameter control and data interaction.

[0042] Sample storage chamber: This is the basic component for storing and discharging tobacco shreds. Its effective storage volume can hold over 200g of tobacco shreds to be tested, meeting the feeding requirements for continuous tobacco shred testing. The sample storage chamber's discharge channel features a tapering design, seamlessly connecting with the feed end of the conveyor belt in the subsequent material distribution unit, ensuring that the tobacco shreds fall along a fixed path and preventing deviation.

[0043] Weighing sensor: A high-precision measuring component that is rigidly connected to the sample library. It can collect the weight data of the tobacco in the sample library in real time and accurately, and convert the weight signal into an electrical signal and transmit it to the computer in real time through a data line, providing accurate data feedback for the closed-loop control of the subsequent output.

[0044] Vacuum and positive pressure ports are auxiliary gas path components for tobacco discharge, working together to form a pneumatic control system for tobacco discharge. The vacuum port is connected to a negative pressure gas source, generating negative pressure adsorption force within the discharge channel during operation. This smoothly adsorbs the tobacco from the sample chamber into the discharge channel, effectively preventing bridging and blockage. The positive pressure port is connected to a positive pressure gas source, generating directional positive pressure thrust during operation. This smoothly and continuously pushes the tobacco in the discharge channel onto the conveyor belt of the subsequent material distribution unit, ensuring no tobacco residue is left after discharge. The on / off status and pressure of the vacuum and positive pressure ports are dynamically controlled by a computer using real-time weight data collected by a weighing sensor, precisely matching the required tobacco discharge volume.

[0045] 2. Tobacco Shred Equalization Conveying Unit:

[0046] This unit is the core conveying and spreading unit for tobacco shreds, achieving uniform spreading and constant-speed transmission of the tobacco shreds. The core components are the equalizing rake 5 and the anti-static conveyor belt. The equalizing rake is rotatably connected above the feed end of the conveyor belt and corresponds to the discharge end of the sample storage. The conveyor belt is powered by a servo drive mechanism, which is electrically connected to a computer. The running speed of the conveyor belt is uniformly controlled by the computer.

[0047] The uniform rake has an adjustable distance from the conveyor belt. After the tobacco leaves are discharged from the sample storage, they fall onto the conveyor belt and are evenly spread into a single layer by the rotating uniform rake. This prevents the tobacco leaves from piling up and overlapping, ensuring the effectiveness of subsequent visual acquisition.

[0048] Anti-static conveyor belt: It serves as the carrier for transporting tobacco shreds. The belt surface is smooth and prevents tobacco shreds from sticking, thus avoiding detection errors caused by tobacco shreds adhering to the surface. The running speed of the conveyor belt can be adjusted in real time by computer according to the detection requirements, and it is linked with the camera frame rate of the subsequent vision acquisition unit to prevent image ghosting.

[0049] 3. Trinocular vision acquisition unit:

[0050] The trinocular vision acquisition unit is the core detection unit of this device and the core of the transformation of the original visible light detection area, enabling simultaneous acquisition of multi-view images of tobacco. The core components include three CCD cameras and three sets of matching LED light sources. The installation and connection relationships of each component are as follows: the three sets of LED light sources correspond one-to-one with the three CCD cameras, arranged coaxially, and all fixed to the periphery of the conveyor belt detection section. One CCD camera and its matching LED light source are installed perpendicular to the conveyor belt surface, with the lens facing the center of the detection section; two CCD cameras and their respective matching LED light sources are symmetrically positioned on both sides of the conveyor belt, with their lenses tilted at 45° towards the center of the detection section. The shooting fields of the three CCD cameras completely overlap, ensuring the acquisition of multi-view images of the same tobacco area; all LED light sources and CCD cameras are electrically connected to a computer, which controls the light source brightness, camera synchronization triggering, and acquisition frame rate.

[0051] Three sets of LED light sources: all are high color rendering and high brightness surface light sources, with uniform light diffusion. The brightness can be adjusted in real time by a computer. During operation, they provide supplementary lighting for the corresponding CCD cameras, avoiding reflections and shadows on the tobacco surface, ensuring clear tobacco image features, and meeting the anti-interference requirements of real-time detection in industrial settings. CCD cameras: high-resolution linear CCD cameras of the same specifications, featuring large depth of field, high sensitivity, and low noise. The acquisition frame rate can be controlled by a computer and is linked to the conveyor belt speed. The computer sends synchronous trigger signals to all three cameras to ensure consistent image acquisition time. The acquired analog image signals are transmitted to the computer in real time for subsequent processing.

[0052] 4. Cleaning Unit:

[0053] The core component is the cleaning roller 6. In the figure, the cleaning roller is symmetrically located at the discharge end and closely fits the surface of the conveyor belt. It rotates synchronously with the conveyor belt. Its function is to remove tobacco residue and debris from the surface of the conveyor belt in real time, ensure the surface cleanliness of the conveyor belt detection section, avoid image acquisition and calculation errors caused by background interference, and ensure the accuracy of the detection results.

[0054] 5. Computer-controlled calculation output unit:

[0055] The computer serves as the control center and data processing core of this device, integrating three major functional modules: database, control software, and parameter output module. It is electrically connected to all electronically controlled components of the device (weighing sensors, vacuum and positive pressure port air circuit controllers, servo drive mechanisms, LED light sources, and three CCD cameras) via data cables. This enables the computer to perform parameter linkage control, data reception, image processing, calculation analysis, and result output for the entire device. The implementation methods of each functional module are as follows:

[0056] Control software: This is the core control module. It sets and adjusts the operating parameters of each component according to the type of tobacco to be inspected, including vacuum and positive pressure, conveyor belt speed, LED light source brightness, CCD camera frame rate, etc. At the same time, it sends synchronous trigger signals to ensure that each component works in coordination.

[0057] Database: This is a data storage module that stores in real time the measurement data from the weighing sensor, the image data acquired by the CCD camera, and the calculated data on the length and volume of tobacco shreds. It supports long-term storage and traceability of batch data.

[0058] Parameter output module: This is the result display module, which outputs the detection parameters such as tobacco length and volume obtained by computer in real time. It also supports statistical analysis of the detection data to meet the needs of refined management and control of tobacco processing.

[0059] like Figure 3 As shown, this application relates to a method for detecting the length and volume of tobacco shreds. Utilizing the aforementioned tri-vision co-operated direct detection device for tobacco shred length and volume, the system parameters are first initialized and the tri-vision camera is jointly calibrated. Then, after standardized tobacco shred transport, the tri-vision camera simultaneously acquires images. After image preprocessing to extract the tobacco shred contour, the actual tobacco shred length is output through contour pose correction, single-view length calculation, and tri-view weighted fusion. An initial 3D point cloud is then obtained through SGBM stereo matching, depth map generation, and ICP registration fusion. A clean 3D point cloud of tobacco shreds is obtained through statistical filtering denoising and RANSAC background segmentation. Finally, the actual tobacco shred volume is output through voxel mesh generation, point cloud mapping, and voxelization calculation. The detection results are output in real time and fully stored for traceability, completing a single detection.

[0060] The specific steps of the detection method are as follows:

[0061] S1. System parameter initialization and camera joint calibration: Set the conveyor belt speed, light source brightness, and camera frame rate according to the tobacco type and match them in a coordinated manner; use Zhang Zhengyou checkerboard calibration method to obtain the intrinsic parameters, distortion coefficients and pose parameters between the three cameras, optimize and unify them to the world coordinate system (the conveyor belt plane is the XY plane, and the Z axis is upward) through the bundle adjustment method, and the reprojection error is ≤0.3 pixels. Establish the pixel-physical coordinate mapping relationship.

[0062] S2. Synchronous acquisition of tobacco shreds and three-lens image: The tobacco shreds to be inspected are quantitatively discharged from the metering and discharge unit to the conveyor belt, and are combed into a single layer of non-overlapping tobacco shreds by the equalizing rake and conveyed to the detection section at a uniform speed; when the tobacco shreds reach the center of the detection section, the computer triggers the external trigger module to control the three cameras to acquire images synchronously under the supplementary lighting of LED light source. The camera directly above acquires the top view contour map, and the cameras on both sides acquire the side view depth map. The image data is transmitted to the computer in real time.

[0063] S3. Image preprocessing and contour extraction: The three-view images are sequentially subjected to grayscale denoising (weighted average method + 3×3 median filtering + morphological closing operation), feature enhancement segmentation (channel weighting method + local adaptive thresholding method), and contour extraction (connected component repair + contour detection, filtering out noise <5 pixels) to obtain the complete contour of a single strand or cluster of tobacco shreds.

[0064] S4. Calculation of actual tobacco length: The contour tilt angle is obtained by the minimum bounding rectangle method, and a rotation matrix is ​​constructed to correct it to a horizontal posture; the corrected contour is sampled at equal intervals, and the actual length of a single view is obtained by fitting a cubic polynomial curve and integrating the arc length by calculus, combined with the pixel equivalent; the final actual length of the tobacco is obtained by weighting and fusing the data with a weight of 0.5 for the camera directly above and 0.25 for each of the cameras on both sides.

[0065] S5. Trinocular Stereo Matching and 3D Point Cloud Reconstruction: Using the image directly above as a reference, perform SGBM semi-global stereo matching (disparity 0-128 pixels, matching block 5×5) with the images on both sides to generate a disparity map and fill in the holes; based on calibration parameters and triangulation principles, convert the disparity map into a depth map and generate a single-view point cloud; convert the point clouds on both sides to the coordinate system of the camera directly above, and after fine registration and deduplication using the ICP algorithm, fuse them into the initial 3D point cloud of tobacco.

[0066] S6. Point cloud post-processing: Statistical filtering is used to remove 3... Outlier points are identified, and the RANSAC algorithm is used to fit and segment the conveyor belt background. Point clouds of tobacco shreds exceeding the preset thickness threshold are retained to obtain a pure 3D point cloud of tobacco shreds.

[0067] S7. Voxelization method for calculating tobacco volume: Set a voxel resolution of 0.5mm, divide the three-dimensional space of tobacco into a cubic voxel mesh and calculate the volume of a single voxel; map the pure point cloud onto the mesh, mark and count the total number of occupied voxels, and obtain the final actual volume of tobacco by calculating the actual volume of tobacco = total number of occupied voxels × volume of a single voxel.

[0068] S8. Output and storage of test results: The computer outputs the length and volume parameters of the tobacco shreds in real time through the human-machine interface; at the same time, it stores the measurement, images, test data and test time, and supports batch data statistical analysis and historical traceability.

[0069] This application also includes two CCD cameras on the side and their matching LED light sources. The original design is to tilt the camera towards the center of the conveyor belt detection section. The tilt angle can be adjusted arbitrarily within the range of 30°-60°. There is no need to fix a single angle. As long as the shooting fields of the three CCD cameras are completely overlapped, the multi-view features of the tobacco can be effectively extracted without affecting the accuracy of subsequent calculations.

[0070] The actual shape of the tobacco shreds can be replaced with an ellipsoid, frustum, or irregular polyhedron fitting model, or a simple calculation model combining projected volume and average thickness can be used. As long as the model size can be determined by the two-dimensional geometric parameters extracted from three perspectives, the volume of the tobacco shreds can be directly calculated, eliminating the dependence on density parameters.

[0071] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A tri-vision co-operated device for direct detection of tobacco length and volume, characterized in that, It includes a tobacco metering and discharging unit, a tobacco uniform conveying unit, a three-eye vision acquisition unit, a cleaning unit, and a computer control and calculation output unit; the above units are mechanically connected in sequence according to the tobacco conveying direction, and all electronic control components are electrically connected to the computer.

2. The tri-vision coordinated direct detection device for tobacco length and volume according to claim 1, characterized in that, The tobacco feeding calculation unit includes a sample library, a weighing sensor, a vacuum port, and a positive pressure port. The weighing sensor is symmetrically arranged at the bottom support of the sample library. The discharge end of the sample library has an air path interface along the direction of tobacco falling. The interface is sealed and connected to the vacuum port and the positive pressure port respectively, and the air path interface is directly opposite the tobacco discharge channel. The electronic control and air path regulation terminals of the sample library, the weighing sensor, and the positive and negative pressure ports are all electrically connected to the computer via data cables.

3. The tri-vision coordinated direct detection device for tobacco length and volume according to claim 2, characterized in that, The sample storage outlet has a tapered structure that seamlessly connects with the feed end of the conveyor belt of the subsequent material equalization and transmission unit; the vacuum port is connected to a negative pressure air source, and the positive pressure port is connected to a positive pressure air source.

4. The tri-vision coordinated direct detection device for tobacco length and volume according to claim 1, characterized in that, The tobacco shred uniform conveying unit includes a uniform rake and a conveyor belt. The uniform rake is rotatably connected above the feed end of the conveyor belt and corresponds to the discharge end of the sample storage.

5. The tri-vision coordinated direct detection device for tobacco length and volume according to claim 1, characterized in that, The tri-lens vision acquisition unit includes three CCD cameras and three sets of LED light sources. The three sets of LED light sources correspond one-to-one with the three CCD cameras and are arranged coaxially. They are all fixed on the periphery of the conveyor belt detection section. One CCD camera and its matching LED light source are installed perpendicular to the conveyor belt surface, with the lens facing the center of the detection section. The other two CCD cameras and their matching LED light sources are symmetrically arranged on both sides of the conveyor belt, with their lenses tilted at 30-60 degrees and pointing towards the center of the detection section. The shooting fields of the three CCD cameras completely overlap.

6. The tri-vision co-operated direct detection device for tobacco length and volume according to claim 1, characterized in that, The cleaning unit is a cleaning roller, which is symmetrically located at the discharge end and closely fits the surface of the conveyor belt, rotating synchronously with the conveyor belt.

7. The tri-vision co-operated direct detection device for tobacco length and volume according to claim 1, characterized in that, The computer-controlled calculation output unit includes control software, a database, and a parameter output module.

8. A method for direct detection of tobacco length and volume using a tri-vision cooperative system, comprising the tri-vision cooperative system for direct detection of tobacco length and volume according to any one of claims 1 to 7, characterized in that, The specific steps are as follows: S1. System parameter initialization and camera joint calibration; S2. Simultaneous acquisition of tobacco shreds and tri-lens images; S3. Image preprocessing and contour extraction; S4. Calculation of actual tobacco length; S5, trinocular stereo matching and 3D point cloud reconstruction; S6. Point cloud post-processing; S7. Calculation of tobacco volume using the analyte method; S8. Output and storage of test results.

Images