A tobacco leaf flow detection method based on 3D TOF camera and sectional area integration

Abstract

This invention discloses a method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D Time-of-Flight (TOF) camera and cross-sectional integration, and its application. It relates to the field of tobacco processing automation technology and includes the following steps: When the conveyor belt is unloaded, a 3D TOF camera fixedly installed above the conveyor belt acquires the reference height plane of the belt; wherein the optical axis of the 3D TOF camera is perpendicular to the bearing plane of the conveyor belt; when the conveyor belt is conveying material, real-time three-dimensional point cloud data containing both the material and the belt is acquired by the 3D TOF camera. This invention, through a 3D TOF camera and cross-sectional integration technology, achieves high-precision, non-contact, intrinsically safe, and highly interference-resistant detection of tobacco leaf flow rate in a tobacco processing line, and simultaneously acquires multi-dimensional process information such as material distribution and foreign matter agglomerates, comprehensively overcoming the problems of limited measurement accuracy, contact damage, radiation hazards, light sensitivity, and limited information in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of tobacco processing automation technology, and in particular to a method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration. Background Technology

[0002] In cigarette manufacturing production lines, continuous and precise flow monitoring of materials such as tobacco leaves and stems during belt conveyor transport is crucial for achieving accurate process control and ensuring the uniformity and stability of the final product quality. Fluctuations in material flow directly affect the processing effects of subsequent critical steps such as flavoring, adding ingredients, and drying. Therefore, finding a high-precision, real-time, and non-destructive online flow monitoring method has always been a technological requirement in this field. Currently, the following technical solutions are mainly used for material flow monitoring in cigarette manufacturing lines: The first type is the nuclear balance or weighing nuclear density meter. This type of device uses the attenuation law of gamma rays as they penetrate materials to estimate the mass of the materials. Although it can achieve high measurement accuracy, it relies on a radioactive source and has inherent radioactive safety hazards. The installation, use, and maintenance of the equipment must follow strict protection and regulatory standards. In addition, the system is expensive and the decommissioning disposal is complicated, which limits its large-scale application.

[0003] Secondly, there are mechanical scraper or leveling devices. Their working principle is to force the material on the conveyor belt to a fixed height using a mechanical structure, and then calculate the volumetric flow rate based on the belt speed. The main drawback of this approach is that the direct physical contact between the scraper and the material can cause significant mechanical damage to fragile materials such as tobacco leaves, disrupting their flaky structure, causing breakage, and directly affecting product quality. Furthermore, for materials with poor flowability or that easily clump together, the uniformity of leveling is difficult to guarantee, thus introducing significant measurement errors.

[0004] Thirdly, there is the estimation method based on 2D visual images. This method involves using an industrial camera to photograph the material from above, extracting the two-dimensional contour or texture information of the image to indirectly estimate the material coverage or volume. This approach only obtains the planar projection information of the material, completely lacking height dimension data. When the thickness of the material accumulation on the conveyor belt changes dynamically, the estimation results based on two-dimensional images will produce fundamental deviations. In addition, this approach is quite sensitive to the lighting conditions in the production environment and the differences in reflectivity caused by changes in material color and moisture content, resulting in insufficient robustness of the measurement.

[0005] Fourthly, there is the single-point or single-line laser ranging method. This method measures the material height at a fixed point or along a straight line along the belt direction by installing a single-point laser displacement sensor above the belt. However, this method has a very limited representative area and cannot obtain a complete cross-sectional profile of the material across the entire belt width. When the material is non-uniformly distributed on the belt, or when there are local accumulations or depressions, the single-point or single-line sampling results cannot accurately reflect the overall material condition, resulting in poor measurement accuracy and representativeness.

[0006] In summary, existing technologies suffer from limitations in measurement accuracy, are prone to causing physical damage to materials, pose safety and protection risks, and cannot accurately reproduce the three-dimensional overall picture of irregularly stacked materials. Therefore, a new solution is urgently needed. Summary of the Invention

[0007] This invention aims to address the shortcomings of existing technologies by proposing a method for detecting tobacco leaf flow rate in tobacco processing lines based on a 3D TOF camera and cross-sectional integration, and its application. The goal is to improve the precision and intelligent control of tobacco processing lines by integrating 3D TOF depth sensing and cross-sectional contour integration technologies. This will solve the technical problems of limited measurement accuracy, easy physical damage to materials, safety and protection hazards, and inability to accurately reproduce the three-dimensional full view of irregular material accumulation.

[0008] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The present invention provides a method for detecting the flow rate of tobacco leaves in a tobacco processing line based on a 3D TOF camera and cross-sectional integration, characterized by the following steps: Step 1: With the conveyor belt unloaded, obtain the reference height plane of the belt using a 3D TOF camera fixedly installed above the conveyor belt; wherein, the optical axis of the 3D TOF camera is perpendicular to the bearing plane of the conveyor belt; Step 2: While the conveyor belt is conveying materials, acquire real-time 3D point cloud data containing the materials and the belt using a 3D TOF camera; Step 3: Subtract the real-time 3D point cloud data from the reference height surface to generate a pure height point cloud of the material, and filter the pure height point cloud of the material to remove outlier noise, thereby obtaining the filtered pure height point cloud of the material. Step 4: Based on the filtered pure height point cloud of the material, set multiple virtual cross-sections perpendicular to the direction of belt movement at equal intervals along the belt movement direction, thereby extracting the height contour data of the material in each virtual cross-section. Step 5: Perform curve fitting on the height profile data to generate a continuous height profile curve; Step 6: Integrate the height profile curve to obtain a set of material cross-sectional area sequences at the virtual cross-section. ,in, Indicates the first The cross-sectional area of ​​the material, The number of virtual cross sections; Step 7: Calculate the average cross-sectional area based on the material's cross-sectional area and the synchronously acquired real-time belt speed. It is used to calculate the instantaneous volumetric flow rate of a material.

[0009] Optionally, step 6 includes: Along the width direction of the belt, the height value is scanned from both sides of the material towards the center, and the position where the first height value exceeds the preset noise threshold is determined as the lower limit and upper limit of integration, respectively. Within the integration interval formed by the lower and upper limits of integration, the height profile curve is subjected to definite integral operation, and the resulting integral value is the cross-sectional area of ​​the material.

[0010] Optional, also includes: Step 8: Multiply the instantaneous volumetric flow rate by the pre-calibrated material density to obtain the instantaneous mass flow rate of the material.

[0011] Optional, also includes: Step 9: Compare the instantaneous volumetric flow rate or instantaneous mass flow rate with the preset flow rate setting value, and calculate the flow rate deviation value; Step 10: Based on the flow deviation value, generate an adjustment control signal using a PID control algorithm, and output the adjustment control signal to the upstream feeding device of the tobacco processing line to adjust the discharge speed or discharge volume of the feeding device, thereby achieving closed-loop stable control of the tobacco leaf flow.

[0012] Optionally, in step 10, the parameters of the PID control algorithm are automatically switched according to the current material flow rate change rate: When the rate of change of flow exceeds the preset rapid change threshold, a high proportional gain coefficient is used to improve the response speed. When the rate of change of flow is less than the preset steady change threshold, a low proportional gain coefficient is used to prevent overshoot oscillation.

[0013] Optional features include material breakage detection: Continuously monitor the material cross-sectional area A or instantaneous mass flow rate Qm at the virtual cross-section; If the cross-sectional area A of the material detected in N consecutive frames is zero or less than the preset material breakage threshold, or the instantaneous mass flow rate Qm detected in N consecutive frames is zero or less than the preset material breakage mass threshold, then it is determined that the material is currently in a material breakage state, and a material breakage alarm signal is output; where N is the preset number of consecutive frames threshold.

[0014] Optional features include material blockage detection: Continuously monitor the material height profile curve H(x) at the virtual cross-section; If the height value of the height profile curve H(x) at any position in the belt width direction exceeds the preset blockage height threshold for a duration exceeding the preset time threshold, or if the material cross-sectional area A exceeds the preset blockage area threshold for a duration exceeding the preset time threshold, then it is determined that the current state is blocked, and a blockage alarm signal and / or deceleration control signal are output to the belt drive mechanism.

[0015] Optional features include material deviation detection: Step A: Within the virtual cross-section, based on the material height profile curve H(x), calculate the distribution center position C of the material in the belt width direction using the formula: In the formula, and These are the lower limit and upper limit of integration for the integration interval, respectively; Step B: Align the distribution center position C with the width center position of the belt. Compare and calculate the offset. ; Step C, if the offset If the deviation exceeds the preset tolerance threshold, it is determined that the material has deviated, and a deviation alarm signal is output. Simultaneously, based on the deviation amount... The direction and magnitude of the material are used to generate a correction control signal to drive the correction device to adjust the conveying posture of the material.

[0016] Optional features include detection of abnormal material clumps: Within the virtual cross-section, the second derivative H″(x) of the height profile curve H(x) is obtained; If there exists a certain position within the integration interval , making If the curvature change threshold is greater than the preset threshold, then the corresponding position is determined. An abnormal material clumping is detected, and the location and size information of the abnormal material clumping are recorded. At the same time, an abnormal clumping warning signal is output.

[0017] Optional features include belt misalignment detection: Based on the reference height surface obtained in step 1, extract the reference edge positions of the two sides of the belt under no-load conditions; While the conveyor belt is conveying materials, the real-time edge positions of both sides of the belt are simultaneously acquired using a 3D TOF camera. The real-time edge position is compared with the reference edge position to calculate the offset of the belt edge; If the offset exceeds the preset belt misalignment threshold, it is determined that the belt body has misaligned and a belt misalignment alarm signal is output.

[0018] Compared with the prior art, this application includes at least one of the following beneficial technical effects: 1. This invention acquires a complete three-dimensional point cloud of the material along the width of the conveyor belt using a 3D TOF camera. Curve fitting is then used to generate a continuous height profile that accurately reflects the material's accumulation pattern. The cross-sectional area is then precisely calculated using an integral method. Compared to 2D vision solutions that only acquire the material's coverage area, and laser ranging solutions that only measure the height of a single point or line, this method fundamentally overcomes the inherent errors caused by uneven material accumulation and irregular distribution, significantly improving flow measurement accuracy.

[0019] 2. The entire measurement process of this invention requires no physical contact between any mechanical parts and the material. Compared to mechanical methods that use scrapers to force leveling, which can easily damage and break the leaf-like structure of tobacco leaves, this method perfectly protects the integrity of fragile materials such as tobacco leaves and has no adverse impact on product quality.

[0020] 3. This invention uses a 3D TOF camera as the core sensor and utilizes the near-infrared time-of-flight principle for three-dimensional measurement, without containing any radiation sources. Compared to nuclear balance solutions that rely on gamma rays and have strict regulatory and protection requirements, this method is completely operator- and environmentally friendly, requires no special safety protection, and makes the entire lifecycle management of the equipment more convenient and safer.

[0021] 4. This invention measures depth information, and its principle ensures that the measurement results are unaffected by changes in ambient lighting, material color differences, or alterations in texture characteristics. Compared to 2D visual estimation schemes that are sensitive to lighting and material reflectivity, this method maintains stable and consistent measurement performance for different batches of tobacco leaves with varying moisture contents, resulting in higher long-term system reliability.

[0022] 5. In addition to outputting instantaneous flow rate, the three-dimensional point cloud data collected by this invention also contains rich information such as material distribution uniformity, packing morphology deviation, and the presence of large clumps or foreign objects. This data can provide strong data support for subsequent process status diagnosis, incoming material uniformity evaluation, and intelligent production line control. Existing other detection methods are unable to extract so many dimensions of process information from the same data source. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to the present invention. Figure 2 This is a three-dimensional point cloud data diagram of tobacco leaf lines in this invention. Detailed Implementation

[0024] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0025] Example 1: like Figure 1 As shown, a tobacco leaf flow detection system based on a 3D TOF camera and cross-sectional integration is used to realize the method of tobacco leaf flow detection in a tobacco processing line. In a specific embodiment, the system mainly includes: a three-dimensional data acquisition module, a speed acquisition module, a data processing module, and a human-computer interaction and output module.

[0026] The core of the 3D data acquisition module is one or more 3D TOF (Time-of-Flight) cameras, which are fixedly mounted above the conveyor belt of the spinning line. Their optical axes are substantially perpendicular to the belt's bearing plane to ensure complete depth information acquisition of the material across the entire width of the belt. The 3D TOF camera actively emits modulated near-infrared light and measures the time difference between its emission and reflection back to each pixel on the sensor, directly calculating the depth value of each point in the scene being photographed, thus generating 3D point cloud data. The camera continuously scans the area of ​​the conveyor belt below at a preset frequency (e.g., 30 frames per second or higher).

[0027] The speed acquisition module is used to acquire the real-time operating speed synchronized with the conveyor belt. In this embodiment, the module uses a high-precision rotary encoder linked to the belt-driven roller. The encoder outputs pulse signals as the roller rotates, and the data processing module calculates the belt's linear speed v in real time by counting the number of pulses per unit time and combining this with the roller's circumference information. Alternatively, it can directly read the belt speed signal provided by the existing control system of the production line (such as a PLC).

[0028] The data processing module is the core computing unit for implementing the method of this invention, and can be an industrial computer or embedded processor with sufficient computing power. This module receives the three-dimensional point cloud data stream from the 3D TOF camera and the belt speed signal from the speed acquisition module, runs the built-in algorithm program, and completes all operations such as point cloud preprocessing, contour fitting, area integration, and flow rate calculation.

[0029] The human-machine interface and output module is connected to the data processing module and mainly includes an industrial display screen and a standard data communication interface. This module is used to display key information such as instantaneous volumetric flow rate, instantaneous mass flow rate, cumulative flow rate, and material cross-sectional profile curve in real time. It can also send flow data and control signals to the production line central control system (PLC / DCS) to provide feedback for the quantitative blending of materials and the dynamic adjustment of flavoring and additives.

[0030] A method for measuring tobacco leaf flow that simultaneously offers high precision, non-contact operation, intrinsic safety, strong environmental adaptability, and rich information output is implemented according to the following steps within a complete measurement cycle: Step S1: System calibration and reference surface acquisition.

[0031] Before the production line starts feeding materials, system calibration is performed first. With the conveyor belt running unloaded, a 3D TOF camera is activated to continuously scan the surface of the empty belt, acquiring multiple frames of depth images of the empty belt. To improve the stability and accuracy of the reference surface, N consecutive frames (e.g., 50 frames) of depth images can be averaged to eliminate the influence of sensor random noise on the reference surface. The averaged depth images are converted into 3D point cloud data. A reference height surface Z0(x, y) representing the surface of the empty belt is generated by plane fitting or by directly taking the average of the depth values ​​of each pixel. This reference surface is stored in the memory of the data processing module as a data matrix, serving as a zero-point reference for subsequent calculations.

[0032] Step S2: Real-time 3D point cloud data acquisition.

[0033] When the production line is feeding materials and the tobacco leaves are being conveyed normally by the belt, such as Figure 1 As shown, a 3D TOF camera continuously scans the material on the conveyor belt at a fixed preset frequency, acquiring a mixed depth image in real time that includes the upper surface of the material and the exposed surface of the conveyor belt. Each acquired depth image is converted into a 3D point cloud dataset Pt containing (X, Y, Z) coordinate information and sent to the data processing module for further processing. Here, the X-axis represents the width direction of the conveyor belt, the Y-axis represents the direction of conveyor belt movement (i.e., the material conveying direction), and the Z-axis represents the height direction.

[0034] Step S3: Extraction and preprocessing of material pure height data.

[0035] This step aims to extract pure height information representing only the thickness of the material itself from the mixed point cloud and to remove noise. For example... Figure 2 As shown, the specific operation is as follows: First, the Z-coordinate values ​​of each data point in the real-time hybrid point cloud Pt obtained in step S2 are subtracted point by point from the height values ​​of the reference height surface Z0 stored in step S1 at the same (x, y) position to obtain the pure height value of the material h = Z - Z0. All pure height values ​​constitute the pure height point cloud data of the material.

[0036] Secondly, to address the outlier noise generated during the transport of tobacco leaves, such as flying fluff and fine debris due to vibration and wind disturbance, this invention preferably employs a Statistical Outlier Removal (SOR) algorithm to filter the pure height point cloud. This algorithm iterates through each point p in the point cloud, quickly searches for its K nearest neighbors by constructing a KD-Tree data structure, and calculates the average distance d(p) from point p to these K neighbors.

[0037] The average distance of all points is calculated, assuming it follows a Gaussian distribution. The global average distance μ and standard deviation σ are then calculated. A standard deviation multiple threshold T is set (e.g., T=1.0). If the average distance d(p) of a point p is greater than μ + T·σ, then point p is identified as an outlier noise point caused by flying fluff or debris and removed from the point cloud data. After processing by this algorithm, a clean, pure height point cloud of the material that accurately reflects the morphology of the main tobacco leaf accumulation is obtained.

[0038] Step S4: Extract cross-sectional profile data.

[0039] In the preprocessed point cloud data of the material's pure height, a virtual cross-section is defined perpendicular to the direction of belt movement (i.e., the Y direction). This cross-section corresponds to a slice region in the point cloud data where the Y coordinate is within a very small range [y-Δy, y+Δy]. The X and Z coordinates (i.e., height values) of all points within this slice region are extracted to form a discrete data point set {xi, hi} representing the material height distribution at this cross-section.

[0040] Step S5: Fitting the material height profile curve.

[0041] The discrete data point set {xi, hi} extracted in step S4 is used to generate a continuous and smooth material height profile curve H(x) in the belt width direction through a curve fitting algorithm.

[0042] This invention preferably uses a cubic B-spline curve as the fitting model. The advantage of choosing a B-spline curve lies in its excellent "local support": when fitting the profile of loose, polymorphic tobacco materials that are prone to irregular slopes and peaks and valleys, adjusting a control point only affects the curve shape of its neighboring intervals, without causing oscillations across the entire curve range. This effectively avoids the "Runge phenomenon" commonly seen at the edges of profiles in high-order global polynomial fitting. The fitting process first determines the initial positions of a set of control points based on the density and distribution of discrete points. Then, the coordinates of the control points that minimize the sum of distance errors between the B-spline curve and all valid data points are solved using the least squares optimization method. Finally, a smooth curve H(x) that highly approximates the actual material surface profile and has a certain degree of noise resistance is obtained.

[0043] Step S6: Calculate the cross-sectional area.

[0044] This step involves performing a definite integral on the height profile curve H(x) to accurately calculate the material area A of the virtual cross-section.

[0045] Before performing integration, the effective integration interval must be adaptively determined. The specific method for determining the interval is as follows: Along the belt width direction (X-axis), scan the height curve H(x) value pixel by pixel or step position from the left boundary xleft towards the center and from the right boundary xright towards the center.

[0046] The first location encountered during the scanning process with a height value greater than the preset noise threshold (e.g., 2mm, which is slightly larger than the sensor's measurement noise amplitude and can effectively eliminate interference from the belt surface and fine dust) is recorded as the integration starting point. and the end point of integration .

[0047] If no point exceeding the threshold is detected during the entire scanning process, it is determined that there is currently no material on the conveyor belt, A=0, and the process proceeds directly to the next frame.

[0048] After determining the integration interval, the cross-sectional area A of the material is calculated using a definite integral: A = ∫ (from Wstart to Wend) H(x) dx. In actual numerical calculations, numerical integration methods such as Simpson's Rule or the composite trapezoidal rule are used to accurately solve the above integral.

[0049] Step S7: Instantaneous flow rate calculation.

[0050] The data processing module simultaneously calculates the volumetric flow rate and mass flow rate of the material according to the following formula: First, calculate the instantaneous volumetric flow rate Qv. Multiply the cross-sectional area A obtained in step S6 by the synchronously acquired real-time belt speed v to obtain the volume of material flowing through the detection cross-section per unit time: Qv = A × v, unit: m3 / s.

[0051] Further, calculate the instantaneous mass flow rate Qm. Multiply the instantaneous volumetric flow rate Qv by the preset calibrated density ρ of the batch of tobacco material (unit: kg / m3) to obtain the mass of material flowing through the detection section per unit time: Qm = Qv × ρ = A × v × ρ, unit: kg / s.

[0052] The bulk density ρ of the material is established by the laboratory through offline measurement of tobacco leaves with different formulations and moisture contents to create a parameter database for the data processing module to call. Alternatively, it can be corrected in real time by integrating an online density measurement module into the system.

[0053] The system repeats steps S2 to S7 in a loop to achieve continuous, high-frequency, online detection of tobacco leaf flow.

[0054] In an application example, the method and system of the present invention are deployed in the tobacco processing workshop of a cigarette factory. A 3D TOF camera is fixedly installed above a key belt conveyor line for transporting materials to the blending cabinet, ensuring its field of view completely covers the full width of the belt. The camera is connected to an industrial computer and simultaneously acquires the real-time belt speed from the conveyor line PLC. After the system is put into operation, the real-time material height profile curve and instantaneous mass flow rate fluctuations can be clearly observed through the human-machine interface. By linking with the underlying PLC control system of the belt conveyor, the measured instantaneous flow rate data is used for closed-loop feedback control with the preceding feeder, effectively suppressing flow rate fluctuations caused by uneven material feeding. Long-term application verification shows that the system significantly improves the stability and uniformity of material feeding, operates without contact with tobacco leaves, poses no radioactive safety hazards, and exhibits good overall performance, optimizing the process performance of subsequent blending and flavoring processes.

[0055] Example 2: Based on Example 1, this embodiment further integrates functions such as flow closed-loop control, material interruption detection, material blockage detection, material deviation detection, abnormal material clumping detection, belt deviation detection, and multi-section average flow calculation, forming a tobacco leaf flow detection and control system for a tobacco processing line that combines accurate detection and intelligent control.

[0056] The system hardware configuration of this embodiment is basically the same as that of Embodiment 1, except that the data processing module has a richer set of algorithm functions, and the human-computer interaction interface has added corresponding alarm prompts and control parameter setting areas. The invention will be further described below with reference to the accompanying drawings.

[0057] I. Closed-loop flow control. Based on steps S1 to S7 above, this embodiment further performs the following steps to achieve closed-loop stable control of tobacco leaf flow: Step S8: Compare the instantaneous volumetric flow rate Qv or instantaneous mass flow rate Qm calculated in step S7 with the preset flow rate setpoint. Compare the preset flow rate settings. The process engineers set the preset flow rate setpoint by calculating the flow rate deviation value based on the production grade and standard process parameters. ,in, This represents the instantaneous flow rate at the current moment.

[0058] Step S9: Set the flow deviation value The input PID control algorithm is expressed in the continuous domain as equation (1): (1) In equation (1), To control the output, This is the proportional gain coefficient. The integral time constant is... The differential time constant is This is the cumulative integral term of the deviation. In practical digital control systems, the incremental PID algorithm is used for discretization, and its expression is given by equation (2): (2) In equation (2), Indicates the first Control increment at each sampling time Indicates the current number Flow deviation value at the time of the second sampling Represents the integral coefficient. Represents the differential coefficient. , , To control the period to be synchronized with the sampling period of the 3D TOF camera.

[0059] Specifically, the PID control parameters in this embodiment can be automatically switched according to the current material flow rate change rate. Let the current flow rate change rate be Equation (3): (3) In equation (3), This represents the instantaneous rate of change of the flow rate at the current moment. Indicates the current number The instantaneous flow rate value of the next sample. This indicates the time interval between two consecutive samples.

[0060] When the rate of change in traffic exceeds the preset rapid change threshold For example, if the threshold is set to ±0.5 kg / s², it can be calibrated according to the actual operating conditions of the production line. This indicates that the system is experiencing severe fluctuations in incoming material, and in this case, a high proportional gain coefficient is used. High proportional gain coefficient Take the benchmark It is 1.5 times faster to improve response speed, enabling the control system to quickly track flow changes and suppress large fluctuations.

[0061] When the rate of change in flow is less than the preset steady change threshold At that time, the threshold for steady change Setting the threshold to ±0.1 kg / s² indicates that the system is in steady-state operation. At this point, switching back to a low proportional gain coefficient is appropriate. Take the benchmark A factor of 0.8 is used to prevent overshoot oscillation and ensure the smoothness of the control process. Hysteresis comparison logic is introduced when switching between the two parameters to avoid control chatter caused by frequent parameter switching.

[0062] Step S10: The generated adjustment control signal u(t) is output to the upstream feeding equipment of the tobacco processing line, such as a quantitative feeder, electronic belt scale feeder, or vibrating feeder, through a standard industrial communication interface such as Profinet, 4-20mA analog signal, or EtherNet / IP protocol, so as to adjust the discharge speed or discharge amount of the feeding equipment in real time, thereby realizing closed-loop stable control of tobacco leaf flow.

[0063] II. Material Shortage Detection. This embodiment performs material shortage detection simultaneously during real-time flow detection. The specific steps are as follows: Step S11: Continuously monitor the material cross-sectional area A at the virtual cross-section obtained in step S6, and the instantaneous mass flow rate Qm obtained in step S7.

[0064] Step S12: Set the material breakage judgment logic: If N consecutive frames are detected, where N is a preset threshold for the number of consecutive frames (e.g., N=10, corresponding to approximately 0.33 seconds), the specific time can be set according to the camera frame rate and system response speed requirements. The material cross-sectional area A is detected to be zero or less than the preset material breakage area threshold. This threshold, slightly greater than zero, is used to avoid misjudgments due to sensor noise and minute impurities on the belt surface. For example, it can be set to 0.5% of the reference area value, or when the instantaneous mass flow rate Qm is zero or less than the preset material breakage mass threshold. If the value is set to 1% of the full capacity, it is determined that the material is currently cut off.

[0065] Step S13: Once a material shortage is detected, the data processing module immediately outputs a material shortage alarm signal, including audible and visual alarms and text prompts, through the human-machine interface. At the same time, it sends a material shortage status signal to the host computer, including the production line central control system PLC / DCS, through the communication interface, reminding the operator to check the upstream material supply, such as whether the feeder hopper is empty or whether the conveying channel is blocked. It can also link and control the upstream equipment to avoid energy waste and equipment wear caused by idling.

[0066] III. Material Blockage Detection. This embodiment also simultaneously performs a material blockage detection function, the specific steps of which are as follows: Step S14: Continuously monitor the material height profile curve H(x) at the virtual cross-section obtained in step S5, and the material cross-sectional area A obtained in step S6.

[0067] Step S15: Set the blockage detection logic: During material conveying, if the height value of the height profile curve H(x) at any position in the belt width direction continuously exceeds the preset blockage height threshold... The time exceeded the preset time threshold. The threshold It can be set according to the height of the belt baffle or the maximum stacking height allowed by the process, for example, set to 90% of the height of the belt baffle. The value is set to 3 seconds to avoid false triggering caused by instantaneous material fluctuations; or the material cross-sectional area A continuously exceeds the preset blockage area threshold. The time exceeds the preset time threshold If the current condition is determined to be a material blockage, the preset blockage area threshold is used. It can be set according to the upper limit of the effective conveying cross-sectional area of ​​the belt. In this embodiment, it is set to 85% of the cross-sectional area of ​​the belt.

[0068] Step S16: Once a blockage is detected, the data processing module outputs a blockage alarm signal to the human-machine interface and simultaneously generates a deceleration control signal to the frequency converter of the belt drive mechanism, causing the belt to slow down or stop to prevent material overflow, equipment overload, or even mechanical damage caused by excessive material accumulation. After the blockage is cleared, the system automatically returns to normal operation.

[0069] IV. Material Deviation Detection. This embodiment simultaneously monitors the uniformity of material distribution along the belt width and detects deviation during flow detection. The specific steps are as follows: Step S17: Within the virtual cross-section, based on the material height profile curve H(x) obtained in step S5, calculate the distribution center position C of the material in the width direction of the belt (i.e., the center of the first moment in the width direction of the material cross-section), and its calculation formula is Equation (4): (4) In equation (4), This represents the material height profile value at location x. These are the position coordinates along the width of the belt. and These are the lower limit and upper limit of the points, respectively.

[0070] Step S18: Align the distribution center position C with the width center position of the belt. (This position is obtained by edge detection of the reference height surface in step S1, or directly calibrated by the mechanical parameters of the belt conveyor.) Compare and calculate the offset. And record whether the offset direction is to the left or to the right.

[0071] Step S19: If the offset ΔC exceeds the preset deviation tolerance threshold This threshold is set based on the material alignment requirements of subsequent processes. For example, if it is set to 5% of the effective width of the belt, it is determined that the material has deviated, and a deviation alarm signal is output through the human-machine interface to prompt the operator to pay attention to whether the landing position of the incoming material on the belt has deviated. At the same time, a correction control signal is generated based on the direction and magnitude of the deviation ΔC, and output to the material guiding device installed upstream of the belt conveyor, such as an adjustable guide plate or a swing-type material distributor, to drive the correction device to adjust the conveying posture of the material, so that the material is gradually restored to the center area of ​​the belt.

[0072] V. Abnormal Material Clump Detection. This embodiment also utilizes the geometric features of the height profile curve to achieve non-contact automatic identification of abnormal material clumps. The specific steps are as follows: Step S20: Within the virtual cross-section, calculate the second derivative H″(x) of the height profile curve H(x) obtained in step S5 with respect to the position x in the width direction. H″(x) reflects the curvature change of the height profile: H″(x) ≈ 0 in flat areas, H″(x) is constant in sloping areas, and the extreme values ​​of H″(x) correspond to the local clump protrusions.

[0073] Step S21: In the integration interval The absolute value of H″(x) is scanned point by point within the interior. If there exists a certain position... , making Greater than the preset curvature change threshold This threshold is set based on the smoothness of the outline of normal tobacco leaf material. It is determined by 2 to 3 times the standard deviation of the statistical mean of the outline curvature of multiple batches of normal material. If an abnormal material clumping, such as wet clumps, stem agglomerates, or foreign objects, is found at that location, it is determined that there is such a location. This indicates the specific width and location of the block that was identified as an anomalous block.

[0074] Step S22: Record the location information of the abnormal mass. This refers to the position and size information of the agglomerate along the width of the belt, and the width of the agglomerate. Based on the distance between the zero-crossing points of H″(x), This indicates the location of the abnormal cluster along the belt width direction. The system will identify a nearby reference location and store the above information as product quality traceability data in the system log. At the same time, it will output an abnormal clumping alert signal to the human-machine interface to remind quality inspectors to pay attention to the incoming material quality of this batch.

[0075] VI. Belt Misalignment Detection. This embodiment further utilizes the three-dimensional perception capability of a 3D TOF camera to achieve real-time monitoring of belt misalignment. The specific steps are as follows: Step S23: Based on the reference height surface obtained in step S1, extract the reference edge positions of both sides of the belt under no-load conditions using the Canny edge detection algorithm or the threshold segmentation method based on height jumps, and denot them as follows: and It is stored in pixel coordinates or world coordinates.

[0076] Step S24: While the conveyor belt is conveying materials, the 3D TOF camera simultaneously acquires real-time 3D point cloud data containing both the material and the belt. The data processing module then uses the same height threshold segmentation method to synchronously extract the real-time edge positions of both sides of the belt from the real-time point cloud. and .

[0077] Step S25: Compare the real-time edge position with the reference edge position, and calculate the left edge offset respectively. and right edge offset Take the larger or average value of the two as the belt misalignment. .

[0078] Step S26: If the belt misalignment ΔE ~ belt ~ exceeds the preset belt misalignment threshold This threshold is set based on the mechanical design parameters and allowable deviation range of the belt conveyor. It is set to 50% of the safe distance between the belt edge and the frame. If this value is exceeded, the belt is considered to be misaligned, and a belt misalignment alarm signal is output. This alarm signal serves two purposes: firstly, it alerts the operator through the human-machine interface; secondly, it can trigger the belt conveyor's automatic correction device, such as a hydraulic or pneumatic alignment roller assembly, to perform correction actions, effectively preventing belt edge wear, tearing, and even machine downtime.

[0079] VII. Multi-section average flow calculation. To further improve measurement accuracy and suppress single-section measurement noise caused by local unevenness or lint on the material surface, this embodiment preferably adopts a multi-virtual cross-section sampling and averaging strategy. The specific steps are as follows: Step S27: In step S4, multiple virtual cross-sections are set at equal intervals along the belt movement direction (i.e., the Y direction), each corresponding to a Y coordinate position. The spacing between adjacent sections is set as Δy, where Δy = 10 mm. This is determined based on the surface roughness of the material and the system's computing resources. The height profile data at each cross section is then obtained.

[0080] Step S28: Steps S5 and S6 respectively perform curve fitting and cross-sectional integration on each of the m virtual cross sections to obtain a set of cross-sectional area sequences. ,in, The number of virtual cross-sections, m = 5 to 10.

[0081] Step S29: Take the average value of the cross-sectional area sequence. Used in step S7 to calculate instantaneous volumetric flow rate. This represents the cross-sectional area of ​​the material calculated by integration at the i-th virtual cross-section, i.e. , Indicates instantaneous volumetric flow rate. This represents the real-time linear speed of the conveyor belt, acquired synchronously. This multi-section averaging strategy effectively smooths out single-section area measurement deviations caused by localized protrusions, depressions, or random lint obstruction on the material surface, significantly improving the stability and repeatability of flow detection.

[0082] Workflow: During actual operation, the system periodically executes the following parallel tasks based on the sampling frame rate of the 3D TOF camera: 3D point cloud acquisition and preprocessing, cross-sectional contour extraction and area integration, multi-section average flow calculation, closed-loop PID control output, and multi-dimensional synchronous monitoring of abnormal states such as blockage, deviation, and clumping. The results of each anomaly detection and flow data are centrally displayed and stored through a human-machine interface, and simultaneously interact with the production line's central control system in real time via an industrial communication network. After actual deployment and verification in a cigarette factory's tobacco processing workshop, the system described in this embodiment, while achieving high-precision flow detection, further enhances the automation control level and abnormal condition early warning capabilities of the tobacco processing production line, providing solid technical support for intelligent manufacturing and lean production in the tobacco processing line.

[0083] It is worth noting that, based on accurate flow detection, this invention further integrates closed-loop control functionality. By adjusting the PID control of the upstream feeding equipment, it effectively suppresses flow fluctuations caused by uneven material delivery, ensuring a stable and controllable material supply for each process in the silk-making line, and significantly improving the process accuracy of subsequent key processes such as flavoring and additive addition. This invention achieves automatic detection and early warning of various abnormal operating conditions such as material shortages, blockages, material deviation, and belt misalignment through software algorithms. It can extract rich equipment operating status information from 3D point cloud data without additional hardware sensors, providing reliable data support for fault diagnosis and preventative maintenance of the production line, and significantly improving the intelligent operation and maintenance level of the silk-making production line. This invention effectively suppresses measurement noise caused by local unevenness on the material surface through a multi-virtual cross-section sampling and averaging strategy; and achieves non-contact automatic identification of abnormal material clumps through curvature mutation detection based on the second derivative, providing a new technical means for incoming material quality control.

[0084] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant teachings of the above embodiments, those skilled in the art can make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration, characterized in that, Includes the following steps: Step 1: With the conveyor belt unloaded, obtain the reference height plane of the belt using a 3D TOF camera fixedly installed above the conveyor belt; wherein, the optical axis of the 3D TOF camera is perpendicular to the bearing plane of the conveyor belt; Step 2: While the conveyor belt is conveying materials, acquire real-time 3D point cloud data containing the materials and the belt using a 3D TOF camera; Step 3: Subtract the real-time 3D point cloud data from the reference height surface to generate a pure height point cloud of the material, and filter the pure height point cloud of the material to remove outlier noise, thereby obtaining the filtered pure height point cloud of the material. Step 4: Based on the filtered pure height point cloud of the material, set multiple virtual cross-sections perpendicular to the direction of belt movement at equal intervals along the belt movement direction, thereby extracting the height contour data of the material in each virtual cross-section. Step 5: Perform curve fitting on the height profile data to generate a continuous height profile curve; Step 6: Integrate the height profile curve to obtain a set of material cross-sectional area sequences at the virtual cross-section. ,in, Indicates the first The cross-sectional area of ​​the material, The number of virtual cross sections; Step 7: Calculate the average cross-sectional area based on the material's cross-sectional area and the synchronously acquired real-time belt speed. It is used to calculate the instantaneous volumetric flow rate of a material.

2. The method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration as described in claim 1, characterized in that, Step 6 includes: Along the width direction of the belt, the height value is scanned from both sides of the material towards the center, and the position where the first height value exceeds the preset noise threshold is determined as the lower limit and upper limit of integration, respectively. Within the integration interval formed by the lower and upper limits of integration, the height profile curve is subjected to definite integral operation, and the resulting integral value is the cross-sectional area of ​​the material.

3. The method for detecting tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 1, characterized in that, Also includes: Step 8: Multiply the instantaneous volumetric flow rate by the pre-calibrated material density to obtain the instantaneous mass flow rate of the material.

4. A method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration, characterized in that, It is based on the flow detection method of claim 3, and includes: Step 9: Compare the instantaneous volumetric flow rate or instantaneous mass flow rate with the preset flow rate setting value, and calculate the flow rate deviation value; Step 10: Based on the flow deviation value, generate an adjustment control signal using a PID control algorithm, and output the adjustment control signal to the upstream feeding device of the tobacco processing line to adjust the discharge speed or discharge volume of the feeding device, thereby achieving closed-loop stable control of the tobacco leaf flow.

5. The method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, In step 10, the parameters of the PID control algorithm are automatically switched according to the current material flow rate change rate: When the rate of change of flow exceeds the preset rapid change threshold, a high proportional gain coefficient is used to improve the response speed. When the rate of change of flow is less than the preset steady change threshold, a low proportional gain coefficient is used to prevent overshoot oscillation.

6. The method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, It also includes material breakage detection: Continuously monitor the material cross-sectional area A or instantaneous mass flow rate Qm at the virtual cross-section; If the cross-sectional area A of the material detected in N consecutive frames is zero or less than the preset material breakage threshold, or the instantaneous mass flow rate Qm detected in N consecutive frames is zero or less than the preset material breakage mass threshold, then it is determined that the material is currently in a material breakage state, and a material breakage alarm signal is output; where N is the preset number of consecutive frames threshold.

7. The method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, It also includes material blockage detection: Continuously monitor the material height profile curve H(x) at the virtual cross-section; If the height value of the height profile curve H(x) at any position in the belt width direction exceeds the preset blockage height threshold for a duration exceeding the preset time threshold, or if the material cross-sectional area A exceeds the preset blockage area threshold for a duration exceeding the preset time threshold, then it is determined that the current state is blocked, and a blockage alarm signal and / or deceleration control signal are output to the belt drive mechanism.

8. The method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, It also includes material deviation detection: Step A: Within the virtual cross-section, based on the material height profile curve H(x), calculate the distribution center position C of the material in the belt width direction using the formula: In the formula, and These are the lower limit and upper limit of integration for the integration interval, respectively; Step B: Align the distribution center position C with the width center position of the belt. Compare and calculate the offset. ; Step C, if the offset If the deviation exceeds the preset tolerance threshold, it is determined that the material has deviated, and a deviation alarm signal is output. Simultaneously, based on the deviation amount... The direction and magnitude of the material are used to generate a correction control signal to drive the correction device to adjust the conveying posture of the material.

9. A method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, It also includes the detection of abnormal material clumps: Within the virtual cross-section, the second derivative H″(x) of the height profile curve H(x) is obtained; If there exists a certain position within the integration interval , making If the curvature change threshold is greater than the preset threshold, then the corresponding position is determined. An abnormal material clumping is detected, and the location and size information of the abnormal material clumping are recorded. At the same time, an abnormal clumping warning signal is output.

10. A method for controlling tobacco leaf flow rate in a tobacco processing line based on a 3D TOF camera and cross-sectional integration according to claim 4, characterized in that, It also includes belt misalignment detection: Based on the reference height surface obtained in step 1, extract the reference edge positions of the two sides of the belt under no-load conditions; While the conveyor belt is conveying materials, the real-time edge positions of both sides of the belt are simultaneously acquired using a 3D TOF camera. The real-time edge position is compared with the reference edge position to calculate the offset of the belt edge; If the offset exceeds the preset belt misalignment threshold, it is determined that the belt body has misaligned and a belt misalignment alarm signal is output.

Images