A method for detecting and characterizing the inner stem of a cigarette.

Abstract

A method for detecting and characterizing internal stems in cigarettes is characterized by: digitizing the three-dimensional structure of the cigarette through industrial CT scanning; obtaining a two-dimensional image of the tobacco distribution inside the cigarette using filtered back projection technology; and then obtaining a three-dimensional reconstruction model of the cigarette using three-dimensional reconstruction technology. This model reflects the state of the tobacco and stems inside the cigarette. Finally, the diameter and length are quantitatively characterized to identify the stems and classify the stems and tobacco. The advantages of this invention compared to existing technologies are: the three-dimensional reconstruction method using CT scanning detects the length, diameter, and quantity of tobacco and stems in cigarettes, obtaining accurate and reliable results; this method can not only identify stems but also quantitatively characterize their diameter, providing guidance for cigarette quality evaluation and control, as well as the optimization and improvement of cigarette manufacturing process parameters. This method is a non-destructive testing method, which can effectively reduce losses during the testing process.

Description

Technical Field

[0001] This invention relates to the field of cigarette product testing technology, and in particular to a method for detecting and quantitatively characterizing the stem marks inside cigarettes using CT three-dimensional reconstruction. Background Technology

[0002] Stems and veins in tobacco leaves easily form stem fragments or clumps during the tobacco processing. These stem fragments must be removed before rolling to reduce the amount of stem fragments in the tobacco. Stem fragments rolled into cigarettes can cause uneven packing, affecting the cigarette's weight and draw resistance stability. They are also a major cause of cigarette punctures and leaks. Furthermore, stem fragments can cause bursting and popping during combustion, affecting the cigarette's combustibility, sensory quality, and consistent draw. Therefore, stem fragment control has always been a key focus and challenge in the process control of cigarette manufacturers.

[0003] The cigarette processing generally reduces the stem content in cigarettes and improves cigarette quality by controlling the stem content in the pounded and re-dried leaves, using air separation in the cigarette making process, and employing stem-removal systems in cigarette making machines. Industry research on stems mainly focuses on controlling the stem content in pounded and re-dried leaves and the research and development of stem separation or removal devices. Currently, the phenomenon of combustion-end bursting is largely under control, but punctures still exist, especially in slim cigarettes where the puncture rate is relatively high.

[0004] Currently, the detection method for stems inside cigarettes is generally manual sampling, which involves manually cutting open the cigarette with a blade to separate the tobacco and stems for visual inspection. This method is inefficient, unable to promptly and comprehensively detect the number of stems inside the cigarette, and is greatly affected by the operator's subjective factors, making it difficult to ensure accuracy. Furthermore, the cigarettes, having been cut open, are unusable, resulting in waste of raw tobacco. Therefore, researching an automated and accurate method for detecting stems in cigarettes is crucial to improving detection efficiency, reducing error rates and process losses, and is of great significance for quality inspection, evaluation, and control in tobacco processing. However, research on methods for detecting and quantitatively characterizing internal stems in cigarettes using CT three-dimensional reconstruction is still lacking.

[0005] Chinese Patent (202210179428.1) discloses a method for detecting cigarette stems based on X-ray vision. This patent utilizes an X-ray device to irradiate the object being tested and obtain a corresponding perspective image of the cigarette. It then uses a generative adversarial network to generate multiple sets of pseudo-labeled samples from the perspective image, filters these pseudo-labeled samples according to screening criteria, and adjusts the training network using the manually labeled samples. Finally, it uses the trained stem classification network to detect stems on the tested cigarette samples. Compared to the method proposed in this patent, there are three main differences: First, the technical means used are different; Chinese Patent (202210179428.1) uses an X-ray detection method, while this patent uses a CT three-dimensional reconstruction method. Secondly, the two methods differ. The Chinese patent (202210179428.1) uses X-ray irradiation of a standard sample to obtain a corresponding cigarette perspective image, then manually annotates the sample to adjust the training network; finally, it uses the trained stem classification network to compare and detect stems in the tested cigarette samples. This invention, however, uses filtered back-projection technology to obtain a two-dimensional image of the tobacco distribution inside the cigarette, then uses three-dimensional reconstruction technology to obtain a three-dimensional reconstruction model of the cigarette, and uses the reconstruction model to obtain the state of the tobacco and stems inside the cigarette. Thirdly, the functions differ. The Chinese patent (202210179428.1) detects the presence of stems in the cigarette through comparison; while the method of this invention can not only detect and determine whether a cigarette contains stems, but also quantitatively analyze the length and diameter of the stems, thus providing a basis for the quality evaluation and control of different types of cigarettes.

[0006] Chinese Patent (202111033497.3) discloses a method for identifying and detecting stems based on image processing, including: acquiring images of tobacco shreds; identifying stems in the tobacco shreds to obtain stems; establishing a fitting model of stem area and stem mass; and calculating the stem content rate in the tobacco shreds. This invention utilizes image processing technology to identify stems; and establishes a fitting model of stem area and mass, thereby calculating the stem content rate of the tobacco shreds. Compared with the method proposed in this patent, the two mainly differ in the following ways: First, the technical means used are different. Chinese Patent (202111033497.3) uses a conventional image acquisition method; this patent uses a CT three-dimensional reconstruction method. Secondly, the two methods differ. Chinese patent (202111033497.3) uses image processing technology to identify the stems and establish a fitting model of the stem area and mass, thus calculating the stem content of the tobacco. This invention uses filtered back-projection technology to obtain a two-dimensional image of the tobacco distribution inside the cigarette, and then uses three-dimensional reconstruction technology to obtain a three-dimensional reconstruction model of the cigarette, obtaining the state of the tobacco and stems inside the cigarette through the reconstruction model. Thirdly, the functions differ. Chinese patent (202111033497.3) calculates the stem content of the tobacco through a fitting model; while this invention's method performs quantitative analysis of the stems themselves, such as measuring length and diameter, thus providing a basis for the quality evaluation and control of different types of cigarettes.

[0007] In summary, the method for detecting and quantitatively characterizing internal stems in cigarettes using industrial CT 3D reconstruction technology, which digitizes the 3D structure of cigarettes through industrial CT scanning, obtains a 2D image of the tobacco distribution inside the cigarette through filtering back projection technology and noise filtering preprocessing technology, and then obtains a 3D reconstruction model of the cigarette through 3D reconstruction technology, can detect the state of tobacco and stems inside the cigarette through the model, and finally uses quantitative characterization of length and diameter to identify and quantify stems, thereby classifying stems. This method still lacks a basis for detecting and quantitatively characterizing internal stems in different types of cigarettes. Summary of the Invention

[0008] The purpose of this invention is to provide a method for detecting and quantitatively characterizing the internal stems of cigarettes based on industrial CT three-dimensional reconstruction. This method can solve the problems of low efficiency and high error rate of existing stem detection technologies. Furthermore, because it uses non-destructive testing technology, it can further improve the utilization efficiency of tobacco raw materials, thereby obtaining higher economic benefits.

[0009] The objective of this invention is achieved through the following technical solution:

[0010] A method for detecting and characterizing internal stems in cigarettes involves digitizing the three-dimensional structure of the cigarette through industrial CT scanning, obtaining a two-dimensional image of the tobacco distribution inside the cigarette using filtered back projection technology, and then obtaining a three-dimensional reconstruction model of the cigarette using three-dimensional reconstruction technology. This model reflects the state of the tobacco and stems inside the cigarette. Finally, the diameter and length are quantitatively characterized to identify the stems and classify the stems and tobacco. The specific steps are as follows:

[0011] 1) Place the cigarette to be tested on the stage of the industrial CT equipment, keep it vertical with clamps and ensure that the tobacco part is within the CT scanning range;

[0012] 2) Start the industrial CT equipment to scan the cigarette, and the digital flat panel detector will transmit and save the received signal to the computer;

[0013] 3) The data processing system in the computer uses the Filtered Back Projection Algorithm (FDK algorithm) to process the acquired data and obtain a two-dimensional image of the distribution of tobacco shreds and stems inside the cigarette.

[0014] 4) The image processing system is used to perform image matching, image smoothing, and image enhancement on the 2D CT image sequence in order to perform edge extraction and image segmentation; interpolation is performed between adjacent 2D CT images to improve the accuracy of the 3D reconstruction model; finally, the 3D structural model is reconstructed using the image processing system.

[0015] 5) Select the median line of the volume profile of each tobacco shred and stem in the 3D reconstruction model of the cigarette. The selection principle is to select the median point of the width and thickness on each cross-section, and connect all the median points along the length to obtain the median line, and calculate the length of the median line.

[0016] 6) Calculate the length N of the median line of the tobacco shreds or stems. Take a point at one end of the median line with a step size of (N / 10), and take a plane that contains the point and is perpendicular to the median line. This plane intersects the outline of the tobacco shreds or stems to obtain the cross-sectional area.

[0017] 7) Using the median point of the cross-sectional area as the center, draw a circle with a radius of 1 pixel. Gradually increase the radius until the circle is externally tangent to the edge of the cross-sectional area, and record the diameter. Repeat this operation at other selected points, compare all diameters, and take the maximum value as the maximum diameter of the measured tobacco or stem.

[0018] 8) Measure the maximum diameter of all tobacco shreds and stems. Those with a median line length > 10.0 mm and a maximum diameter > 3.0 mm are considered stems; the rest are tobacco shreds. The determination of stem values ​​here is based on the definition of stems given by Henan Tobacco Company in its statistical analysis of stems in tobacco shreds.

[0019] 9) The cigarettes to be measured can be slim cigarettes, medium cigarettes, or regular cigarettes.

[0020] The cigarette imaging technology employs CT scanning imaging technology and three-dimensional reconstruction technology.

[0021] The advantages of this invention are: (1) The length, width and quantity of tobacco shreds and stems in cigarettes are detected by CT scanning three-dimensional reconstruction, and the results are real and reliable; (2) This method uses mature and reliable industrial technology, which is more efficient and accurate than manual identification of stems; (3) This method can not only identify stems, but also quantitatively characterize the diameter and length of stems, providing guidance for the quality evaluation and control of cigarettes, as well as the optimization and improvement of parameters of cigarette rolling process; (4) This method is a non-destructive testing method, which can effectively reduce the loss during the testing process and improve economic efficiency. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention;

[0023] Figure 2 A diagram of a cone-beam scanning structure with a flat panel detector;

[0024] In the figure, γ is the angle between the ray and the central ray, β is the angle formed by the central ray and the y-axis, called the projection angle, and κ is the cone angle in the Z-axis direction of the cone beam.

[0025] In the figure, the ab coordinate system is the coordinate system on the virtual detector. In this patent, a, b and a(x,y,β) and b(x,y,z,β) are the same, all representing the coordinate information on the virtual detector. Among them, x, y, z and β represent the expression factors of the ab coordinate system in the xyz coordinate system, that is, (a,b) can be converted to (x,y,z) and express the same spatial position information in different coordinate systems. Detailed Implementation

[0026] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0027] 1. Place the cigarette to be tested on the stage of the CT equipment, keep it vertical with clamps and ensure that the tobacco part is within the CT scan range; the cigarette sample to be tested needs to be placed in the equilibration chamber for temperature and humidity equilibration before the experiment. The specific values ​​are: temperature 22℃, relative humidity 65%, equilibration time of more than 24 hours.

[0028] 2. Start the industrial CT equipment to scan the cigarette. The digital flat panel detector transmits the received signals and saves them to the computer. Before scanning the cigarette, the industrial CT equipment needs to be pre-set with the following parameters: image size of 2048x2048, X-ray source tube voltage of 100kV, X-ray source tube current of 70uA, scanning thickness of 0.004mm, scanning interval of 0.004mm, CT scanning method of cone-beam scanning, and CT scanning mode of Normal scanning.

[0029] 3. The data processing system uses the Filtered Back Projection (FDK) algorithm to process the acquired data, obtaining a two-dimensional image of the tobacco distribution inside the cigarette. The FDK algorithm mainly includes several steps: pre-weighting of the projection data, one-dimensional filtering, and back projection.

[0030] (1) First, the projection data is weighted using a function similar to cosine, and the distance and angle difference between the voxel and the source point are appropriately corrected.

[0031] (2) Then, perform one-dimensional filtering in the horizontal direction on the projection data of different projection angles;

[0032] (3) Perform a weighted back projection of the filtered data using a cone-beam array. The weight function in the back projection depends on the distance from the reconstructed point to the focal point. See the attached diagram for a cone-beam scanning structure with a planar detector. Figure 2 .

[0033] The FDK reconstruction algorithm for flat panel detectors can be expressed by the following formula:

[0034]

[0035]

[0036] Where R represents the radius of rotation. Let g(a) represent the filtered projection data, g(a) be the filtering function, and U(x,y,β) represent the distance of the reconstructed pixel in the xy plane to the X-ray source.

[0037] Here, a and b represent the coordinates on the virtual detector.

[0038]

[0039] a(x,y,β) represents the horizontal position of the corresponding ray on the virtual detector, and the corresponding vertical position can be represented as...

[0040]

[0041] The weighted function can be decomposed into the following expression:

[0042]

[0043] Z is the Z-axis coordinate of the point to be reconstructed. Similarly, we can obtain the length of U based on geometric relationships.

[0044] U(x,y,β)=R+xcosβ+ysinβ

[0045] The weighting factor U is similar to the weighting factor in the two-dimensional filtering back projection algorithm. Geometrically, it is related to the projection of the line connecting the ray source point and the reconstructed image onto the intermediate ray.

[0046] 4. The image processing system in the computer performs image matching, smoothing, and enhancement operations on the 2D CT image sequence to facilitate edge extraction and image segmentation; it also interpolates between adjacent 2D CT images to improve the accuracy of the 3D reconstruction model; finally, the image processing system reconstructs the 3D structural model. In CT scan images, the spacing in the X and Y directions is equal during sampling, with a larger difference only in the Z direction. Therefore, linear interpolation is performed in the Z direction. The calculation formula is:

[0047]

[0048] In the formula, P1 and P2 are the CT values ​​of corresponding points in adjacent CT images, a1 and a2 are the Z-axis distances of the two corresponding points from the interpolation point, and P is the CT value of the interpolation point.

[0049] Multifractal spectrum technology is used to extract the internal and contour features of each cigarette CT image. The images are then analyzed using their visual appeal. The range of spectral values ​​in the multifractal spectrum theory is set, and the edges of the cigarette CT images are detected and extracted. The contour of the three-dimensional image is drawn using the edge information extracted from each cigarette CT image. The original cigarette CT image and the contour image extracted by using linear interpolation are then integrated to form a three-dimensional reconstruction model.

[0050] 5. Select the median line of the volume profile of each tobacco shred and stem in the 3D reconstruction model of the cigarette. The selection principle is to select the median point of the width and thickness on each cross-section, and connect all the median points along the length to obtain the median line, and calculate the length of the median line.

[0051] 6. Calculate the length N of the median line of the tobacco shreds or stems. Take a point at one end of the median line with a step size of (N / 10). Take a plane that contains the point and is perpendicular to the median line. This plane intersects the outline of the tobacco shreds or stems to obtain the cross-sectional area.

[0052] 7. Using the median point of the cross-sectional area as the center, draw a circle with a radius of 1 pixel. Gradually increase the radius until the circle is externally tangent to the edge of the cross-sectional area, and record the diameter. Repeat this operation at other selected points, compare all diameters, and take the maximum value as the maximum diameter of the measured tobacco or stem.

[0053] 8. Measure the maximum diameter of all tobacco shreds and stem sticks. Those with a median line length > 10.0 mm and a maximum diameter > 3.0 mm are stem sticks, and the rest are tobacco shreds.

Claims

1. A method for detecting and characterizing the stem residue inside a cigarette, characterized in that: The process involves digitizing the three-dimensional structure of a cigarette using industrial CT scanning. A two-dimensional image of the tobacco distribution inside the cigarette is obtained through filtered back-projection technology. Then, a three-dimensional reconstruction model of the cigarette is created using 3D reconstruction technology. This model reflects the state of the tobacco and stems inside the cigarette. Finally, the diameter and length are quantitatively characterized to identify the stems and classify the stems and tobacco. The specific steps are as follows: 1) Place the cigarette to be tested on the stage of the industrial CT equipment, keep it vertical with clamps and ensure that the tobacco part is within the CT scanning range; 2) Start the industrial CT equipment to scan the cigarette, and the digital flat panel detector will transmit and save the received signal to the computer; 3) The data processing system in the computer uses a filtering back projection algorithm to process the acquired data and obtain a two-dimensional image of the distribution of tobacco shreds and stems inside the cigarette. 4) The image processing system is used to perform image matching, image smoothing, and image enhancement operations on the 2D CT image sequence to facilitate edge extraction and image segmentation; interpolation is performed between adjacent 2D CT images to improve the accuracy of the 3D reconstruction model; finally, the 3D structural model is reconstructed using the image processing system. 5) Select the median line of the volume profile of each tobacco shred and stem in the 3D reconstruction model of the cigarette. The selection principle is to select the median point of the width and thickness on each cross-section, and connect all the median points along the length to obtain the median line, and calculate the length of the median line. 6) Calculate the length N of the median line of the tobacco shreds or stems. Take a point at one end of the median line with a step size of N / 10. Take a plane that contains the point and is perpendicular to the median line. This plane intersects the outline of the tobacco shreds or stems to obtain the cross-sectional area. 7) Draw a circle with the median point of the cross-sectional area as the center and a radius of 1 pixel. Gradually increase the radius until the circle is externally tangent to the edge of the cross-sectional area and record the diameter. Repeat this operation at other selected points, compare all diameters, and take the maximum value as the maximum diameter of the tobacco or stem shreds being measured. 8) Measure the maximum diameter of all tobacco shreds and stem sticks. Those with a median line length > 10.0 mm and a maximum diameter > 3.0 mm are stem sticks, and the rest are tobacco shreds.

2. The method for detecting and characterizing the inner stem of a cigarette according to claim 1, characterized in that: In step (1), the cigarette sample to be tested needs to be placed in a equilibration chamber for temperature and humidity equilibration before the experiment. The specific values ​​are: temperature 22℃, relative humidity 65%, and equilibration time of more than 24 hours.

3. The method for detecting and characterizing the inner stem of a cigarette according to claim 1, characterized in that: The CT equipment is set with the following parameters: image size of 2048x2048 for dedicated cigarette structure scanning, X-ray source tube voltage of 100 kV, X-ray source tube current of 70 uA, scanning thickness of 0.004 mm, scanning interval of 0.004 mm, CT scanning mode of cone-beam scanning, and CT scanning mode of Normal scanning.

4. The method for detecting and characterizing the inner stem of a cigarette according to claim 1, characterized in that: The cigarettes to be tested are slim cigarettes, medium cigarettes, or regular cigarettes.

5. The method for detecting and characterizing the inner stem of a cigarette according to claim 1, characterized in that: In step (3), the filtering back projection algorithm mainly includes several steps: pre-weighting of the projection data, one-dimensional filtering, and back projection. (1) First, the projection data is weighted using a function similar to cosine, and the distance and angle difference between the voxels and the source point are appropriately corrected; (2) Then, one-dimensional filtering is performed on the projection data at different projection angles in the horizontal direction; (3) Perform a weighted back projection of the filtered data using a cone beam. The weight function in the back projection depends on the distance from the reconstructed point to the focal point. The FDK reconstruction algorithm for flat panel detectors can be expressed by the following formula: ； ； Where R represents the radius of rotation. This represents the filtered projection data, where g(a) is the filtering function. () represents the distance of the reconstructed pixel in the xy plane to the X-ray source; ； This represents the horizontal position of the corresponding ray on the virtual detector, and the corresponding vertical position can be represented as... ； The weighted function can be decomposed into the following expression: ； Z is the Z-axis coordinate of the point to be reconstructed. Similarly, we can obtain the length of U based on geometric relationships. ； The weighting factor U is similar to the weighting factor in the two-dimensional filtering back projection algorithm. Geometrically, it is related to the projection of the line connecting the ray source point and the reconstructed image onto the intermediate ray.

6. The method for detecting and characterizing the inner stem of a cigarette according to claim 1, characterized in that: The specific process of step (4) is as follows: In the CT scan image, the spacing in the X and Y directions is equal during sampling, while the spacing in the Z direction has a larger difference. Therefore, a linear interpolation is performed in the Z direction; the calculation formula is: ； In the formula, P1 and P2 are the CT values ​​of corresponding points in adjacent CT images, a1 and a2 are the Z-axis distances of the two corresponding points from the interpolation point, and P is the CT value of the interpolation point.

Images