A method for detecting velocity and concentration of two-dimensional fluidized bed cut tobacco

Abstract

The present application belongs to the technical field of tobacco detection, and particularly relates to a two-dimensional fluidized bed tobacco speed and concentration detection method. Image data collected is labeled to construct a data set, the position of tobacco is similarity matched, the speed momentum is updated and iterated according to the displacement speed between two frames of tobacco, the detection of tobacco is completed, the speed of tobacco movement is calculated according to the position information of tobacco under the same label in combination with the collection frame rate, the visual angle range is divided into several regions from the horizontal and vertical directions, the tobacco in the regions is identified, gradient classified, and then the tobacco concentration distribution is calculated. The tobacco concentration and speed detection method controls the fluidization speed of tobacco stem separation through the positioning and speed detection of tobacco, so that the tobacco entering the subsequent section has a lower tobacco stem rate, less crushing, and ensures the product quality.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of tobacco detection, and particularly relates to a two-dimensional fluidized bed tobacco speed and concentration detection method. BACKGROUND

[0002] In the process of cigarette production, the quality of tobacco determines the quality of cigarettes. The separation of stems and tobacco is an important processing technology in the processing of tobacco. Larger stem particles can cause the puncture of cigarette paper, resulting in quality defects of cigarette products. Therefore, it is necessary to separate stems and tobacco in the reprocessing of tobacco.

[0003] The commonly used method at present is to separate leaf tobacco and stems by using a fluidized bed. The separation of stems and tobacco by using a fluidized bed has the characteristics of high efficiency and convenience. However, in the fluidized state, the different fluidization speeds will have a certain influence on the separation effect of stems and tobacco. When the speed is low, the separation effect of stems and tobacco is poor. When the speed is high, the leaf tobacco collides in the fluidized bed, thereby causing crushing and affecting the quality of tobacco. SUMMARY

[0004] The purpose of the application is to provide a two-dimensional fluidized bed tobacco speed and concentration detection method, which controls the fluidization speed of stem and tobacco separation by positioning and speed detection of tobacco, so that the tobacco entering the subsequent section has a lower stem content and less crushing, thereby ensuring the product quality.

[0005] To achieve the above purpose, the application is implemented by the following technical scheme:

[0006] A two-dimensional fluidized bed tobacco speed and concentration detection method, comprising the following steps:

[0007] S1, collecting image data of tobacco passing through a fluidized chamber;

[0008] S2, labeling part of the tobacco image, importing a deep learning model that needs to be trained, training the deep learning model, and automatically labeling the remaining tobacco image by using the trained model to construct a data set;

[0009] S3, using the first frame image as the starting image, and sequentially numbering the tobacco from left to right as tobacco i number recognized by the deep learning model, wherein i=1…n;

[0010] S4, setting an initial speed momentum value, predicting the position of the tobacco in the first frame image in the tobacco position in the second frame image, and performing similarity matching between the real tobacco position in the second frame image and the predicted position;

[0011] S5, if the similarity matching value exceeds the threshold value, it is considered that the matching is successful, and the tobacco with the maximum matching value is selected as the same tobacco number in the previous frame.

[0012] S6, if the similarity matching value is less than the set threshold value, it is considered that the matching fails, and the tobacco is skipped, and the matching value of the number of other tobaccos in the frame image is calculated;

[0013] S7, the displacement speed between the two frames of tobacco is calculated, the speed momentum is updated and iterated, and the detection process of steps S3 to S7 is sequentially dynamically iterated until the last tobacco detection is completed;

[0014] S8, after the above detection process is finished, the position information of the tobacco under the same label is calculated, and the speed of the tobacco motion is calculated combined with the frame rate of the collection;

[0015] S9, then according to the tobacco number, the tobacco image corresponding to the tobacco number in the image is sliced to obtain a single tobacco image;

[0016] S10, the collection angle range is divided into several regions from the horizontal and vertical directions respectively, and the divided regions are marked as Qi respectively, wherein i is a natural number;

[0017] S11, the area of Qi is calculated and marked as Mi;

[0018] S12, the tobaccos in the divided regions are identified respectively, and the tobacco particle size is gradient classified, and the number of tobacco particles of each size in each gradient is counted and marked as nij, wherein i and j are natural numbers;

[0019] S13, taking the tobacco particles of a set size as a reference, a tobacco particle size correction coefficient is calculated ;

[0020] S14, the tobacco concentration ;

[0021] S15, then according to the region label, the number of tobacco particles in each region is counted to obtain the tobacco concentration distribution.

[0022] Further, in step S1, the fluidized bed device is used to adjust the fluidization speed, and the image data of the tobacco passing through the fluidized chamber is collected by the industrial CCD camera.

[0023] Further, in step S2, part of the tobacco image is labeled by using X-AnyLabeling-GPU labeling tool.

[0024] Further, in step S4, the Kalman filter is used to predict the position of the tobacco in the first frame image in the second frame.

[0025] Further, in step S4, the similarity matching formula is:

[0026] Wherein, Si-1 is the similarity matching value at i-1 moment, Ki is the control input at i moment, and Zi is the observation value at i moment.

[0027] Further, in step S8, the speed of the tobacco is calculated in combination with the frame rate of the CCD camera.

[0028] Further, the final calculated speed value and the tobacco concentration distribution value of the tobacco are also saved.

[0029] The beneficial effects of the present application are:

[0030] The tobacco concentration and speed detection method of the present application controls the fluidization speed of the tobacco stem separation by positioning and speed detection of the tobacco, so that the tobacco entering the subsequent section has a lower tobacco stem rate and less crushing, thereby ensuring product quality. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 The detection flowchart of the present application.

[0032] Figure 2 The detection model structure diagram of the present application. DETAILED DESCRIPTION

[0033] The technical solutions of the present application will be described in detail below in combination with the drawings. The following examples are only exemplary and can only be used to explain and illustrate the technical solutions of the present application, but cannot be interpreted as a limitation of the technical solutions of the present application.

[0034] Referring to Figure 1 and Figure 2 , the present application provides a two-dimensional fluidized bed tobacco speed and concentration detection method, which comprises the following steps:

[0035] S1, collecting experimental samples, in the present application, the fluidization speed is adjusted by a fluidized bed device, and an industrial CCD camera is used to collect image data of the tobacco passing through the fluidization bin, and in other embodiments of the present application, all devices capable of collecting images can be used.

[0036] S2, using X-AnyLabeling-GPU labeling tool to label part of the tobacco images, in the present application, the part of the tobacco images is usually not less than 1 / 3 frames of the collected images within a set time, the deep learning model to be trained is imported, the deep learning model is trained, the remaining tobacco images are automatically labeled by using the trained model, and a data set is constructed.

[0037] S3, adopt the first frame image as the starting image, and sequentially number the cut tobacco from left to right according to the cut tobacco identified by the deep learning model, and the number is tobacco i number; the number of the cut tobacco is a sequential number, such as a natural number sequential number, i = 1…n.

[0038] S4, set an initial speed momentum value, predict the position of the cut tobacco in the first frame image in the second frame using the Kalman filter, and perform similarity matching between the actual cut tobacco position and the predicted position in the second frame image, and the similarity matching value S i The formula is: , wherein Si-1 is the similarity matching value at i-1 time, Ki is the control input at i time, and Zi is the observation value at i time.

[0039] S5, if the similarity matching value exceeds the threshold value, it is considered that the Kalman filter matching is successful, and the cut tobacco with the maximum matching value is selected as the same cut tobacco number in the previous frame.

[0040] S6, if the similarity matching value is less than the set threshold value, it is considered that the Kalman filter matching fails, and the cut tobacco is skipped, and the matching value of other cut tobacco numbers in the frame image is calculated.

[0041] S7, calculate the displacement speed between the two frames of cut tobacco, update and iterate the speed momentum in the Kalman filter, and sequentially dynamically iterate according to the detection process of steps S3 to S7 until the last cut tobacco is detected.

[0042] S8, after the above detection process is completed, the position information of the cut tobacco under the same label is calculated, and the speed of the cut tobacco is calculated in combination with the acquisition frame rate of the industrial CCD camera.

[0043] S9, then according to the cut tobacco label, the cut tobacco image in the image is sliced to obtain a single cut tobacco image, and the final calculated speed value of the cut tobacco is saved as a txt file.

[0044] S10, divide the view angle of the industrial CCD camera into several regions from the horizontal and vertical directions respectively, and mark the divided regions as Q i , wherein i = 0, 1, 2, 3, 4…

[0045] S11, calculate the area of Qi and mark it as M i .

[0046] S12, identify the cut tobacco in each divided region respectively, identify the cut tobacco particle size, and perform gradient classification on the cut tobacco particle size. Meanwhile, the number of cut tobacco particles of each size in each gradient is counted and marked as n ij . Wherein i = 0, 1, 2, 3, 4…, wherein j = 0, 1, 2, 3, 4…

[0047] S13, calculate the tobacco particle size correction coefficient based on a certain size .

[0048] S13, calculate the tobacco concentration .

[0049] S14, then according to the area label, count the number of tobacco particles in each area to obtain the tobacco concentration distribution, and save the final calculated concentration value of the tobacco as a txt file.

[0050] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application rather than limit them, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the present application, and these modifications or equivalent replacements also cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed, characterized in that, Includes the following steps: S1. Collect image data of tobacco shreds as they pass through the fluidization chamber; S2. Label some tobacco images, import the deep learning model to be trained, train the deep learning model, and use the trained model to automatically label the remaining tobacco images to build a dataset. S3. Using the first frame image as the starting image, the tobacco shreds identified by the deep learning model are numbered sequentially from left to right as tobacco shred number i, where i=1…n; S4. Set an initial velocity-momentum value, predict the position of the tobacco in the first frame image and the position of the tobacco in the second frame image, and perform similarity matching between the actual tobacco position in the second frame image and the predicted position. S5. If the similarity matching value exceeds the threshold, the match is considered successful, and the tobacco with the largest matching value is selected and marked as the same tobacco number in the previous frame. S6. If the similarity matching value is less than the set threshold, the matching is considered to have failed, and the tobacco shred is skipped. The matching value of other tobacco shred numbers in the frame image is calculated. S7. Calculate the displacement velocity between the two frames of tobacco, update and iterate the velocity momentum, and dynamically iterate in sequence according to the detection process from step S3 to step S7 until the last tobacco is detected. S8. After the above detection process is completed, calculate and extract the position information of the tobacco shreds under the same label, and calculate the speed of the tobacco shred movement in combination with the acquisition frame rate. S9. Then, based on the tobacco shred number, slice the tobacco shred image corresponding to the tobacco shred number in the image to obtain a single tobacco shred image. S10. Divide the field of view into several regions horizontally and vertically, and label each region as Qi, where i is a natural number. S11. Calculate the area of ​​Qi and label it as Mi; S12. Identify the tobacco shreds within the defined area, classify the tobacco shred particles by size gradient, and count the number of tobacco shred particles of each size within each gradient, labeled as nij, where i and j are natural numbers. S13. Calculate the tobacco particle size correction factor based on the set size of the tobacco particles. ; S14. Calculate the tobacco concentration. ; S15. Subsequently, based on the area labels, the number of tobacco particles in each area is statistically analyzed to obtain the tobacco concentration distribution.

2. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, In step S1, the fluidization speed is adjusted using a fluidized bed device, and an industrial CCD camera is used to collect image data of the tobacco as it passes through the fluidization chamber.

3. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, In step S2, the X-AnyLabeling-GPU annotation tool is used to annotate some of the tobacco images.

4. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, In step S4, the position of the tobacco shreds in the first frame image is predicted to be the position of the tobacco shreds in the second frame image using a Kalman filter.

5. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, In step S4, the similarity matching formula is: , among which, S i-1 K represents the similarity matching value at time i-1. i Z is the control input at time i. i Let represent the observation value at time i.

6. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, In step S8, the speed of the tobacco movement is calculated by combining the acquisition frame rate of the CCD camera.

7. The method for detecting the velocity and concentration of tobacco in a two-dimensional fluidized bed according to claim 1, characterized in that, It also includes the final calculated speed value and tobacco concentration distribution value for preserving the tobacco.

Images