A method for online measurement of multiple characteristics of tobacco using machine learning

Abstract

The present invention provides a method for measuring multiple characteristics of tobacco shreds online using machine learning, comprising the following steps: S1: obtaining tobacco shreds pictures under the camera respectively through a camera; S2: calculating the characteristics of tobacco shreds width, length, area, perimeter and the like in the picture respectively through a machine vision image algorithm; S3: when the method for measuring multiple characteristics of tobacco shreds online using machine learning provided by the present invention, respectively obtain the width value of tobacco shreds in a natural state and the width value of tobacco shreds in a flattened state. The method for measuring the length of tobacco shreds online in a natural state provided by the present invention can be statistically calculated by calculating the number of midline points of tobacco shreds. The method for measuring the perimeter of tobacco shreds online in a natural state provided by the present invention can calculate the perimeter of tobacco shreds by calculating the pixel points of the outline of tobacco shreds, and by setting the algorithm, the perimeter of tobacco shreds in a separated state can be calculated. According to the characteristic values ​​of tobacco shreds, the staff can grasp the on-site process quality in real time and respond in time.

Description

Technical Field

[0001] The present invention relates to the field of tobacco quality detection, and in particular to a method for online measuring multiple characteristics of tobacco using machine learning. Background Art

[0002] In the cigarette production process, the tobacco quality needs to be tested during the tobacco drying process, that is, the various characteristic values ​​of the tobacco need to be tested.

[0003] In the prior art, workers select 20 tobacco shreds, and the width of the selected tobacco shreds is very uniform, without irregular shapes such as bends and burrs. Then the tobacco shreds are neatly arranged on a cutting board, and then placed on the existing tobacco shred width measuring instrument in the factory to measure the tobacco shred width. Five points are usually randomly selected on a tobacco shred to measure the tobacco shred width, and then the measured tobacco shred width values ​​are averaged, which represents the width of the current tobacco shred, and the width of other tobacco shreds is measured in the same way.

[0004] However, the tobacco width measured by the tobacco width measuring instrument in this measurement method cannot fully represent whether the tobacco width of this batch meets the standard setting value. On the one hand, the number of tobaccos measured in the laboratory is limited, and on the other hand, the tobaccos measured in the laboratory are all extremely uniform. In addition, there are problems such as human errors in the manual operation process. For the entire batch, it is not enough to represent the tobacco width of the entire batch.

[0005] In the prior art, workers select a certain weight of tobacco shreds and put them into a vibrating screen machine with different mesh sizes for vibrating screen operation to separate long filaments, medium filaments, short filaments and broken filaments, so as to determine the long filament rate, medium filament rate, short filament rate and broken filament rate of this batch of tobacco shreds.

[0006] However, this measurement method cannot accurately separate long filaments, medium filaments, short filaments and broken filaments, because the curl shapes of tobacco shreds are different, and it is very likely that the long filaments will be screened out and mixed into the medium filaments due to the curl shape.

[0007] Therefore, it is necessary to provide a method for online measuring characteristic value tobacco length in a natural state to solve the above technical problems.

[0008] Among the existing measuring equipment in tobacco factories, there is no application specifically for calculating the area and circumference of tobacco.

[0009] Therefore, it is necessary to provide these two methods for online measuring the characteristic values ​​of cut tobacco area and circumference in the natural state to supplement the characteristic values ​​of cut tobacco. Summary of the invention

[0010] The present invention provides a method for measuring multiple characteristics of shredded tobacco online by using machine learning, which solves the problem that the current method for measuring characteristic values ​​of shredded tobacco has errors and is difficult to represent the characteristic values ​​of a whole batch of shredded tobacco.

[0011] In order to solve the above technical problems, the present invention provides a method for online measuring multiple characteristics of tobacco using machine learning, comprising the following steps:

[0012] S1: Obtain tobacco pictures under the camera through the camera;

[0013] S2: Calculate the width, length, area, perimeter and other characteristics of the tobacco in the image through machine vision image algorithms;

[0014] S3: When calculating the width, the tobacco width value in the natural state and the tobacco width value in the flattened state are obtained respectively, and the tobacco width value in the natural state and the tobacco width value in the flattened state are fitted and trained using a neural network model. After the network model fitting training, the tobacco width value in the natural state is predicted.

[0015] S4: When calculating the length, the center line of the tobacco is found, and the pixel points of the center line are calculated, and the length of the tobacco is calculated according to the pixel accuracy.

[0016] S5: When calculating the area, the outline of the tobacco is found according to the tobacco in the separated state, the internal outline pixels are calculated, and the area of ​​the tobacco is calculated.

[0017] S6: When calculating the perimeter, the contour of the tobacco is found according to the tobacco in the separated state, the number of pixels of the outer contour is calculated, and the perimeter of the tobacco is calculated.

[0018] The tobacco width value algorithm in the machine vision image algorithm specifically includes the following steps:

[0019] S31: separating the shredded tobacco by a shredded tobacco separation device;

[0020] S32: photographing the separated tobacco shreds with a camera, and performing width algorithm processing on the tobacco shreds within a field of view;

[0021] S33: extract the center line of the tobacco shreds, obtain the distance from a random point on the center line of the tobacco shreds to the edge of the tobacco shreds, set it as d1, that is, the width value of the tobacco shreds d2 = d1*2;

[0022] S34: traverse all points on the center line of the tobacco and take the average value as the width value of the tobacco.

[0023] The tobacco length value algorithm in the machine vision image algorithm specifically includes the following steps:

[0024] S41: separating the shredded tobacco by a shredded tobacco separation device;

[0025] S42: photographing the separated tobacco shreds with a camera, and performing length algorithm processing on the tobacco shreds within a field of view;

[0026] S43: extracting the center line of the tobacco shreds, obtaining all pixel points on the center line of the tobacco shreds, and calculating the length of the tobacco shreds according to the pixel accuracy of the camera.

[0027] The tobacco area value algorithm in the machine vision image algorithm specifically includes the following steps:

[0028] S51: separating the shredded tobacco by a shredded tobacco separation device;

[0029] S52: photographing the separated tobacco by a camera, and performing area algorithm processing on the tobacco within the field of view;

[0030] S53: extracting the outer contour of the tobacco shreds, obtaining all the outer contour points on a piece of tobacco shreds, calculating the area of ​​the inner pixels of the contour points, and calculating the area of ​​the tobacco shreds according to the pixel accuracy.

[0031] The tobacco circumference value algorithm in the machine vision image algorithm specifically includes the following steps:

[0032] S61: separating the shredded tobacco by a shredded tobacco separation device;

[0033] S62: photographing the separated tobacco by a camera, and performing perimeter algorithm processing on the tobacco within a field of view;

[0034] S63: extracting the outer contour of the tobacco shreds, obtaining all the outer contour points on a piece of tobacco shreds, calculating the number of contour points, and calculating the circumference of the tobacco shreds according to pixel accuracy.

[0035] Preferably, the tobacco separation device includes a support member, a separation box is connected to the upper side of the support member, a placement slot tube is provided at the lower center of the separation box, the bottom end of the placement slot tube is connected to a drive tube, one end of the drive tube is connected to an air intake pipe, and one end of the air intake pipe is connected to the output end of the fan.

[0036] Preferably, a driving mechanism is provided on one side of the driving tube, and a covering mechanism is provided on the separation box and on the side where the groove tube is placed, and the driving mechanism includes a mounting cover, and the mounting cover is fixed to one side of the driving tube, and the mounting cover is rotatably connected to a main shaft, and a driving impeller is fixedly connected to the surface of the main shaft and located inside the mounting cover, and one side of the driving impeller extends to the interior of the driving tube.

[0037] Preferably, the capping mechanism comprises a threaded shaft, the threaded shaft is rotatably connected to the mounting cover via a bearing seat, one end of the threaded shaft is fixedly connected to a driven bevel gear, one end of the main shaft is fixedly connected to a driving bevel gear, and the driving bevel gear is meshed with the driven bevel gear.

[0038] Preferably, the sealing mechanism further comprises a receiving groove, which is opened inside the lower side wall of the separation box, and one side of the receiving groove is connected to the interior of the placement groove tube, and a sealing block is arranged inside the receiving groove, and a vent hole is opened on the sealing block.

[0039] Preferably, a slide groove is provided at the bottom of the inner wall of the storage groove, a threaded sleeve is threadedly connected to the surface of the threaded shaft, a connecting frame is fixedly connected to the upper side of the threaded sleeve, and the upper end of the connecting frame passes through the slide groove and is fixedly connected to the sealing block.

[0040] Preferably, the support member includes a support seat, a strip groove is provided on the top of the support seat, a sliding rod is fixedly connected to the inside of the strip groove, a sliding sleeve is slidably connected to the surface of the sliding rod, the upper side of the sliding sleeve is fixedly connected to the bottom of the separation box, an elastic member is sleeved on the surface of the sliding rod and located on one side of the sliding sleeve, and a vibration motor is fixedly installed on one side of the separation box.

[0041] Preferably, a side door is provided on one side of the separation box, and a material pushing mechanism is provided inside the separation box and on a side away from the side door via an adsorption sheet.

[0042] Preferably, the pushing mechanism comprises a pushing plate, a magnetic block is embedded and fixed on one side of the pushing plate, and a handle is arranged on the upper side of the pushing plate.

[0043] Compared with the related art, the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention has the following beneficial effects:

[0044] The present invention provides a method for online measuring multiple characteristic values ​​of tobacco in a natural state. By setting the algorithm, the actual width of the tobacco in a flattened state can be estimated based on the width of the tobacco in a naturally curled state on site, so that the staff can grasp the on-site process quality in real time and respond in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of the first embodiment of the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention;

[0046] Figure 2 for Figure 1 The step diagram of the machine vision width algorithm shown;

[0047] Figure 3 for Figure 1 The neural network framework diagram shown;

[0048] Figure 4 A flowchart of the second embodiment of the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention;

[0049] Figure 5 for Figure 4 The local structural schematic diagram shown;

[0050] Figure 6 for Figure 4 A schematic diagram of the structure of the support member shown;

[0051] Figure 7 This is a flowchart of the steps of the third embodiment of the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention.

[0052] Numbers in the figure:

[0053] 1. Support member, 101. Support seat, 102. Strip groove, 103. Sliding rod, 104. Elastic member,

[0054] 2. Separation box, 3. Vibration motor, 4. Placement slot pipe, 5. Drive pipe,

[0055] 6. Driving mechanism, 61. Mounting cover, 62. Main shaft, 63. Driving impeller, 64. Active bevel gear,

[0056] 7. Capping mechanism, 71. Threaded shaft, 72. Threaded sleeve, 73. Connecting frame, 74. Receiving groove, 75. Sealing block, 76. Bearing seat, 77. Driven bevel gear, 78. Slide groove,

[0057] 8. Box cover, 9. Fill light, 10. Camera, 11. Handle,

[0058] 12. Sliding sleeve,

[0059] 13. Intake pipe,

[0060] 14. Side door,

[0061] 15. Pushing mechanism, 151. Pushing plate, 152. Handle, 153. Magnetic block,

[0062] 16. Adsorption sheet. DETAILED DESCRIPTION

[0063] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0064] First embodiment

[0065] Please refer to Figure 1 , Figure 2 and Figure 3 ,in, Figure 1 A flowchart of the first embodiment of the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention; Figure 2 for Figure 1 The step block diagram of the machine vision image algorithm shown; Figure 3 for Figure 1 The neural network framework diagram shown in the figure uses machine learning to measure multiple characteristics of tobacco online, including the following steps:

[0066] S1: Obtain tobacco pictures under the camera through the camera;

[0067] S2: Calculate the width, length, area, perimeter, and type dimension of the tobacco in the image through machine vision image algorithms;

[0068] S3: When calculating the width, the tobacco width value in the natural state and the tobacco width value in the flattened state are obtained respectively, and the tobacco width value in the natural state and the tobacco width value in the flattened state are fitted and trained using a neural network model. After the network model fitting training, the tobacco width value in the natural state is predicted.

[0069] S4: When calculating the length, find the center line of the tobacco, calculate the pixel points of the center line, and calculate the length of the tobacco according to the pixel accuracy;

[0070] S5: When calculating the area, find the outline of the tobacco in the separated state, and calculate the area of ​​the tobacco according to the outline;

[0071] S6: When calculating the circumference, find the outline of the tobacco shreds according to the separated state, and calculate the circumference of the tobacco shreds according to the outline;

[0072] The tobacco width value algorithm in the machine vision image algorithm specifically includes the following steps:

[0073] S31: separating the shredded tobacco by a shredded tobacco separation device;

[0074] S32: photographing the separated tobacco shreds with a camera, and performing width algorithm processing on the tobacco shreds within a field of view;

[0075] S33: extract the center line of the tobacco shreds, obtain the distance from a random point on the center line of the tobacco shreds to the edge of the tobacco shreds, set it as d1, that is, the width value of the tobacco shreds d2 = d1*2;

[0076] S34: traverse all points on the center line of the tobacco and take the average value as the width value of the tobacco.

[0077] The tobacco length value algorithm in the machine vision image algorithm specifically includes the following steps:

[0078] S41: separating the shredded tobacco by a shredded tobacco separation device;

[0079] S42: photographing the separated tobacco shreds with a camera, and performing length algorithm processing on the tobacco shreds within a field of view;

[0080] S43: extracting the center line of the tobacco shreds, obtaining all pixel points on the center line of the tobacco shreds, and calculating the length of the tobacco shreds according to the pixel accuracy of the camera.

[0081] The tobacco area value algorithm in the machine vision image algorithm specifically includes the following steps:

[0082] S51: separating the shredded tobacco by a shredded tobacco separation device;

[0083] S52: photographing the separated tobacco by a camera, and performing area algorithm processing on the tobacco within the field of view;

[0084] S53: extracting the outer contour of the tobacco shreds, obtaining all the outer contour points on a piece of tobacco shreds, calculating the area of ​​the inner pixels of the contour points, and calculating the area of ​​the tobacco shreds according to the pixel accuracy.

[0085] The tobacco circumference value algorithm in the machine vision image algorithm specifically includes the following steps:

[0086] S61: separating the shredded tobacco by a shredded tobacco separation device;

[0087] S62: photographing the separated tobacco by a camera, and performing perimeter algorithm processing on the tobacco within a field of view;

[0088] S63: extracting the outer contour of the tobacco shreds, obtaining all the outer contour points on a piece of tobacco shreds, calculating the number of contour points, and calculating the circumference of the tobacco shreds according to pixel accuracy.

[0089] Method for online detection of multiple characteristic values ​​of tobacco: When measuring the width, the photos taken by the industrial camera are transferred to the computer, and the center line of each tobacco strand is calculated using the traditional machine vision algorithm. The distance from the center line of the tobacco strand to the edge of the tobacco strand is then calculated based on the center line, and then multiplied by two to obtain the width of each tobacco strand and keep a record. The average value of all widths is calculated as the final width result;

[0090] After that, each natural tobacco is flattened and marked so that the natural tobacco and the flattened tobacco correspond to each other. The width of the flattened tobacco is calculated using the same traditional machine vision algorithm and recorded. Figure 2 shown.

[0091] Next, we will start to collect a lot of data, because we need to use machine learning methods to find the numerical correspondence between the width of the same tobacco in its natural state and the width of the tobacco in its flattened state. We need to collect a lot of data to train the machine learning model.

[0092] After the model is trained, it is tested with data that has not participated in the training to determine the effect of the model training. If the test data results of the model are too different from the training data results, the training model or training data is adjusted and training is performed again until the model test data results meet the requirements;

[0093] After the model is successfully trained, we can know the corresponding relationship between the width of the tobacco in the naturally curled state and the width of the tobacco in the flattened state. Based on this corresponding relationship, we can roughly infer the width of the tobacco in the flattened state based on the width of the tobacco in the natural state, thus achieving the ultimate goal.

[0094] When measuring the length, the photos taken by the industrial camera are transferred to the computer, and the traditional machine vision algorithm is used to find the center line of each tobacco strand. The length of each tobacco strand is then calculated based on the number of pixels in the center line and the accuracy of the pixels.

[0095] When measuring the circumference of tobacco, the photos taken by the industrial camera are transferred to the computer, and the traditional machine vision algorithm is used to find the outline of each tobacco strand. The circumference of the tobacco strand is then calculated based on the number of outline points of the tobacco strand.

[0096] When measuring the area, the photos taken by the industrial camera are transferred to the computer, and the traditional machine vision algorithm is used to find the outline of each tobacco strand. Then the area of ​​the tobacco strand is calculated based on the number of pixels inside the outline of the tobacco strand.

[0097] Compared with the related art, the method for online measuring multiple characteristics of shredded tobacco using machine learning provided by the present invention has the following beneficial effects:

[0098] The method for measuring multiple characteristics of tobacco shreds online using machine learning provided by the present invention can estimate the actual width in the flattened state based on the width of the tobacco shreds in the natural curling state on site by setting the algorithm; the method for measuring the length of tobacco shreds online in the natural state provided by the present invention can calculate the length of tobacco shreds by setting the algorithm. The method for measuring the area of ​​tobacco shreds online in the natural state provided by the present invention can calculate the area of ​​tobacco shreds in the separated state by setting the algorithm. The method for measuring the circumference of tobacco shreds online in the natural state provided by the present invention can calculate the circumference of tobacco shreds in the separated state by setting the algorithm. This allows staff to grasp the on-site process quality in real time and respond in a timely manner.

[0099] Second embodiment

[0100] Please refer to Figure 4 , Figure 5 and Figure 6 Based on the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the first embodiment of the present application, the second embodiment of the present application proposes another method for online measuring multiple characteristic values ​​of tobacco in a natural state. The second embodiment is only a preferred method of the first embodiment, and the implementation of the second embodiment will not affect the independent implementation of the first embodiment.

[0101] Specifically, the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the second embodiment of the present application is different in that, in the method for online measuring multiple characteristic values ​​of tobacco in a natural state,

[0102] The tobacco separation device includes a support member 1, a separation box 2 is connected to the upper side of the support member 11, a placement slot tube 4 is arranged at the lower center of the separation box 2, the bottom end of the placement slot tube 4 is connected to a drive tube 5, one end of the drive tube 5 is connected to an air intake pipe 13, and one end of the air intake pipe 13 is connected to the output end of the fan.

[0103] The fan is a blower, and preferably four support members 1 are arranged at the four corners of the bottom of the separation box 2. A valve is arranged on the lower side of the slot tube 4, and a switch for closing the blower is embedded on one side of the inner wall of the slot tube 4.

[0104] A driving mechanism 6 is provided on one side of the driving tube 5, and a covering mechanism 7 is provided on the separation box 2 and on the side where the groove tube 4 is placed. The driving mechanism 6 includes a mounting cover 61, and the mounting cover 61 is fixed to one side of the driving tube 5. The upper portion of the mounting cover 61 is rotatably connected to a main shaft 62, and a driving impeller 63 is fixedly connected to the surface of the main shaft 62 and located inside the mounting cover 61, and one side of the driving impeller 63 extends to the interior of the driving tube 5.

[0105] The switch is located at a side away from the capping mechanism 7, and the penetration point between the main shaft 62 and the mounting cover 61 is mechanically sealed.

[0106] The capping mechanism 7 includes a threaded shaft 71, which is rotatably connected to the mounting cover 61 via a bearing seat 76, one end of the threaded shaft 71 is fixedly connected to a driven bevel gear 77, one end of the main shaft 62 is fixedly connected to a driving bevel gear 64, and the driving bevel gear 64 is meshed with the driven bevel gear 77.

[0107] The sealing mechanism 7 also includes a receiving groove 74, which is opened inside the lower side wall of the separation box 2, and one side of the receiving groove 74 is connected to the interior of the placement groove tube 4. A sealing block 75 is arranged inside the receiving groove 74, and a ventilation hole is opened on the sealing block 75.

[0108] By opening the air holes, when the sealing block 75 has not completely sealed the slot for placing the slot tube 4, the air flow can flow into the interior of the separation box 2 through the air holes, wherein the diameter of the air holes is much smaller than the length of the tobacco shreds, wherein the length of the receiving slot 74 is longer than the sealing block 75, and when the end of the sealing block 75 moves into the receiving slot 74, the air flow has blown the tobacco into the interior of the separation box 2.

[0109] A slide groove 78 is provided at the bottom of the inner wall of the receiving groove 74 , a threaded sleeve 72 is threadedly connected to the surface of the threaded shaft 71 , a connecting frame 73 is fixedly connected to the upper side of the threaded sleeve 72 , and the upper end of the connecting frame 73 passes through the slide groove 78 and is fixedly connected to the sealing block 75 .

[0110] The support member 1 includes a support seat 101, a strip groove 102 is opened on the top of the support seat 101, a sliding rod 103 is fixedly connected inside the strip groove 102, a sliding sleeve 12 is slidably connected to the surface of the sliding rod 103, the upper side of the sliding sleeve 12 is fixedly connected to the bottom of the separation box 2, an elastic member 104 is sleeved on the surface of the sliding rod 103 and located on one side of the sliding sleeve 12, and a vibration motor 3 is fixedly installed on one side of the separation box 2.

[0111] By providing an elastic member 104, the separation box 2 can be movably connected, and when the vibration motor 3 vibrates, the separation box 2 can be driven to vibrate. The elastic member 104 is preferably a spring;

[0112] A box cover 8 is arranged on the top of the separation box 2, a fill light 9 and a camera 10 are arranged on the lower side of the box cover 8, and an observation window is arranged on the box cover 8.

[0113] The working principle of the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the present invention is as follows:

[0114] First, open the box cover 8, place the tobacco to be photographed in the placement slot tube 4, then close the box cover 8, turn on the fan and vibration motor 3, and open the valve on the placement slot tube 4 while turning on the fan;

[0115] The fan drives the airflow into the driving tube 5 through the air inlet pipe 13, and then the airflow enters the interior of the separation box 2 through the placement groove tube 4. The airflow can blow up the tobacco, so that the tobacco can be scattered into the interior of the separation box 2;

[0116] The vibration motor 3 can also drive the entire separation box 2 to vibrate horizontally, thereby helping the tobacco to be more evenly distributed at the bottom of the separation box 2;

[0117] At the same time, when the airflow passes through the driving pipe 5, it drives the impeller 63 to rotate, which drives the impeller 63 to drive the main shaft 62 to rotate, the main shaft 52 drives the active bevel gear 64 to rotate, the active bevel gear 64 drives the driven bevel gear 77 to rotate, the driven bevel gear 77 drives the threaded shaft 71 to rotate, the threaded shaft 71 drives the threaded sleeve 72 to rotate, and the threaded sleeve 72 drives the sealing block 75 to move to the inside of the slotted pipe 4 through the connecting frame 73;

[0118] The mouth of the placement groove tube 4 is gradually sealed. When the sealing block 75 completely seals the placement groove tube 4, one end of the sealing block 75 touches the switch located on one side of the inner wall of the placement groove tube 4, and the fan is turned off.

[0119] By providing the sealing block 75, it is possible to prevent some tobacco that has not been scattered outside from entering the interior of the placement groove tube 4 again;

[0120] When the particles are evenly dispersed, the vibration motor 3 is turned off;

[0121] The fill light 9 can be turned on, and the camera 10 can be used for photo processing;

[0122] The threaded shaft 71 can be manually rotated in the reverse direction to allow the sealing block 75 to enter the receiving groove 74 again.

[0123] Compared with the related art, the method for online measuring multiple characteristics of shredded tobacco using machine learning provided by the present invention has the following beneficial effects:

[0124] The tobacco inside the placement slot tube 4 is blown up and dispersed by setting an ascending airflow, and the separation box 2 is driven to vibrate by the vibration motor 3, so that the tobacco can be more evenly distributed at the bottom of the inner wall of the separation box 2;

[0125] By setting up the sealing mechanism 7, after the airflow blows the tobacco out of the placement slot tube 4, the airflow drives the impeller 63 to rotate, thereby driving the sealing block 75 in the sealing mechanism 7 to gradually seal the mouth of the placement slot tube 4, thereby preventing a small amount of tobacco from entering the placement slot tube 4 again when the fan is turned off. At the same time, the airflow is used as power, and there is no need to set up additional power, which is more energy-saving and environmentally friendly.

[0126] Third embodiment

[0127] Please refer to Figure 7 Based on the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided in the second embodiment of the present application, the third embodiment of the present application proposes another method for online measuring multiple characteristic values ​​of tobacco in a natural state.

[0128] Specifically, the difference between the method for online measuring multiple characteristic values ​​of tobacco in a natural state provided by the third embodiment of the present application is that, in the method for online measuring multiple characteristic values ​​of tobacco in a natural state, a side door 14 is provided on one side of the separation box 2, and a pushing mechanism 15 is provided on the side inside the separation box 2 away from the side door 14 through an adsorption sheet 16.

[0129] The pushing mechanism 15 includes a pushing plate 151 , a magnetic block 153 is embedded and fixed on one side of the pushing plate 151 , and a handle 152 is disposed on the upper side of the pushing plate 151 .

[0130] The side door 14 is provided with a lock core, and is connected to the separation box 2 through the lock core;

[0131] The adsorption sheet 16 is a metal sheet that can be adsorbed by a magnet, preferably made of stainless steel. The magnetic blocks 153 are preferably provided in two groups, symmetrically arranged on both sides of the push plate 151, with two in each group;

[0132] When the tobacco is photographed, the side door 14 can be opened, and the handle 152 can be used to drive the push plate 151 to clean the tobacco at the bottom of the separation box 2 out of the side door. The cleaning operation is simple.

[0133] After cleaning, the magnetic block 153 on one side of the push plate 151 can be adsorbed and limited by the adsorption sheet 16 .

[0134] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for online measurement of multiple characteristics of tobacco using machine learning, characterized in that: The following steps are involved: S1: Obtain tobacco pictures under the camera through the camera; S2: Calculate the width, length, area, and perimeter features of the tobacco in the image through machine vision image algorithms; S3: When calculating the width, the width of the tobacco in the natural state and the width of the tobacco in the flattened state are obtained respectively, and the width of the tobacco in the natural state and the width of the tobacco in the flattened state are fitted and trained using a neural network model, and the width of the tobacco in the natural state is predicted after the network model fitting training; S4: When calculating the length, find the center line of the tobacco, calculate the pixel points of the center line, and calculate the length of the tobacco according to the pixel accuracy; S5: When calculating the area, find the outline of the tobacco in the separated state, and calculate the area of ​​the tobacco according to the pixel points inside the outline; S6: When calculating the circumference, the outline of the tobacco is found according to the separated tobacco, and the circumference of the tobacco is calculated according to all the pixel points of the outline. The tobacco width value algorithm in the machine vision image algorithm specifically includes the following steps: S31: separating the shredded tobacco by a shredded tobacco separation device; S32: photographing the separated tobacco shreds with a camera, and performing width algorithm processing on the tobacco shreds within a field of view; S33: extract the center line of the tobacco shreds, obtain the distance from a random point on the center line of the tobacco shreds to the edge of the tobacco shreds, set it as d1, that is, the width value of the tobacco shreds d2=d1*2; S34: traverse all points on the center line of the tobacco, and take the average value as the width value of the tobacco; The tobacco length value algorithm in the machine vision image algorithm specifically includes the following steps: S41: separating the shredded tobacco by a shredded tobacco separation device; S42: photographing the separated tobacco shreds with a camera, and performing length algorithm processing on the tobacco shreds within a field of view; S43: extracting the center line of the tobacco, obtaining all pixel points on the center line of the tobacco, and calculating the length of the tobacco according to the pixel accuracy of the camera; The tobacco area value algorithm in the machine vision image algorithm specifically includes the following steps: S51: separating the shredded tobacco by a shredded tobacco separation device; S52: photographing the separated tobacco by a camera, and performing area algorithm processing on the tobacco within the field of view; S53: extracting the outer contour of the tobacco, obtaining all the outer contour points on a piece of tobacco, calculating the area of ​​the inner pixels of the contour points, and calculating the area of ​​the tobacco according to the pixel accuracy; The tobacco circumference value algorithm in the machine vision image algorithm specifically includes the following steps: S61: separating the shredded tobacco by a shredded tobacco separation device; S62: photographing the separated tobacco by a camera, and performing perimeter algorithm processing on the tobacco within a field of view; S63: Extract the outer contour of the tobacco, obtain all the outer contour points on a piece of tobacco, calculate the number of contour points, and calculate the circumference of the tobacco according to pixel accuracy. The tobacco separation device includes a support member, and the upper side of the support member is connected to a separation box. A placement groove tube is provided at the lower center of the separation box. The bottom end of the placement groove tube is connected to a driving tube, one end of the driving tube is connected to an air intake pipe, and one end of the air intake pipe is connected to the output end of the fan.

2. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 1, characterized in that: A driving mechanism is provided on one side of the driving tube, and a covering mechanism is provided on the separation box and on the side where the groove tube is placed. The driving mechanism includes a mounting cover, which is fixed to one side of the driving tube. A main shaft is rotatably connected to the upper portion of the mounting cover, and a driving impeller is fixedly connected to the surface of the main shaft and located inside the mounting cover, and one side of the driving impeller extends to the interior of the driving tube.

3. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 2, characterized in that: The capping mechanism comprises a threaded shaft, which is rotatably connected to the mounting cover via a bearing seat, one end of the threaded shaft is fixedly connected to a driven bevel gear, one end of the main shaft is fixedly connected to a driving bevel gear, and the driving bevel gear is meshed with the driven bevel gear.

4. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 3, characterized in that: The capping mechanism also includes a receiving groove, which is opened inside the lower side wall of the separation box, and one side of the receiving groove is connected to the inside of the placement groove tube. A sealing block is arranged inside the receiving groove, and a vent hole is opened on the sealing block.

5. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 4, characterized in that: A sliding groove is provided at the bottom of the inner wall of the receiving groove, a threaded sleeve is threadedly connected to the surface of the threaded shaft, a connecting frame is fixedly connected to the upper side of the threaded sleeve, and the upper end of the connecting frame passes through the sliding groove and is fixedly connected to the sealing block.

6. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 1, characterized in that: The support member includes a support seat, a strip groove is provided on the top of the support seat, a sliding rod is fixedly connected to the inside of the strip groove, a sliding sleeve is slidably connected to the surface of the sliding rod, the upper side of the sliding sleeve is fixedly connected to the bottom of the separation box, an elastic member is sleeved on the surface of the sliding rod and located on one side of the sliding sleeve, and a vibration motor is fixedly installed on one side of the separation box.

7. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 1, characterized in that: A side door is arranged on one side of the separation box, and a material pushing mechanism is arranged on a side of the separation box away from the side door through an adsorption sheet.

8. The method for online measuring multiple characteristics of shredded tobacco using machine learning according to claim 7, characterized in that: The pushing mechanism comprises a pushing plate, a magnetic block is embedded and fixed on one side of the pushing plate, and a handle is arranged on the upper side of the pushing plate.

Images