Roast duck cake conveyor based on visual online detection and classification

Abstract

The invention discloses a roast duck cake conveyor based on visual online detection and classification, and relates to the technical field of cake quality detection, the roast duck cake conveyor comprises a conveying mechanism 1, a cake guide mechanism 2 and a detection module 3, and the cake guide mechanism 2 performs corresponding work adjustment according to the detection result of the detection module 3. The roast duck cake online detection device can perform multi-dimensional comprehensive detection on the size, shape (roundness), thickness and the like of roast duck cakes, classifies unqualified products, inferior-quality products and qualified products, can improve the detection quality and reduce waste through fine sorting, is designed according to the principle of roast duck cake online detection and is combined with actual equipment, and the detection efficiency is improved. According to the invention, in the online detection process, orderly classification is realized in the conveying process, the automation is improved, and the detection efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of pancake quality inspection technology, and in particular to a Peking duck pancake conveyor based on visual online detection and classification. Background Technology

[0002] As a traditional food, Peking duck pancakes typically undergo a production process including dough mixing, shaping, baking or steaming, cooling, and packaging. In industrial production, after being formed by a shaping machine, the pancakes are conveyed to the subsequent packaging process via a conveyor belt. However, due to factors such as raw material ratios, equipment wear and tear, and process fluctuations, the finished products often exhibit quality defects such as inconsistent size, distorted shape (oval or broken), uneven thickness, and foreign matter adhering to the surface.

[0003] Currently, most Peking duck pancake production lines still rely on manual visual inspection for surface quality, supplemented by simple measuring tools for random sampling. This method has the following drawbacks: Inefficient and inaccurate: Manual inspection is difficult to match the speed of high-speed automated production lines, and long-term work can easily lead to visual fatigue, resulting in missed or false detections, especially for minute thickness changes and shape deviations that are difficult to quantify accurately. Lack of comprehensive grading standards: Existing testing methods can often only distinguish between "qualified" and "unqualified" products. They cannot accurately sort out "substandard" products that do not affect edibility based on appearance indicators such as size, shape, and thickness. This results in some products with poor appearance but are edible being directly discarded or mixed with qualified products, causing waste of resources or affecting the product image.

[0004] Therefore, designing a conveyor system capable of online, non-contact, multi-dimensional comprehensive detection of the size, shape (roundness), and thickness of Peking duck pancakes, and automatically performing multi-level classification based on the detection results, has become a pressing technical problem in this field. To address this, a Peking duck pancake conveyor based on visual online detection and classification is proposed. Summary of the Invention

[0005] In order to improve the existing roast duck pancake detection technology mentioned in the background technology, which is inefficient and lacks precision, cannot meet the needs of automated production lines, and lacks automatic comprehensive grading capabilities, some products with poor appearance but are edible are directly discarded or mixed with qualified products, resulting in resource waste or damage to the product image.

[0006] This invention provides a visual online detection and classification-based roast duck pancake conveyor, employing the following technical solution: A visual online detection and classification-based Peking duck pancake conveyor includes, The conveying mechanism is divided into a front conveying area A and a rear conveying area B according to the conveying process, and into a first channel I, a second channel II, a third channel III and a fourth channel IV according to the collection path of the roast duck pancakes; The pancake guiding mechanism is used to change the transportation channel during the conveying process of roast duck pancakes. It includes a path guiding component, a front non-conforming product guiding component, a front secondary product guiding component, a rear non-conforming product guiding component one, a rear non-conforming product guiding component two, and a rear secondary product guiding component located at the top of the conveying mechanism. The path guiding component, the front non-conforming product guiding component, and the front secondary product guiding component are located in the front conveying area A, while the rear non-conforming product guiding component one, the rear non-conforming product guiding component two, and the rear secondary product guiding component are located in the rear conveying area B. The detection module is used to detect whether the roast duck pancakes are qualified or defective during transportation on the conveyor mechanism. The qualified pancakes are divided into unqualified products and acceptable products after the unqualified product detection. The acceptable products are divided into substandard products and qualified products after the defective product detection. The detection module includes a front acceptable product detection component, a front qualified product detection component, a rear acceptable product detection component 1, a rear acceptable product detection component 2, and a rear qualified product detection component. The front acceptable product detection component, the front qualified product detection component, the rear acceptable product detection component 1, the rear acceptable product detection component 2, and the rear qualified product detection component are respectively set before the front unqualified product guide component, the front substandard product guide component, the rear unqualified product guide component 1, the rear unqualified product guide component 2, and the rear substandard product guide component, and the detection feedback is correspondingly sent to the matching guide component.

[0007] Preferably, the front receptive item detection component, the rear receptive item detection component one, and the rear receptive item detection component two utilize images acquired by a camera to detect foreign objects, colors, damage, receptive size, and receptive thickness. The pre-qualified product inspection component and the post-qualified product inspection component use cameras to acquire images based on defective product size inspection, roundness inspection, and defective product thickness inspection; Whether roast duck pancakes are acceptable is determined by whether there is a food safety risk or serious appearance defect. Whether roast duck pancakes are substandard is determined by the size of appearance defects while ensuring they are safe to eat.

[0008] Preferably, the foreign object detection uses a deep convolutional neural network to identify foreign objects and determine whether a roast duck pancake is acceptable. Color detection analyzes the hue (H) and saturation (S) distribution of the roast duck pancake area using the HSV color space, calculates the color difference with the standard pancake, and determines whether the roast duck pancake is acceptable. Damage detection calculates the circumscribed rectangle, area, and perimeter of the Peking duck pancake outline through contour analysis. Based on the area of ​​contour defects, a threshold 'a' is set to determine whether the Peking duck pancake is acceptable. Acceptable size detection calculates the equivalent diameter area of ​​the roast duck pancake, the equivalent circle diameter, the minimum circumcircle diameter, and the maximum incircle diameter using a binarized mask. A diameter threshold b is set to determine whether the roast duck pancake is acceptable. The thickness of the roast duck pancake can be measured by laser up and down through a through-beam method. A thickness threshold c is set to determine whether the roast duck pancake is acceptable. Defective product size inspection uses a binary mask to calculate the equivalent diameter area of ​​the roast duck pancake, the equivalent circle diameter, the minimum circumcircle diameter, and the maximum incircle diameter, and sets a diameter threshold d to determine whether the roast duck pancake is qualified. Roundness detection uses the defective product size detection value to calculate the roundness of the roast duck pancake, sets a roundness threshold e, and determines whether the roast duck pancake is qualified. The thickness inspection of defective products is the same as that of acceptable thickness inspection. The thickness of the roast duck pancake is measured by laser up and down through a beam, and a thickness threshold f is set to determine whether the roast duck pancake is qualified.

[0009] Preferably, the conveying mechanism includes two sets of parallel front frames and rear frames. Conveying rollers are provided at both ends of the front and rear frames. A front conveyor belt located on the front frame and a rear conveyor belt located on the rear frame are provided on the conveying rollers. The output ends of the front and rear conveyor belts are respectively provided with flip plate outer shafts at the front and rear frames. A flip plate assembly is provided in the flip plate outer shaft. The flip plate assembly is used to flip the roast duck pancake and move it away from the conveyor belt. A cover is provided on the top of the front and rear frames.

[0010] Preferably, the flip plate assembly includes a flipping shaft controlled by a flipping motor, which is connected to the outer shaft of the flip plate. A flipping claw is provided at the top of the flipping shaft, and a baffle is provided at the top of the flipping claw. The front conveyor belt and the rear conveyor belt are discontinuous according to the position of the flipping claw.

[0011] Preferably, the path guiding component is located in the second channel II in a trumpet shape, used to guide the roast duck pancakes to concentrate in the second channel II; The front non-conforming product guide assembly, the front secondary product guide assembly, the rear non-conforming product guide assembly 1, the rear non-conforming product guide assembly 2, and the rear secondary product guide assembly can be raised and lowered to guide the roast duck pancakes to change the conveying channel. The front non-conforming product guide assembly is used for roast duck pancakes from the second channel II to the first channel I, the front secondary product guide assembly is used for roast duck pancakes from the second channel II to the third channel III, the rear non-conforming product guide assembly 1 is used for roast duck pancakes from the second channel II to the first channel I, the rear non-conforming product guide assembly 2 is used for roast duck pancakes from the third channel III to the fourth channel IV, and the rear secondary product guide assembly is used for roast duck pancakes from the second channel II to the third channel III.

[0012] Preferably, the front defective product guiding component, the front secondary product guiding component, the rear defective product guiding component 1, the rear defective product guiding component 2, the rear secondary product guiding component, and the path guiding component of the dough guiding mechanism have the same composition structure, including a lifting column fixed to the top of the conveying mechanism, a main guide plate at the bottom of the lifting column, a telescopic rod in the inner cavity of the main guide plate, a movable guide plate at the output end of the telescopic rod, and the movable guide plate is slidably disposed in the main guide plate. The number of the above-mentioned structures in the front defective product guiding component, the front secondary product guiding component, the rear defective product guiding component 1, the rear defective product guiding component 2, and the rear secondary product guiding component is the same as one set, and the number of the above-mentioned structures in the path guiding component is two sets, and the overall structure is a gradually contracting trumpet shape along the conveying direction; Preferably, the control feedback of the telescopic poles in the front nonconforming product guide assembly, the front secondary product guide assembly, the rear nonconforming product guide assembly I, the rear nonconforming product guide assembly II, and the rear secondary product guide assembly is associated with the acceptable size detection or defective size detection in the corresponding front acceptable product detection assembly, the front acceptable product detection assembly, the rear acceptable product detection assembly I, the rear acceptable product detection assembly II, and the rear acceptable product detection assembly.

[0013] Preferably, the quality of the roast duck pancakes flowing out from the first channel I, the second channel II, the third channel III, and the fourth channel IV are respectively unqualified, qualified, acceptable substandard, and unqualified.

[0014] In summary, the present invention has the following beneficial effects: 1. This invention can comprehensively inspect the size, shape (roundness), and thickness of roast duck pancakes from multiple dimensions, classify unqualified, substandard, and qualified products, and improve inspection quality and reduce waste through refined sorting; 2. This invention presents the principle design for online detection of roast duck pancakes, and combines it with actual equipment to perfectly match the online detection process. During the transportation process, it achieves orderly classification, improves automation, and increases detection efficiency. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the detection principle of the roast duck pancake combined equipment of the present invention; Figure 2 For the present invention in Figure 1 Based on this, a flowchart of the detection principle is combined with the dough guiding mechanism and the detection module; Figure 3 This is a schematic diagram of the overall structure of the present invention; Figure 4 This is a schematic diagram of the conveying mechanism structure after the cover is removed according to the present invention; Figure 5 This is a top view of the detection range of the disc guide mechanism and the detection module behind the cover of the present invention. Figure 6 This is a view showing the extent of the front conveying area, rear conveying area, first channel, second channel, third channel, and fourth channel of the present invention. Figure 7 This is a schematic diagram of the structure of the flap outer shaft and flap assembly of the present invention; Figure 8 This is a cross-sectional view of the conveyor roller of the rear conveyor belt in this invention; Figure 9 This is a schematic diagram of the dough guiding mechanism of the present invention; Figure 10 This is a cross-sectional view of the main guide plate, telescopic rod, and movable guide plate in the dough guiding mechanism of the present invention; Figure 11 This is a flowchart illustrating the detection principle of the roast duck pancake of the present invention.

[0016] In the picture: 1. Conveying mechanism; 11. Front frame; 12. Rear frame; 13. Conveying roller; 14. Front conveyor belt; 15. Rear conveyor belt; 16. Outer shaft of the flip plate; 17. Flip plate assembly; 171. Tilting shaft; 172. Filting claw; 173. Baffle; 18. Cover; 2. Dough guiding mechanism; 21. Front non-conforming product guiding assembly; 22. Front secondary product guiding assembly; 23. Rear non-conforming product guiding assembly one; 24. Rear non-conforming product guiding assembly two; 25. Rear secondary product guiding assembly; 26. Path guiding assembly; 201. Lifting column; 202. Main guide plate; 203. Telescopic pole; 204. Moving guide plate; 3. Detection module; 31. Front acceptable product detection component; 32. Front qualified product detection component; 33. Rear acceptable product detection component one; 34. Rear acceptable product detection component two; 35. Rear qualified product detection component; A. Front conveying area; B. Rear conveying area; I. First Channel; II. Second Channel; III. Third Channel; IV. Fourth Channel. Detailed Implementation

[0017] The following is in conjunction with the appendix Figures 1-11 The present invention will be described in further detail below.

[0018] In the current testing process for roast duck pancakes produced in factories, the testing methods can often only distinguish between "qualified" and "unqualified" levels, and cannot carry out more refined sorting. As a result, some products that are not good in appearance but are edible are directly discarded or mixed with qualified products, causing waste of resources or affecting the product image.

[0019] Specific testing items generally include the following: Foreign object detection: Foreign objects such as hair and plastic pieces attached to the roast duck pancakes pose a serious food safety risk and are considered unqualified if present. Color inspection: If the roast duck pancakes have abnormal colors such as mold, charred black or green, they pose a serious food safety risk and are therefore unqualified. Damage detection: The degree of damage to the roast duck pancakes generally does not affect the safety of consumption. Based on the degree of damage, they can be divided into unqualified, substandard, and qualified products. Size inspection: Generally, this refers to the diameter of the roast duck pancake. Generally, its food safety is not affected. Based on the deviation of the diameter from the standard size range, it can be divided into unqualified, substandard, and qualified products. Thickness inspection: The thickness of the roast duck pancakes generally does not affect the safety of consumption (excessive thickness may lead to undercooking and food safety risks). Based on the deviation of the thickness from the standard range, they can be divided into unqualified, substandard, and qualified products. Roundness inspection: The roundness of the roast duck pancake does not affect the safety of consumption. Under the condition that the damage inspection and size inspection are qualified, the roundness basically only affects the appearance. Based on the deviation of the roundness from the standard range of size, it can be divided into defective products and qualified products.

[0020] Considering that Peking duck pancakes have two sides (side A and side B), and both sides need to be inspected during testing, a design was created as follows: Figure 11 The detection process shown is as follows: Based on the above problems and the design ideas of the detection process, this embodiment of the invention provides a Peking duck pancake conveyor based on visual online detection and classification, including a conveying mechanism 1, a pancake guiding mechanism 2 and a detection module 3. The pancake guiding mechanism 2 makes corresponding adjustments according to the detection results of the detection module 3.

[0021] The conveying mechanism 1 is divided into a front conveying area A and a rear conveying area B according to the conveying process, and into a first channel I, a second channel II, a third channel III and a fourth channel IV according to the collection path of the roast duck pancakes.

[0022] The pancake guiding mechanism 2 is used to change the transportation channel during the conveying process of roast duck pancakes. It includes a path guiding component 26, a front non-conforming product guiding component 21, a front secondary product guiding component 22, a rear non-conforming product guiding component one 23, a rear non-conforming product guiding component two 24, and a rear secondary product guiding component 25 located at the top of the conveying mechanism 1. The path guiding component 26, the front non-conforming product guiding component 21, and the front secondary product guiding component 22 are located in the front conveying area A, while the rear non-conforming product guiding component one 23, the rear non-conforming product guiding component two 24, and the rear secondary product guiding component 25 are located in the rear conveying area B.

[0023] The path guide component 26 is located in the second channel II in a trumpet shape, used to guide the roast duck pancakes to concentrate in the second channel II, ensuring that the roast duck pancakes are centered as much as possible in the second channel II. The front non-conforming product guide component 21, the front secondary product guide component 22, the rear non-conforming product guide component one 23, the rear non-conforming product guide component two 24, and the rear secondary product guide component 25 can be raised and lowered to guide the roast duck pancakes to change the conveying channel. The front non-conforming product guide component 21 is used for roast duck pancakes from the second channel II to the first channel I, the front secondary product guide component 22 is used for roast duck pancakes from the second channel II to the third channel III, the rear non-conforming product guide component one 23 is used for roast duck pancakes from the second channel II to the first channel I, the rear non-conforming product guide component two 24 is used for roast duck pancakes from the third channel III to the fourth channel IV, and the rear secondary product guide component 25 is used for roast duck pancakes from the second channel II to the third channel III.

[0024] The detection module 3 is used to detect whether the roast duck pancakes are qualified or defective during transportation on the conveyor mechanism 1. In this application, the roast duck pancakes are divided into unqualified products and acceptable products after the qualification test. The acceptable products are divided into substandard products and qualified products after the defective product test. The detection module 3 includes a front acceptable product detection component 31, a front qualified product detection component 32, a rear acceptable product detection component one 33, a rear acceptable product detection component two 34, and a rear qualified product detection component 35. The front acceptable product detection component 31, the front qualified product detection component 32, the rear acceptable product detection component one 33, the rear acceptable product detection component two 34, and the rear qualified product detection component 35 are respectively set before the front unqualified product guide component 21, the front substandard product guide component 22, the rear unqualified product guide component one 23, the rear unqualified product guide component two 24, and the rear substandard product guide component 25, respectively. Figures 5-6 As shown. In addition, it should be noted that, in addition to the measuring equipment required for each inspection item, the front acceptable product inspection component 31, the front qualified product inspection component 32, the rear acceptable product inspection component one 33, the rear acceptable product inspection component two 34, and the rear qualified product inspection component 35 also need to have a camera for acquiring images and a trigger sensor. The trigger sensor can be a laser through-beam sensor, which takes an image when the roast duck pancake passes through the trigger sensor.

[0025] For details, please refer to Figures 1-2 Guided by the principles of the detection process and combined with actual equipment, the process for conveying the roast duck pancakes in this application is as follows: S1. After being made by the forming machine, the roast duck pancake is placed directly or indirectly into the second channel II, and the path guide component 26 roughly centers the roast duck pancake.

[0026] S2. The roast duck pancake is inspected by the pre-acceptable product detection component 31. If it is acceptable, it continues to stay in the second channel II. If it is unacceptable, it enters the first channel I through the pre-defective product guide component 21.

[0027] After S3 and S2, the roast duck pancakes remaining in the second channel II are inspected by the pre-qualified product inspection component 32. If they pass the inspection, they remain in the second channel II. If they fail, they pass through the pre-secondary product guide component 22 and enter the third channel III. After that, the roast duck pancakes in the second channel II and the third channel III are flipped over.

[0028] After S4 and S3, the roast duck pancakes in the second channel II are inspected by the acceptable product detection component 1 33. If they are acceptable, they remain in the second channel II; if they are unacceptable, they pass through the non-conforming product guide component 1 23 and enter the first channel I. After S3, the roast duck pancakes in the third channel III are inspected by the acceptable product detection component 2 34. If they are acceptable, they remain in the second channel III; if they are unacceptable, they pass through the non-conforming product guide component 2 24 and enter the fourth channel IV.

[0029] The roast duck pancakes remaining in the second channel II after S5 and S4 are inspected by the qualified product inspection component 35. If they pass the inspection, they remain in the second channel II; otherwise, they pass through the secondary product guide component 25 and enter the third channel III.

[0030] S6. Collect the roast duck pancakes left on the first channel I, the second channel II, the third channel III, and the fourth channel IV, which correspond to defective products, qualified products, acceptable substandard products, and defective products, respectively.

[0031] Specifically, the front acceptable product inspection component 31, the rear acceptable product inspection component 1 33 and the rear acceptable product inspection component 2 34 use cameras to acquire images based on foreign object detection, color detection, damage detection, acceptable size detection and acceptable thickness detection, while the front qualified product inspection component 32 and the rear qualified product inspection component 35 use cameras to acquire images based on defective product size detection, roundness detection and defective product thickness detection.

[0032] Among them, whether roast duck pancakes are acceptable is determined by whether there is a food safety risk or serious appearance defects, and whether roast duck pancakes are substandard is determined by the size of appearance defects under the condition that they are safe to eat.

[0033] Foreign object detection uses a deep convolutional neural network to identify foreign objects and determine whether a roast duck pancake is acceptable. Color detection analyzes the hue (H) and saturation (S) distribution of the roast duck pancake area using the HSV color space, calculates the color difference with the standard pancake, and determines whether the roast duck pancake is acceptable. Damage detection calculates the circumscribed rectangle, area, and perimeter of the Peking duck pancake outline through contour analysis. Based on the area of ​​contour defects, a threshold 'a' is set to determine whether the Peking duck pancake is acceptable. Acceptable size detection calculates the equivalent diameter area of ​​the roast duck pancake, the equivalent circle diameter, the minimum circumcircle diameter, and the maximum incircle diameter using a binarized mask. A diameter threshold b is set to determine whether the roast duck pancake is acceptable. The thickness of the roast duck pancake can be measured by laser up and down through a through-beam method. A thickness threshold c is set to determine whether the roast duck pancake is acceptable. Defective product size inspection uses a binary mask to calculate the equivalent diameter area of ​​the roast duck pancake, the equivalent circle diameter, the minimum circumcircle diameter, and the maximum incircle diameter, and sets a diameter threshold d to determine whether the roast duck pancake is qualified. Roundness detection uses the defective product size detection value to calculate the roundness of the roast duck pancake, sets a roundness threshold e, and determines whether the roast duck pancake is qualified. The thickness inspection of defective products is the same as that of acceptable thickness inspection. The thickness of the roast duck pancake is measured by laser up and down through a beam, and a thickness threshold f is set to determine whether the roast duck pancake is qualified.

[0034] To better illustrate the above testing items, the following table is provided for supplementary explanation: It should be noted that the data in the table above is for reference and supplementary explanation only, and can be adjusted according to the actual situation. Furthermore, the threshold range can be determined through a large number of sample calibrations and experiments, thus making it applicable to the detection and classification of other types of pancakes.

[0035] Here, we will take color detection as an example to provide further practical instructions: Due to the influence of lighting, a uniform background light source can be provided.

[0036] First, establish reference standards, including static and dynamic standards. Static reference: Use a standard color chart or a customized Peking duck pancake color chart, photographed under the same lighting conditions, as a color benchmark; or take the average color of multiple qualified Peking duck pancakes as a reference. Dynamic standard: Based on production batches, collect images of several qualified products daily, calculate their average color as the daily reference, and eliminate the impact of fluctuations in raw materials and processes between batches.

[0037] Then, when extracting the color difference ΔE feature, the following features can be extracted for the entire Peking duck pancake: Maximum color difference: The maximum value of ΔE in the entire cake, reflecting the most serious defect.

[0038] Average color difference: The average ΔE value of the entire cake, reflecting the overall color deviation.

[0039] Color difference exceeding the standard area ratio: The proportion of the total area of ​​pixels whose ΔE exceeds a certain threshold.

[0040] Color difference distribution histogram: statistically analyzes the distribution of ΔE, which can be used to determine uniformity.

[0041] Finally, set a color difference threshold: for example, ΔE≤3 is acceptable, and anything exceeding this threshold is unacceptable.

[0042] refer to Figures 3-4The conveying mechanism 1 includes two sets of parallel front frames 11 and rear frames 12. Conveying rollers 13 are installed at both ends of the front and rear frames 11 and 12. A front conveyor belt 14 is installed on the front frame 11 and a rear conveyor belt 15 is installed on the rear frame 12. The roast duck pancakes to be inspected move from one end of the front conveyor belt 14 to the rear conveyor belt 15 and then leave the conveying mechanism 1. During this process, they are inspected and classified. The output ends of the front conveyor belt 14 and the rear conveyor belt 15 are respectively located at the front frame 11 and the rear frame 12, each equipped with a flip-plate outer shaft 16. A flip-plate assembly 17 is installed in the flip-plate outer shaft 16. The flip-plate assembly 17 is used to flip the roast duck pancakes and move them away from the conveyor belts. The flip-plate assembly 17 on the front conveyor belt 14 is used to flip the roast duck pancakes and move them onto the rear conveyor belt 15. The flip-plate assembly 17 on the rear conveyor belt 15 is used to remove the roast duck pancakes from the conveying mechanism 1. A cover 18 is provided on the top of the front frame 11 and the rear frame 12. The cover 18 can be a frame fixed to the front frame 11 and the rear frame 12, creating a relatively sealed space for a portion of the front conveyor belt 14 and the rear conveyor belt 15. Openings are only provided at the inlet and outlet ends of the roast duck pancakes, reducing the contamination of the roast duck pancakes by the external environment during the inspection process and providing a relatively stable inspection environment. Therefore, it is conceivable to install a light source at the top of the inner cavity of the cover 18 to improve the consistency of the inspection environment. Furthermore, a cooling fan can be installed at the top of the inner cavity of the cover 18, in front of the front conveyor belt 14 and the rear conveyor belt 15. This not only provides heat dissipation for the roast duck pancakes made directly in the forming machine, facilitating subsequent packaging, but also cleans the surface of the roast duck pancakes through airflow.

[0043] refer to Figure 7 The flip plate assembly 17 includes a flipping shaft 171 controlled by a flipping motor, which is connected to the outer shaft 16 of the flip plate. A flipping claw 172 is provided on the top of the flipping shaft 171, and a baffle 173 is provided on the top of the flipping claw 172. The first channel I, the second channel II, the third channel III, and the fourth channel IV are all controlled by the flip plate assembly 17 separately. In order to improve the working efficiency and accuracy of the flip plate assembly 17, a photoelectric sensor can be installed at a suitable position in the inner cavity of the cover 18. When the roast duck pancake passes by and blocks the light, the sensor triggers the corresponding flipping motor to realize the flipping and transfer operation. The position of the sensor can be reasonably selected according to the actual situation. For example, the projection position of the sensor can be close to the tip of the flipping claw 172 but not on the projection of the flip plate assembly 17. Through the delayed trigger design, the accurate flipping and transfer operation can also be achieved by considering the moving speed of the roast duck pancake. The front conveyor belt 14 and the rear conveyor belt 15 are discontinuous according to the position of the flipper 172. To ensure the relative position of the front conveyor belt 14 and the rear conveyor belt 15 remains stable during operation, grooves adapted to the front conveyor belt 14 and the rear conveyor belt 15 can be formed on the conveyor roller 13, such as... Figure 8 As shown.

[0044] The front defective product guide assembly 21, the front secondary product guide assembly 22, the rear defective product guide assembly 1 23, the rear defective product guide assembly 24, the rear secondary product guide assembly 25, and the path guide assembly 26 of the dough guide mechanism 2 have the same structure, only the size is different according to the actual situation. It includes a lifting column 201 fixed on the top of the conveying mechanism 1. The bottom of the lifting column 201 is provided with a main guide plate 202. The inner cavity of the main guide plate 202 is provided with a telescopic rod 203. The output end of the telescopic rod 203 is provided with a movable guide plate 204, and the movable guide plate 204 is slidably disposed in the main guide plate 202. Under normal circumstances, the lifting column 201 is in a shortened state. It only extends when the corresponding detection component detects a defect in the moving roast duck pancake. This causes the main guide plate 202 and the moving guide plate 204 to press against the front conveyor belt 14 and the rear conveyor belt 15, changing the position of the roast duck pancake. Then, the lifting column 201 immediately shortens again. (Refer to...) Figures 5-6 The telescopic pole 203 can adjust the position of the movable guide plate 204, better guiding the roast duck pancake to change its position after changing the channel according to its size, so that the roast duck pancake is as close to the center of the channel as possible.

[0045] In the preceding non-conforming product guide assembly 21, the preceding secondary product guide assembly 22, the following non-conforming product guide assembly 1 23, the following non-conforming product guide assembly 24, and the following secondary product guide assembly 25, the number of the above-mentioned structures is the same for all groups. In the path guide assembly 26, the number of the above-mentioned structures is two groups, and the overall structure is a gradually narrowing funnel shape along the conveying direction. (Refer to...) Figure 6 .

[0046] The roast duck pancakes flowing out from the first channel I, the second channel II, the third channel III, and the fourth channel IV correspond to the quality of unqualified products, qualified products, acceptable substandard products, and unqualified products, thus achieving the effect of classified collection.

[0047] In another embodiment, the control feedback of the telescopic poles 203 in the front non-conforming product guide assembly 21, the front secondary product guide assembly 22, the rear non-conforming product guide assembly 23, the rear non-conforming product guide assembly 24, and the rear secondary product guide assembly 25 is associated with the acceptable size detection or defective product size detection in the corresponding front acceptable product detection assembly 31, the front acceptable product detection assembly 32, the rear acceptable product detection assembly 33, the rear acceptable product detection assembly 24, and the rear acceptable product detection assembly 35. The size of each roast duck pancake can be determined using the size detection results from the front acceptable product detection assembly 31, the front acceptable product detection assembly 32, the rear acceptable product detection assembly 33, the rear acceptable product detection assembly 24, and the rear acceptable product detection assembly 35, and thus fed back to the telescopic poles 203, enabling real-time position adjustment of the moving guide plate 204, further improving the position of the roast duck pancakes closer to the center of the channel after changing the channel.

Claims

1. A Peking duck pancake conveyor based on visual online detection and classification, characterized in that, include, The conveying mechanism (1) is divided into a front conveying area A and a rear conveying area B according to the conveying process, and into a first channel (Ⅰ), a second channel (Ⅱ), a third channel (Ⅲ) and a fourth channel (Ⅳ) according to the collection path of the roast duck pancakes. The pancake guiding mechanism (2) is used to change the transport channel during the conveying of roast duck pancakes. It includes a path guiding component (26), a front non-conforming product guiding component (21), a front secondary product guiding component (22), a rear non-conforming product guiding component one (23), a rear non-conforming product guiding component two (24), and a rear secondary product guiding component (25) located at the top of the conveying mechanism (1). The path guiding component (26), the front non-conforming product guiding component (21), and the front secondary product guiding component (22) are located in the front conveying area A, and the rear non-conforming product guiding component one (23), the rear non-conforming product guiding component two (24), and the rear secondary product guiding component (25) are located in the rear conveying area B. The detection module (3) is used to detect whether the roast duck pancakes are qualified and whether they are defective during the transportation process on the conveyor mechanism (1). The roast duck pancakes are divided into unqualified products and acceptable products after the qualification test. The acceptable products are divided into substandard products and qualified products after the defective product test. The detection module (3) includes a front acceptable product detection component (31), a front qualified product detection component (32), a rear acceptable product detection component one (33), a rear acceptable product detection component two (34), and a rear qualified product detection component (35). 35) The front acceptable product inspection component (31), the front qualified product inspection component (32), the rear acceptable product inspection component one (33), the rear acceptable product inspection component two (34) and the rear qualified product inspection component (35) are respectively set before the front unqualified product guide component (21), the front secondary product guide component (22), the rear unqualified product guide component one (23), the rear unqualified product guide component two (24) and the rear secondary product guide component (25), and the detection is fed back to the matching guide component.

2. A Peking duck pancake conveyor based on visual online detection and classification according to claim 1, characterized in that: The front receptive item detection component (31), the rear receptive item detection component one (33) and the rear receptive item detection component two (34) use cameras to acquire images based on foreign object detection, color detection, damage detection, receptive size detection and receptive thickness detection; The pre-qualified product inspection component (32) and the post-qualified product inspection component (35) use cameras to acquire images based on defective product size inspection, roundness inspection and defective product thickness inspection; Whether roast duck pancakes are acceptable is determined by whether there is a food safety risk or serious appearance defect. Whether roast duck pancakes are substandard is determined by the size of appearance defects while ensuring they are safe to eat.

3. A Peking duck pancake conveyor based on visual online detection and classification according to claim 2, characterized in that: Foreign object detection uses a deep convolutional neural network to identify foreign objects and determine whether a roast duck pancake is acceptable. Color detection analyzes the hue (H) and saturation (S) distribution of the roast duck pancake area using the HSV color space, calculates the color difference with the standard pancake, and determines whether the roast duck pancake is acceptable. Damage detection calculates the bounding rectangle, area, and perimeter of the Peking duck pancake outline through contour analysis. Based on the area of ​​contour defects, a threshold 'a' is set to determine whether the Peking duck pancake is acceptable. The acceptable size detection calculates the equivalent diameter (area equivalent circle diameter), minimum circumcircle diameter, and maximum incircle diameter of the Peking duck pancake using a binarized mask, sets a diameter threshold b, and determines whether the Peking duck pancake is acceptable. The thickness of the roast duck pancake can be measured by laser up and down through a through-beam method. A thickness threshold c is set to determine whether the roast duck pancake is acceptable. The defective product size inspection uses a binary mask to calculate the equivalent diameter (area equivalent circle diameter), minimum circumcircle diameter, and maximum incircle diameter of the roast duck pancake, and sets a diameter threshold d to determine whether the roast duck pancake is qualified. Roundness detection uses the defective product size detection value to calculate the roundness of the roast duck pancake, sets a roundness threshold e, and determines whether the roast duck pancake is qualified. The thickness inspection of defective products is the same as that of acceptable thickness inspection. The thickness of the roast duck pancake is measured by laser up and down through a beam, and a thickness threshold f is set to determine whether the roast duck pancake is qualified.

4. A Peking duck pancake conveyor based on visual online detection and classification according to claim 1, characterized in that: The conveying mechanism (1) includes two sets of parallel front frames (11) and rear frames (12). Conveying rollers (13) are provided at both ends of the front frame (11) and the rear frame (12). A front conveyor belt (14) located on the front frame (11) and a rear conveyor belt (15) located on the rear frame (12) are provided on the conveyor rollers (13). The output ends of the front conveyor belt (14) and the rear conveyor belt (15) are respectively provided with flip plate outer shaft rods (16) at the front frame (11) and the rear frame (12). A flip plate assembly (17) is provided in the flip plate outer shaft rod (16). The flip plate assembly (17) is used to flip the roast duck pancake and move it away from the conveyor belt. A cover (18) is provided on the top of the front frame (11) and the rear frame (12).

5. A Peking duck pancake conveyor based on visual online detection and classification according to claim 4, characterized in that: The flip plate assembly (17) includes a flipping shaft (171) controlled by a flipping motor, which is connected to the outer shaft (16) of the flip plate. A flipping claw (172) is provided on the top of the flipping shaft (171), and a baffle (173) is provided on the top of the flipping claw (172). The front conveyor belt (14) and the rear conveyor belt (15) are discontinuous according to the position of the flipping claw (172).

6. A Peking duck pancake conveyor based on visual online detection and classification according to claim 1, characterized in that: The path guiding component (26) is located in the second channel (II) in a trumpet shape, and is used to guide the roast duck pancakes to concentrate in the second channel (II); The front non-conforming product guide assembly (21), the front secondary product guide assembly (22), the rear non-conforming product guide assembly one (23), the rear non-conforming product guide assembly two (24), and the rear secondary product guide assembly (25) can be raised and lowered to guide the roast duck pancakes to change the conveying channel. The front non-conforming product guide assembly (21) is used for roast duck pancakes from the second channel (II) to the first channel (I), the front secondary product guide assembly (22) is used for roast duck pancakes from the second channel (II) to the third channel (III), the rear non-conforming product guide assembly one (23) is used for roast duck pancakes from the second channel (II) to the first channel (I), the rear non-conforming product guide assembly two (24) is used for roast duck pancakes from the third channel (III) to the fourth channel (IV), and the rear secondary product guide assembly (25) is used for roast duck pancakes from the second channel (II) to the third channel (III).

7. A Peking duck pancake conveyor based on visual online detection and classification according to claim 1, characterized in that: The front non-conforming product guide assembly (21), the front secondary product guide assembly (22), the rear non-conforming product guide assembly one (23), the rear non-conforming product guide assembly two (24), the rear secondary product guide assembly (25), and the path guide assembly (26) of the dough guide mechanism (2) have the same structure, including a lifting column (201) fixed on the top of the conveying mechanism (1), a main guide plate (202) at the bottom of the lifting column (201), and a telescopic electric rod (203) in the inner cavity of the main guide plate (202). The output end of the rod (203) is provided with a movable guide plate (204), and the movable guide plate (204) is slidably disposed in the main guide plate (202). The number of the above structures in the front non-conforming product guide assembly (21), the front secondary product guide assembly (22), the rear non-conforming product guide assembly one (23), the rear non-conforming product guide assembly two (24), and the rear secondary product guide assembly (25) is the same. The number of the above structures in the path guide assembly (26) is two, and the overall structure is a gradually contracting trumpet shape along the conveying direction.

8. A Peking duck pancake conveyor based on visual online detection and classification according to claim 7, characterized in that: The control feedback of the telescopic pole (203) in the front nonconforming product guide assembly (21), the front secondary product guide assembly (22), the rear nonconforming product guide assembly one (23), the rear nonconforming product guide assembly two (24), and the rear secondary product guide assembly (25) is associated with the acceptable size detection or defective size detection in the corresponding front acceptable product detection assembly (31), the front acceptable product detection assembly (32), the rear acceptable product detection assembly one (33), the rear acceptable product detection assembly two (34), and the rear acceptable product detection assembly (35).

9. A Peking duck pancake conveyor based on visual online detection and classification according to claim 1, characterized in that: The roast duck pancakes flowing out from the first channel (Ⅰ), the second channel (Ⅱ), the third channel (Ⅲ), and the fourth channel (Ⅳ) correspond to the following quality levels: unqualified, qualified, acceptable substandard, and unqualified.

Images