Turnout type sorting device and sorting method

Abstract

The invention relates to a switch type sorting device and a sorting method, which solves the technical problems that the existing fruit and vegetable sorting device is easy to cause damage to fragile fruits and vegetables or fruits and vegetables with high appearance requirements and has low sorting efficiency. The device comprises a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt, a switch type diversion mechanism and a plurality of trays. The stainless steel tubular conveyor belt is provided with a plurality of stainless steel tubes. The switch type diversion mechanism is provided with a main track, a segmented main track 1, a segmented main track 2, a diversion track 1, a diversion track 2, a grading track 1, a grading track 2, a merging track, a return track 1, a return track 2, a reversing finger 1, an electromagnet connecting seat 1, an electromagnet 1, a reversing finger 2, an electromagnet connecting seat 2 and an electromagnet 2. The tray is slidably connected with the stainless steel tube, and the tray is provided with a guide shaft.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit and vegetable sorting, and in particular to a switch type sorting device and a sorting method. Background Art

[0002] In the field of fruit and vegetable sorting technology, grading equipment usually makes fruits and vegetables of different grades fall into corresponding grading areas to achieve grading and sorting. Please refer to the invention patent application with application publication number CN117463598A. However, due to the height difference, it is easy to cause damage to fruits and vegetables. It is only suitable for fruits and vegetables with thicker skin, or for sorting scenes with low requirements for the appearance of fruits and vegetables. It is not suitable for fruits and vegetables that are easily damaged or have high requirements for appearance (such as cucumbers). In addition, the sorting efficiency needs to be improved.

[0003] There is also a sorting device in the prior art, which uses a flexible tray and is mainly suitable for fragile spherical fruits and vegetables of various sizes, such as peaches and watermelons. The grading principle is that the fruits and vegetables are loaded on independent flexible round trays and conveyed to the visual module for detection via a conveyor belt. After arriving at the grading area, the flexible round tray is hit by an external force, changes the moving direction, moves to the corresponding grading area, and completes the grading. By using a flexible tray with a certain degree of coverage and a high degree of freedom, there is no height difference in the grading process of fruits and vegetables, which protects fragile fruits and vegetables to a certain extent. However, there is still a problem of easily causing damage to fragile fruits and vegetables, and the sorting efficiency is low.

[0004] In the prior art, there are some technical means for fruit and vegetable grading judgment, which is to collect images of fruits and vegetables through cameras, further extract features based on the texture and color of the images, and then grade them. However, this method is easily affected by the ambient light intensity and the image background, resulting in poor grading effect, low grading accuracy and low efficiency. Summary of the invention

[0005] The present invention aims to solve the technical problems that the existing fruit and vegetable sorting devices are prone to damage fragile fruits and vegetables or fruits and vegetables with high appearance requirements, have low sorting efficiency, and have low accuracy and efficiency in grading according to the texture and color of fruit and vegetable images through machine vision. A switch-type sorting device and sorting method are provided to avoid damage and have high operating efficiency.

[0006] The present invention provides a turnout type sorting device, comprising a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt, a turnout type diverting mechanism and a plurality of trays, wherein the driving mechanism is connected to the supporting frame;

[0007] The stainless steel tube conveyor belt includes a left chain, a right chain and a plurality of stainless steel tubes, wherein the plurality of stainless steel tubes are connected between the left chain and the right chain; the driving mechanism is used to drive the right chain and the left chain to move;

[0008] The turnout-type diversion mechanism includes a top track support tube, a main track, a segmented main track 1, a segmented main track 2, a diversion track 1, a diversion track 2, a graded track 1, a graded track 2, a bottom track support tube, a merging track, a return track 1, a return track 2, a reversing finger 1, an electromagnet connection seat 1, an electromagnet 1, a reversing finger 2, an electromagnet connection seat 2 and an electromagnet 2, a plurality of top track support tubes are fixedly connected to the support frame, a plurality of bottom track support tubes are fixedly connected to the support frame, a plurality of bottom track support tubes are located on a horizontal plane B, a plurality of top track support tubes are located on a horizontal plane A, and the horizontal plane B is located below the horizontal plane A; the main track, the segmented main track 1 and the segmented main track 2 are respectively It is fixedly connected to multiple top-level rail support tubes, the main track, the segmented main track one and the segmented main track two are located on the same straight line and close to one side of the support frame, a gap is provided between the front end of the segmented main track one and the rear end of the main track, a gap is provided between the front end of the segmented main track two and the rear end of the segmented main track one, the main track, the segmented main track one and the segmented main track two are distributed along the longitudinal direction, the shunt track one is fixedly connected to multiple top-level rail support tubes, the shunt track one is inclinedly arranged with the straight line where the main track is located as the reference, the front end of the shunt track one is close to the rear end of the main track, the electromagnet connector one is connected to the top-level rail support tube, the electromagnet one is connected to the electromagnet connector one, and the reversing finger One is connected to the rotating shaft of electromagnet one, and the reversing finger one is located between the rear end of the main track and the front end of the shunt track one; the shunt track two is fixedly connected to multiple top-level track support tubes, and the shunt track two is tilted with the straight line of the segmented main track one as the reference, and the front end of the shunt track two is close to the rear end of the segmented main track one, and the shunt track two is located between the shunt track one and the segmented main track two, the electromagnet connecting seat two is connected to the top-level track support tube, the electromagnet two is connected to the electromagnet connecting seat two, the reversing finger two is connected to the rotating shaft of electromagnet two, and the reversing finger two is located between the front end of the shunt track two and the rear end of the segmented main track one; the graded track one is fixedly connected to multiple top-level track support tubes, and the graded track The first is close to the rear end of the shunt track one, the grading track one is located outside the rear end of the shunt track one, and the grading track one is parallel to the main track; the grading track two is fixedly connected to a plurality of top track support tubes, the grading track two is close to the rear end of the shunt track two, the grading track two is located outside the rear end of the shunt track two, and the grading track two is parallel to the segmented main track one; the merging track is fixedly connected to a plurality of bottom track support tubes, along the vertical direction, the merging track is located directly below the main track, the return track one is fixedly connected to a plurality of bottom track support tubes, the return track one is arranged obliquely with the straight line where the merging track is located, the rear end of the return track one is close to the merging track, and the front end of the return track one is located below the rear end of the grading track one;The return track 2 is fixedly connected to a plurality of bottom track support pipes, and is arranged obliquely with respect to the straight line where the merging track is located, with the rear end of the return track 2 close to the merging track, and the front end of the return track 2 located below the rear end of the graded track 2; the main track, the segmented main track 1, the segmented main track 2, the diversion track 1, the diversion track 2, the graded track 1, the graded track 2, the merging track, the return track 1 and the return track 2 are all located in the space surrounded by the stainless steel tubular conveyor belt;

[0009] The tray is slidably connected to the stainless steel tube of the stainless steel tube conveyor belt, and the tray is provided with a guide shaft;

[0010] The guide shaft can move along the main track, the guide shaft can move along the segmented main track one, the guide shaft can move along the segmented main track two, the guide shaft can move along the diversion track one, the guide shaft can move along the diversion track two, the guide shaft can move along the graded track one, the guide shaft can move along the graded track two, the guide shaft can move along the regression track one, the guide shaft can move along the regression track two, and the guide shaft can move along the merging track.

[0011] Preferably, the tray is slidably connected to two stainless steel tubes of the stainless steel tube conveyor belt.

[0012] Preferably, the tray includes a base and a lifting component, the lifting component is connected to the top of the base, and the guide shaft is connected to the bottom of the base; the base is provided with a slide groove, and the stainless steel pipe is slidably connected to the slide groove.

[0013] Further preferably, the lifting component is a plurality of contoured lifting plates, and the plurality of contoured lifting plates are arranged in a line.

[0014] Preferably, the driving mechanism includes a driving motor, a driving shaft, a driven shaft, a first driving sprocket, a second driving sprocket, a first driven sprocket and a second driven sprocket, the driving motor is connected to the supporting frame, the driving shaft is rotatably connected to the supporting frame, the driven shaft is rotatably connected to the supporting frame, the output shaft of the driving motor is connected to the driving shaft, the first driving sprocket is connected to the right end of the driving shaft, the second driving sprocket is connected to the left end of the driving shaft, the first driven sprocket is connected to the right end of the driven shaft, and the second driven sprocket is connected to the left end of the driven shaft; the left chain of the stainless steel tubular conveyor belt is connected between the second driving sprocket and the second driven sprocket, and the right chain is connected between the first driving sprocket and the first driven sprocket.

[0015] Preferably, the main track is provided with a vertical portion, the segmented main track one is provided with a vertical portion, the segmented main track two is provided with a vertical portion, the diversion track one is provided with a vertical portion, the graded track one is provided with a vertical portion, the diversion track two is provided with a vertical portion, the graded track two is provided with a vertical portion, the merging track is provided with a vertical portion, the return track one is provided with a vertical portion, and the return track two is provided with a vertical portion.

[0016] The present invention also provides a turnout type sorting device, comprising a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt, a turnout type diverting mechanism and a plurality of trays, wherein the driving mechanism is connected to the supporting frame;

[0017] The stainless steel tube conveyor belt includes a left chain, a right chain and a plurality of stainless steel tubes, wherein the plurality of stainless steel tubes are connected between the left chain and the right chain; the driving mechanism is used to drive the right chain and the left chain to move;

[0018] The turnout-type shunting mechanism comprises a top-level rail support tube, a main track, a segmented main track, a shunting rail, a graded rail, a bottom-level rail support tube, a merging rail, a return rail, a reversing finger, an electromagnet connector and an electromagnet. A plurality of top-level rail support tubes are fixedly connected to the support frame, a plurality of bottom-level rail support tubes are fixedly connected to the support frame, a plurality of bottom-level rail support tubes are located on a horizontal plane B, a plurality of top-level rail support tubes are located on a horizontal plane A, and horizontal plane B is located below horizontal plane A; the main track and the segmented main track are respectively fixedly connected to a plurality of top-level rail support tubes, the main track and the segmented main track are located on the same straight line and close to one side of the support frame, a spacing is provided between the front end of the segmented main track and the rear end of the main track, the main track and the segmented main track are distributed along the longitudinal direction, the shunting rail is fixedly connected to a plurality of top-level rail support tubes, the shunting rail is arranged tilted with respect to the straight line where the main track is located, and the front end of the shunting rail is close to At the rear end of the main track, the electromagnet connection seat 1 is connected to the top track support tube, the electromagnet 1 is connected to the electromagnet connection seat 1, the reversing finger 1 is connected to the rotating shaft of the electromagnet 1, and the reversing finger 1 is located between the rear end of the main track and the front end of the diversion track 1; the graded track 1 is fixedly connected to multiple top track support tubes, the graded track 1 is close to the rear end of the diversion track 1, the graded track 1 is located outside the rear end of the diversion track 1, and the graded track 1 is parallel to the main track; the merging track is fixedly connected to multiple bottom track support tubes, along the vertical direction, the merging track is located directly below the main track, the return track 1 is fixedly connected to multiple bottom track support tubes, the return track 1 is arranged obliquely with the straight line where the merging track is located, the rear end of the return track 1 is close to the merging track, and the front end of the return track 1 is located below the rear end of the graded track 1; the main track, the segmented main track 1, the diversion track 1, the graded track 1, the merging track and the return track 1 are all located in the space surrounded by the stainless steel tubular conveyor belt;

[0019] The tray is slidably connected to the stainless steel tube of the stainless steel tube conveyor belt, and the tray is provided with a guide shaft;

[0020] The guide shaft can move along the main track, the guide shaft can move along the segmented main track, the guide shaft can move along the diversion track, the guide shaft can move along the graded track, the guide shaft can move along the return track, and the guide shaft can move along the merging track.

[0021] Preferably, the main track is provided with a vertical portion, the segmented main track is provided with a vertical portion, the diversion track is provided with a vertical portion, the graded track is provided with a vertical portion, the merging track is provided with a vertical portion, and the return track is provided with a vertical portion.

[0022] Preferably, the rear end of the diverter track is close to the other side of the supporting frame.

[0023] Preferably, the tray is slidably connected to two stainless steel tubes of the stainless steel tube conveyor belt.

[0024] The present invention also provides a sorting method using a turnout type sorting device. After the turnout type sorting device starts working, the fruits and vegetables on the tray in the input stage are graded to obtain grade data, and then the electromagnet one or the electromagnet two are controlled to act according to the grade data to sort the fruits and vegetables.

[0025] The process of obtaining grade data is carried out according to the following steps:

[0026] The first step is to establish a fruit and vegetable key point detection image dataset;

[0027] Step S101, image acquisition, using a camera to shoot fruits and vegetables on a tray directly below it from a bird's-eye view, to acquire a large number of fruit and vegetable images;

[0028] Step S102, image annotation, annotating key points in the collected fruit and vegetable image data to construct a fruit and vegetable key point detection image data set;

[0029] The second step is to use the fruit and vegetable key point detection image dataset to train the YOLOv8-pose model or RTMPose model to obtain the fruit and vegetable key point detection model;

[0030] The third step is to calculate the fruit and vegetable grading index;

[0031] Input the images of fruits and vegetables to be graded collected by the camera into the key point detection model of fruits and vegetables to obtain the key point recognition results;

[0032] Calculate the quantitative size index based on the obtained key points;

[0033] The fourth step is to determine the grade based on the quantitative size index.

[0034] Preferably, the grading to obtain grade data is performed on cucumbers; step S102, there are 10 key points, point 1 is the top of the cucumber, point 2 is the handle end of the cucumber, points 3 and 4 are edge points 3 cm away from the top of the cucumber, points 5 and 6 are edge points in the middle of the cucumber, points 7 and 8 are edge points 3 cm away from the handle end of the cucumber, point 9 is the maximum bend of the cucumber, and point 10 is the midpoint of the boundary between the handle and the body of the cucumber;

[0035] In the third step, the key point recognition results have 10 key points, and the quantitative size indicators are calculated based on the 10 key points. The quantitative size indicators include D (bow height), D (transverse longitude difference), and D (melon handle ratio);

[0036] D (bow height) is calculated by the following formula (1):

[0037]

[0038] D (difference in transverse longitude) is calculated by the following formula (2):

[0039] D(diameter difference)=max(|D(3,4)-D(5,6)|,|D(7,8)-D(5,6)|) (2);

[0040] D (melon-to-handle ratio) is calculated by the following formula (3):

[0041]

[0042] In formula (3), D (melon length) is calculated by the following formula (4):

[0043] D(melon length) = D(1,9') + D(2,9') (4);

[0044] In the above formulas (1), (2) and (3), D(1,2) represents the distance between point 1 and point 2, D(2,10) represents the distance between point 2 and point 10, D(3,4) represents the distance between point 3 and point 4, D(5,6) represents the distance between point 5 and point 6, D(7,8) represents the distance between point 7 and point 8, 9' is the distance between point 5 and point 6, moving point 9 at the maximum bending point of the cucumber to the middle of the cucumber, D(1,9') represents the distance between point 1 and point 9', D(2,9') represents the distance between point 2 and point 9', and D(9',1_2) represents the perpendicular distance from point 9' to the straight line between point 1 and point 2.

[0045] The present invention also provides a sorting method using a turnout type sorting device. After the turnout type sorting device starts working, the fruits and vegetables on the tray in the input stage are graded to obtain grade data, and then the electromagnet one or the electromagnet two are controlled to act according to the grade data to sort the fruits and vegetables.

[0046] The process of obtaining grade data is carried out according to the following steps:

[0047] The first step is to establish a fruit and vegetable key point detection image dataset;

[0048] Step S101, image acquisition, using a camera to shoot fruits and vegetables on a tray directly below it from a bird's-eye view, to acquire a large number of fruit and vegetable images;

[0049] Step S102, image annotation, annotating key points in the collected fruit and vegetable image data to construct a fruit and vegetable key point detection image data set;

[0050] The second step is to establish an appearance classification dataset. The fruits and vegetables on the tray are photographed by a camera from a top and side view to collect a large number of fruit and vegetable images, and then the fruits and vegetables are labeled to obtain an appearance classification dataset.

[0051] The third step is to train the neural network model using the fruit and vegetable key point detection image dataset to obtain the fruit and vegetable key point detection model, and use the appearance classification dataset to train the convolutional neural network model to obtain the appearance classification model based on the convolutional neural network;

[0052] The fourth step is (1) calculating the fruit and vegetable grading index; inputting the image of the fruit and vegetable to be graded collected by the camera into the fruit and vegetable key point detection model to obtain the key point recognition result, and calculating the quantitative size index based on the obtained key points;

[0053] (2) inputting the images of fruits and vegetables to be graded collected by the camera into an appearance classification model based on a convolutional neural network, and the appearance classification model based on the convolutional neural network outputs the appearance category level;

[0054] Step 5: Determine the grade based on the quantitative size index;

[0055] Step 6. Combine the appearance category grade with the grade determined in step 5 and determine the lower grade or the same grade as the final grade.

[0056] The beneficial effects of the present invention are: realizing a turnout type sorting method, and the sorting speed is fast and the efficiency is high; the whole sorting process is completed without height difference, and will not cause damage to fruits and vegetables, and can provide good protection for fruits and vegetables. It is not limited to fruits and vegetables that are fragile or have high appearance requirements, and can also be applied to other fruits and vegetables.

[0057] It is not limited to sorting fruits and vegetables, but can also sort any other items that can be sorted.

[0058] The cucumber grading effect is better, with higher accuracy, faster speed and higher efficiency; it has strong stability and is not easily affected by ambient light intensity and image background. Based on the deep convolutional neural network algorithm, the key points of the cucumbers in the camera field of view are detected quickly and accurately; the bow height, horizontal diameter difference, cucumber handle ratio and other indicators are calculated based on the detected key points; the true value of the above indicators is calculated based on the real size of the reference object (such as the tray on which the cucumbers are placed), and the cucumber grade is obtained by comparing it with the cucumber grade standard.

[0059] It can replace human labor in cucumber grading tasks, greatly reducing labor costs. The grading method is not limited to cucumbers, but can also be applied to other fruit and vegetable grading tasks.

[0060] The cucumber grading effect is better, with higher accuracy, faster speed and higher efficiency; it has strong stability and is not easily affected by ambient light intensity and image background.

[0061] Further features of the present invention will be clearly described in the following description of the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is an axonometric view of a turnout type fruit and vegetable sorting device;

[0063] Figure 2 yes Figure 1 In the structure shown, the schematic diagram of the drive mechanism and the stainless steel tubular conveyor belt;

[0064] Figure 3 yes Figure 2 A partial enlarged view of

[0065] Figure 4 yes Figure 1 A schematic diagram of the track connected to the track support tube in the structure shown;

[0066] Figure 5 It is a structural schematic diagram of a main track, a segmented main track 1, a segmented main track 2, a diversion track 1, a diversion track 2, a graded track 1 and a graded track 2 installed on a plurality of top-level track support tubes;

[0067] Figure 6 yes Figure 5 a bottom view of the structure shown;

[0068] Figure 7 yes Figure 5 A partial enlarged view of a middle reversing finger;

[0069] Figure 8 yes Figure 5 A partial enlarged view of the second position of the middle reversing finger;

[0070] Fig. 9 yes Figure 6 A partial enlarged view of one part of the electromagnet;

[0071] Fig.10 This is a state diagram in which the free end of the reversing finger 1 is close to the rear end of the main track;

[0072] Fig.11 This is the axonometric view of the reversing finger 1;

[0073] Fig.12 yes Fig.11 A top view of the reversing finger 1 shown;

[0074] Fig.13 It is an axonometric view of the pallet;

[0075] Fig.14 It is the main view of the tray;

[0076] Fig.15 In the figure, Figure (a) is a top view of the pallet, and Figure (b) is a schematic diagram of dimensioning the grooves on the contoured lifting plate based on the top view;

[0077] Fig.16 It is the right side view of the tray;

[0078] Fig.17 It is an axonometric view of the pallet;

[0079] Fig.18 It is a schematic diagram of the connection between the tray and two stainless steel pipes;

[0080] Fig.19 It is a schematic diagram of the state in which the guide shaft of the pallet moves along the vertical portion of the main track;

[0081] Fig. 20 yes Figure 4 A partial enlarged view of the middle diversion track 1 and the graded track 1;

[0082] Fig.21 yes Figure 4 A partial enlarged view of the second location of the middle grade track;

[0083] Fig. 22 This is a schematic diagram of the input stage of the entire turnout type fruit and vegetable sorting device working process, where the guide shaft of the tray moves along the vertical part of the main track;

[0084] Fig.23 This is a schematic diagram of the state in which the guide shafts of a part of the trays move against the vertical part of the diversion track 1 and the guide shafts of another part of the trays move against the vertical part of the diversion track 2 during the diversion stage of the working process of the whole turnout type fruit and vegetable sorting device;

[0085] Fig.24 It is a schematic diagram of the working process of the turnout type fruit and vegetable sorting device;

[0086] Fig.25 is an image from the training dataset created to train the cucumber detection model;

[0087] Fig.26 It is a graphic representation of the 10 key points of the cucumber definition;

[0088] Fig. 27 This is a diagram of key point annotation of a cucumber image;

[0089] Fig.28 This is a diagram of key point annotation of a cucumber image;

[0090] Fig.29 This is a diagram of key point annotation of a cucumber image;

[0091] Fig.30 This is a schematic diagram of the cucumber key point detection model based on YOLOv8-pose

[0092] Fig.31 This is a schematic diagram of the cucumber key point detection model based on RTMPose;

[0093] Fig.32 is the detection result output by the cucumber key point detection model;

[0094] Fig.33 This is a schematic diagram of the visual system of the automatic cucumber grading equipment;

[0095] Fig.34 It is a diagram showing the quantitative size indicators calculated based on the key points of the cucumber.

[0096] Fig.35 is the cucumber image captured by the camera from a top-down perspective;

[0097] Fig.36 is Fig.33 Flowchart of the system shown with the cucumber appearance classification model added.

[0098] Explanation of symbols in the figure:

[0099] 1. Support legs, 2. Left longitudinal support plate, 3. Right longitudinal support plate, 4. Stainless steel tubular conveyor belt, 4-1. Left chain, 4-2. Right chain, 4-3. Stainless steel tube, 5. Driving motor, 6. First driving sprocket, 7. Second driving sprocket, 8. First driven sprocket, 9. Second driven sprocket, 10. Top track support tube, 11. Main track, 11-1. Vertical part, 12. Segmented main track 1, 13. Segmented main track 2, 14. Diverter track 1, 14-1. Vertical part, 15. Diverter track 2, 15-1. Vertical part, 16. Graded track 1, 16-1. Vertical part, 17. Graded track 2, 17-1. Vertical part, 18. Bottom track support tube, 19. Merging track, 20. Return track 1, 21. Return Return track two, 22. Reversing finger one, 23. Electromagnet connecting seat one, 24. Electromagnet one, 25. Reversing finger two, 26. Electromagnet connecting seat two, 27. Electromagnet two; 28. Tray, 28-1. Base, 28-1-1. First slide slot, 28-1-2. Second slide slot, 28-1-3. Rib plate, 28-2. Guide shaft, 28-3. Profiled lifting plate, 28-3-1. Groove; 29. ​​Cucumber, 30. Cucumber, 31. Cucumber, 32. Cucumber, 33. Cucumber, 34. Cucumber, 35. Cucumber, 36. Cucumber, 37. Cucumber, 38. Cucumber, 39. Cucumber, 40. Cucumber, 41. Cucumber; 42. Tray, 43. Tray, 44. Tray, 45. Tray, 46. Tray, 47. Tray, 48. Tray. DETAILED DESCRIPTION

[0100] The present invention will be further described in detail below with reference to the accompanying drawings by way of specific embodiments.

[0101] like Figure 1-4 As shown, the switch type fruit and vegetable sorting device includes a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt 4, a switch type diverting mechanism, and a plurality of trays, and the driving mechanism is connected to the supporting frame.

[0102] The support frame includes a left longitudinal support plate 2, a right longitudinal support plate 3 and a support leg 1, and the left longitudinal support plate 2 and the right longitudinal support plate 3 are fixedly mounted on the support leg 1 respectively.

[0103] The driving mechanism includes a driving motor 5, a driving shaft, a driven shaft, a first driving sprocket 6, a second driving sprocket 7, a first driven sprocket 8, and a second driven sprocket 9. The driving motor 5 is installed at the front end of the right longitudinal support plate 3. The driving shaft is rotatably connected between the front end of the left longitudinal support plate 2 and the front end of the right longitudinal support plate 3. The driven shaft is rotatably connected between the rear end of the left longitudinal support plate 2 and the rear end of the right longitudinal support plate 3. The output shaft of the driving motor 5 is connected to the driving shaft, the first driving sprocket 6 is connected to the right end of the driving shaft, the second driving sprocket 7 is connected to the left end of the driving shaft, the first driven sprocket 8 is connected to the right end of the driven shaft, and the second driven sprocket 9 is connected to the left end of the driven shaft. The driving motor 5 drives the first driving sprocket 6 and the second driving sprocket 7 to rotate.

[0104] The stainless steel tubular conveyor belt 4 includes a left chain 4-1, a right chain 4-2 and a plurality of stainless steel tubes 4-3, and the plurality of stainless steel tubes 4-3 are connected between the left chain 4-1 and the right chain 4-2. The stainless steel tubular conveyor belt 4 is connected to the driving mechanism in such a way that the left chain 4-1 is connected between the second driving sprocket 7 and the second driven sprocket 9, and the right chain 4-2 is connected between the first driving sprocket 6 and the first driven sprocket 8. In the layout of the entire turnout-type fruit and vegetable sorting device, the plurality of stainless steel tubes 4-3 are located in the transverse direction. When the driving motor 5 is working, the first driving sprocket 6 and the second driving sprocket 7 respectively drive the right chain 4-2 and the left chain 4-1 to move, thereby driving the plurality of stainless steel tubes 4-3 to move in accordance with the rotation direction. Figure 2 Move in the direction indicated by the arrow.

[0105] The turnout-type diversion mechanism includes a top-level track support tube 10, a main track 11, a segmented main track 1 12, a segmented main track 2 13, a diversion track 1 14, a diversion track 2 15, a graded track 1 16, a graded track 2 17, a bottom-level track support tube 18, a merging track 19, a return track 1 20, a return track 2 21, a reversing finger 1 22, an electromagnet connecting seat 1 23, an electromagnet 1 24, a reversing finger 2 25, an electromagnet connecting seat 2 26, and an electromagnet 2 27.

[0106] like Figure 4 As shown, multiple top rail support tubes 10 are fixedly connected between the left longitudinal support plate 2 and the right longitudinal support plate 3. Multiple top rail support tubes 10 are located in the space surrounded by the stainless steel tubular conveyor belt 4, that is, from the side, the multiple stainless steel tubes 4-3 of the stainless steel tubular conveyor belt 4 are divided into upper stainless steel tubes and lower stainless steel tubes, and multiple top rail support tubes 10 are located between the upper stainless steel tubes and the lower stainless steel tubes. Figure 1The plurality of bottom rail support tubes 18 are fixedly connected between the left longitudinal support plate 2 and the right longitudinal support plate 3, and the plurality of bottom rail support tubes 18 are also located in the space surrounded by the stainless steel tubular conveyor belt 4. The plurality of bottom rail support tubes 18 are located on the horizontal plane B, and the plurality of top rail support tubes 10 are located on the horizontal plane A, and the horizontal plane B is located below the horizontal plane A.

[0107] like Figure 4-6As shown, the main track 11, the segmented main track 1 12 and the segmented main track 2 13 are fixedly connected to multiple top-level track support tubes 10 by screws, the main track 11, the segmented main track 1 12 and the segmented main track 2 13 are located on the same straight line, the main track 11, the segmented main track 1 12 and the segmented main track 2 13 are close to the left longitudinal support plate 2, there is a certain distance between the front end of the segmented main track 1 12 and the rear end of the main track 11 (the distance can leave an installation space for the electromagnet connecting seat 1 23), there is a certain distance between the front end of the segmented main track 2 13 and the rear end of the segmented main track 1 12 (the distance can leave an installation space for the electromagnet connecting seat 2 26), the main track 11, the segmented main track 1 12 and the segmented main track 2 13 are distributed along the longitudinal direction (i.e., along the movement direction of the stainless steel tubular conveyor belt 4). The shunt track 14 is fixedly installed on multiple top-level track support tubes 10 with screws. The shunt track 14 is tilted and arranged with the straight line where the main track 11 is located as a reference. The rear end of the shunt track 14 is close to the right longitudinal support plate 3, and the front end of the shunt track 14 is close to the rear end of the main track 11. The electromagnet connecting seat 23 is fixedly connected to the top-level track support tube 10, and the electromagnet 24 is fixedly connected to the electromagnet connecting seat 23. The reversing finger 22 is connected to the rotating shaft of the electromagnet 24, and the reversing finger 22 is located between the rear end of the main track 11 and the front end of the shunt track 14. The shunt track 15 is fixedly installed on multiple top track support tubes 10 by screws. The shunt track 15 is tilted with the straight line of the segmented main track 12 as the reference. The front end of the shunt track 15 is close to the rear end of the segmented main track 12. The shunt track 15 is located between the shunt track 14 and the segmented main track 13. The electromagnet connection seat 26 is fixedly connected to the top track support tube 10. The electromagnet 27 is fixedly connected to the electromagnet connection seat 26. The reversing finger 25 is connected to the rotating shaft of the electromagnet 27. The reversing finger 25 is located between the front end of the shunt track 15 and the rear end of the segmented main track 12. The graded track 16 is fixedly connected to multiple top track support tubes 10 by screws. The graded track 16 is close to the rear end of the shunt track 14. The graded track 16 is located outside the rear end of the shunt track 14. The graded track 16 is parallel to the main track 11. The grading track 17 is fixedly connected to the multiple top track support tubes 10 by screws, the grading track 17 is close to the rear end of the shunt track 15, the grading track 17 is located outside the rear end of the shunt track 15, and the grading track 17 is parallel to the segmented main track 12. The merging track 19 is fixedly connected to the multiple bottom track support tubes 18 by screws; in the vertical direction, the merging track 19 is located directly below the main track 11.The return track 1 20 is fixedly connected to the multiple bottom track support tubes 18 by screws. The return track 1 20 is arranged obliquely on the straight line where the merging track 19 is located. The rear end of the return track 1 20 is close to the merging track 19, and the front end of the return track 1 20 is located below the rear end of the graded track 1 16. The return track 2 21 is fixedly connected to the multiple bottom track support tubes 18 by screws. The return track 2 21 is arranged obliquely on the straight line where the merging track 19 is located. The rear end of the return track 2 21 is close to the merging track 19, and the front end of the return track 2 21 is located below the rear end of the graded track 17.

[0108] The main track 11, the segmented main track 1 12, the segmented main track 2 13, the diversion track 1 14, the diversion track 2 15, the graded track 1 16, the graded track 2 17, the merging track 19, the return track 1 20 and the return track 2 21 are all located in the space surrounded by the stainless steel tubular conveyor belt 4.

[0109] The structure of the reversing finger 22 is as follows: Fig.11 and 12 The structure of the second reversing finger 25 is the same as that of the first reversing finger 22.

[0110] refer to Figure 5 and 7 , the reversing finger 1 22 is in the initial state, in which the reversing finger 1 22 is not connected to the rear end of the main track 11. At this time, when the electromagnet 1 24 is activated, the shaft of the electromagnet 1 24 rotates a certain angle, and the shaft of the electromagnet 1 24 drives the reversing finger 1 22 to rotate a certain angle toward the rear end of the main track 11, and the free end of the reversing finger 1 22 is close to the rear end of the main track 11. Fig.10 , so as to realize the connection between the reversing finger 1 22 and the rear end of the main track 11.

[0111] The working mode of the reversing finger 25 is the same as that of the reversing finger 1 22. The rotating shaft of the electromagnet 27 rotates a certain angle, driving the reversing finger 25 to rotate a certain angle toward the rear end of the segmented main track 12, and the free end of the reversing finger 25 is close to the rear end of the segmented main track 12. Figure 8 , to achieve the connection between the reversing finger 25 and the rear end of the segmented main track 12.

[0112] refer to Figure 4 , 5 , 7 and 19, the main track 11 is provided with a vertical portion 11-1, and similarly, the segmented main track 1 12 is also provided with a vertical portion, and the segmented main track 2 13 is also provided with a vertical portion. Figure 4 , 5 and 20, the diversion track 14 is provided with a vertical portion 14-1, and the grading track 16 is provided with a vertical portion 16-1. Figure 4 ,8 and 21, the second diversion track 15 is provided with a vertical portion 15-1, and the second grading track 17 is provided with a vertical portion 17-1. Similarly, the merging track 19 is also provided with a vertical portion, the first return track 20 is also provided with a vertical portion, and the second return track 21 is also provided with a vertical portion.

[0113] like Figure 13-16 As shown, the tray 28 includes a base 28-1, a guide shaft 28-2, and a contoured lifting plate 28-3. The guide shaft 28-2 is fixedly connected to the bottom of the base 28-1, and seven contoured lifting plates 28-3 are respectively connected to the top of the base 28-1, and the seven contoured lifting plates 28-3 are arranged in a line. The base 28-1 is provided with a first slide groove 28-1-1 and a second slide groove 28-1-2, and the first slide groove 28-1-1 and the second slide groove 28-1-2 are arranged side by side, and the first slide groove 28-1-1 and the second slide groove 28-1-2 are arranged along the length direction of the base 28-1. A groove 28-3-1 is provided on the contoured lifting plate 28-3. When the cucumber is lifted by the seven contoured lifting plates, different parts of the cucumber are located in the groove 28-3-1 of each contoured lifting plate 28-3; preferably, the lengths of the grooves on the 5th, 6th and 7th contoured lifting plates are equal, L5=L6=L7, the lengths of the grooves on the 1st, 2nd, 3rd and 4th contoured lifting plates gradually decrease, L1>L2>L3>L4, the length of the groove on the 5th contoured lifting plate is less than the length of the groove on the 4th contoured lifting plate (L5<L4); the center points of L1, L2, L3, L4, L5, L6 and L7, a total of 7 points, can be connected into an arc, imitating the arc of the cucumber, that is, the 7 grooves form a limiting space adapted to the shape of the cucumber, refer to Fig.24 The grooves on the 5th, 6th and 7th contoured lifting plates are used to limit the tail of the cucumber (the tail of the cucumber refers to the end without thorns), and the grooves on the 1st, 2nd, 3rd and 4th contoured lifting plates are used to limit the parts other than the tail of the cucumber. The limiting space formed by the combination of the grooves on multiple contoured lifting plates can stably and reliably lift and prevent the cucumbers from rolling or falling off the tray due to emergency stops, speed changes and other working conditions.

[0114] like Fig.18 As shown, when the tray 28 cooperates with the two stainless steel tubes 4-3 on the stainless steel tube conveyor belt 4, the two stainless steel tubes 4-3 pass through the first slide groove 28-1-1 and the second slide groove 28-1-2 respectively, so that the base 28-1 can slide along the two stainless steel tubes 4-3. The sliding of the base 28-1 along the two stainless steel tubes 4-3 is sliding in the transverse direction of the stainless steel tube conveyor belt 4. Figure 1 and 2Many trays 28 are connected with the stainless steel tubes on the stainless steel tube conveyor belt 4. When the stainless steel tube conveyor belt 4 is working, the stainless steel tubes move in the longitudinal direction, thereby carrying the trays 28 to move in the longitudinal direction. The trays can be used for height difference-free sorting operations to avoid damage to fragile fruits and vegetables or fruits and vegetables with high appearance requirements.

[0115] In order to reduce the weight of the tray 28, the first slide groove 28-1-1 and the second slide groove 28-1-2 of the base 28-1 are formed by eight ribs arranged in a line. Fig.14 and 17 and 16, the base 28-1 is provided with eight ribs 28-1-3, and two symmetrically arranged U-shaped grooves are provided on both sides of each rib 28-1-3, so that the eight U-shaped grooves on one side of the eight ribs form the first slide 28-1-1, and the eight U-shaped grooves on the other side of the eight ribs form the second slide 28-1-2. The design of multiple ribs can reduce the weight while ensuring the matching of the slide and the stainless steel pipe. The light weight of the tray is conducive to reducing the inertia of movement, reliable sliding, and improving the efficiency of the sorting operation.

[0116] like Fig.19 and 22 As shown, during the input stage of the entire operation process of the turnout type fruit and vegetable sorting device, the guide shaft 28 - 2 of the tray 28 moves along the vertical portion 11 - 1 of the main track 11 . Fig. 22 In the figure, the direction indicated by the upper arrow is the moving direction of the upper tray, and the direction indicated by the lower arrow is the moving direction of the lower tray.

[0117] The working process of the above-mentioned turnout type fruit and vegetable sorting device is described below:

[0118] refer to Fig.24 A cucumber 29 is placed on the multiple contoured lifting plates 28-3 of each tray 28, the driving mechanism is started, the stainless steel tubular conveyor belt 4 circulates, and the multiple trays on the upper layer move in the direction indicated by the arrow in the figure. In the input stage of the working process, the guide shaft 28-2 of the tray moves along the vertical portion 11-1 of the main track 11 (such as Fig. 22 shown).

[0119] The cucumbers on the tray at the input stage are graded by machine vision technology. The controller sends control instructions to the electromagnet 1 24 and the electromagnet 2 27 according to the grade data, so that the electromagnet 1 24 and the electromagnet 2 27 are activated. The reversing finger 1 22 is not connected with the rear end of the main track 11 in the initial state (such as Figure 5 , 7 and 19), the rotating shaft of the electromagnet 24 rotates a certain angle, so that the free end of the reversing finger 22 is close to the rear end of the main track 11 (as shown in FIG. Fig.10As shown in FIG. 1 , the guide shaft of the tray moves along the reversing finger 22 to the diverting track 14, and the guide shaft slides against the vertical portion 14-1 of the diverting track 14, and finally the tray moves obliquely forward along the diverting track 14 (toward the output end of the switch type fruit and vegetable sorting device). Fig.24 The trays where cucumbers 30, 31, 32, 33, 34 and 35 are located are all moving along the diversion track 14. The guide shafts of these trays move from the rear end of the diversion track 14 to the grading track 16 and slide against the vertical part 16-1 (moving in the longitudinal direction). When the rotating shaft of the electromagnet 24 is reversed by a certain angle to return the reversing finger 22 to the initial state, no trays will be diverted to the diversion track 14 at this time, and the trays on the main track 11 will move directly from the rear end of the main track 11 to the front end of the segmented main track 12, and continue to move forward along the segmented main track 12. Fig.24 The trays on which cucumbers 36 and 37 are shown are both moved along the segmented main track 12.

[0120] For the pallet on the segmented main track 12, the reversing finger 25 is not connected with the rear end of the segmented main track 12 in the initial state, and the rotating shaft of the electromagnet 27 rotates a certain angle to make the free end of the reversing finger 25 close to the rear end of the segmented main track 12 (such as Figure 8 As shown in FIG. 1 , the guide shaft of the tray moves along the reversing finger 25 to the diverting track 15, and the guide shaft slides against the vertical portion 15-1 of the diverting track 15, and finally the tray moves obliquely forward along the diverting track 15 (toward the output end of the switch type fruit and vegetable sorting device). Fig.24 Cucumbers 38, 39, 40 and 41 shown in the figure all move along the diverter track 15. The guide shafts of these trays move from the rear end of the diverter track 15 to the grading track 17 and slide against the vertical part 17-1 (move in the longitudinal direction). When the rotating shaft of the electromagnet 27 is reversed by a certain angle to return the reversing finger 25 to the initial state, no tray will be diverted to the diverter track 15. The tray on the segmented main track 12 will move directly from its rear end to the front end of the segmented main track 13 and continue to move forward along the segmented main track 13. Fig.24 The displayed pallet 42, pallet 43, pallet 44, pallet 45, pallet 46, pallet 47 and pallet 48 all move along the segmented main track 13.

[0121] It can be seen that at the output end of the switch-type fruit and vegetable sorting device, three positions located in the horizontal direction output three grades of cucumbers respectively. The cucumbers placed on tray 42, tray 43, tray 44, tray 45, tray 46, tray 47 and tray 48 belong to grade one, cucumber 38, cucumber 39, cucumber 40 and cucumber 41 belong to grade two, and cucumber 30, cucumber 31, cucumber 32, cucumber 33, cucumber 34 and cucumber 35 belong to grade three.

[0122] After the cucumbers at the three positions at the output end of the turnout type fruit and vegetable sorting device are taken away or dropped, the corresponding trays are moved to the lower layer, and the guide shafts of trays 42, 43, 44, 45, 46, 47 and 48 are moved along the merging track 19, and then moved to the rear end of the merging track 19, turned upward and then moved to the front end of the main track 11 to achieve circulation. The guide shafts of the trays where cucumbers 30, 31, 32, 33, 34 and 35 are located are separated from the rear end of the grading track 16, and then moved to the front end of the return track 20 and slid against the vertical part of the return track 20, and slid all the way to the merging track 19, and then turned upward from the rear end of the merging track 19 and moved to the front end of the main track 11.

[0123] The guide shaft of the tray where cucumbers 38, cucumber 39, cucumber 40 and cucumber 41 are located detaches from the rear end of the grading track 2 17, then moves to the front end of the return track 21 and slides against the vertical part of the return track 21, slides all the way to the merging track 19, then flips upward from the rear end of the merging track 19 and moves to the front end of the main track 11.

[0124] It can be seen that due to the rapid response and action of the electromagnet, the corresponding reversing finger moves quickly, and the tray is diverted quickly, which can meet the requirements of rapid movement of the tray and the high efficiency of grading at the output end of the turnout type fruit and vegetable sorting device. The entire sorting process is completed without height difference, the cucumbers will not roll over, will not cause damage to the cucumbers, and can provide good protection for the cucumbers.

[0125] There is no height difference and no strong impact force during the sorting process.

[0126] It should be noted that, for the tray 28, the shapes and sizes of the grooves on the multiple contoured lifting plates 28-3 can be changed to accommodate lifting of other stick-shaped fruits and vegetables except cucumbers.

[0127] It should be noted that, for the tray 28, other structural lifting components can also be used to replace the multiple contoured lifting plates 28-3 to achieve the lifting function of stick-shaped fruits and vegetables such as cucumbers, or to achieve the lifting of fruits and vegetables other than stick-shaped fruits and vegetables, or to achieve the lifting of items other than fruits and vegetables (the items can be any items that can be lifted).

[0128] It should be noted that the material of the tray can be fiber-reinforced nylon to increase wear resistance.

[0129] Regarding the process of classifying cucumbers on the tray at the input stage according to their quality by machine vision technology, the present invention discloses a cucumber grading method based on computer vision with higher grading accuracy and efficiency, referring to Fig.33 , mainly including the following steps:

[0130] The first step is to establish a cucumber key point detection image dataset.

[0131] Step S101, image acquisition. The tray directly below the cucumber is photographed by a camera from a top-down perspective, a large number of cucumbers are prepared, and the cucumbers are placed one by one on the tray in sequence, and the camera is controlled to capture an image of each cucumber to obtain a cucumber image. Fig.25 This is one of the cucumber images collected.

[0132] Step S102, image annotation. For the collected cucumber image data, the cucumber key points in the image are annotated for training the cucumber key point detection model. First, the cucumber key points are defined. The cucumber grade division of the Chinese cucumber grading standard is specified in Table 1 below.

[0133] Table 1

[0134]

[0135] According to the cucumber classification regulations in Table 1, cucumbers are mainly graded based on color, damage degree and quantitative indicators. The color and damage degree of cucumbers can create a deep learning classification network to classify cucumbers in the image according to the grade classification standards. The quantitative indicators include the cucumber bow height, transverse diameter difference and cucumber handle ratio, so the present invention defines 10 key points of cucumbers based on quantitative indicators, such as Fig.26 As shown, points 1-10 in the figure are respectively represented as: point 1 is the top of the cucumber, point 2 is the handle end of the cucumber, point 3 and point 4 are edge points 3 cm away from the top of the cucumber, point 5 and point 6 are edge points in the middle of the cucumber, point 7 and point 8 are edge points 3 cm away from the handle end, point 9 is the maximum curvature of the cucumber, and point 10 is the midpoint of the boundary between the handle and the body of the cucumber.

[0136] Secondly, the key points of the cucumber image are annotated. The specific annotation process can be: step (1), use OpenCV to automatically detect the tray in the image, obtain the correspondence between the image pixels and the actual size according to the actual size of the tray, and then obtain the actual size of a pixel in the image; step (2), start to annotate the key points, and the annotation order is: step 1), mark the top position point of the cucumber as point 1, and mark the handle position point of the cucumber as point 2. Step 2), click the top position point of the cucumber and move a distance along the direction of the cucumber body and then release it. The line connecting the click point and the current mouse position will be displayed in real time during movement (such as Fig. 27 After releasing the mouse, draw an auxiliary line perpendicular to the melon body 3 cm from the top of the melon (as shown in the black line in the middle) to assist the operator in determining the appropriate position. Fig. 27 As shown by the purple line in the figure, the purple auxiliary line is perpendicular to the black line, that is, perpendicular to the "line connecting the click point and the current mouse position"). 3cm is determined according to the actual size of the pixel. Step 3), click the end of the cucumber and move a distance along the direction of the cucumber body and then release it. The line connecting the click point and the current mouse position will be displayed in real time while moving (as shown in the figure below). Fig.28 As shown by the black line on the left in the figure, the purple auxiliary line is perpendicular to the black line, that is, perpendicular to the "line connecting the click point and the current position of the mouse"). After releasing the mouse, draw an auxiliary line perpendicular to the melon body 3cm away from the handle end (as shown in the figure below). Fig.28 3cm is determined based on the actual size of the pixel. Step 4), based on the position of the top and handle of the melon, draw an auxiliary line perpendicular to the line between the top and handle of the melon (as shown in the purple line on the left). Fig.28 Step 5), determine the positions of point 3, point 4, point 5, point 6, point 7, and point 8 in turn according to the three auxiliary lines obtained. Step 6), mark the maximum bending point of the cucumber as point 9, and mark the midpoint of the boundary between the handle and the body of the cucumber as point 10. Fig.29 As shown. So far, the key point annotation of a cucumber image has been completed. After annotating all cucumber images according to the above method, a cucumber key point detection image dataset is constructed.

[0137] The second step is to train the cucumber key point detection model.

[0138] The cucumber key point detection dataset is divided into a training set and a validation set according to an 8:2 ratio. The training set is used to train the cucumber key point detection model, and the validation set is used to evaluate the performance of the trained model. Key point detection is mainly divided into two ways, one is top-down and the other is bottom-up. From top to bottom, first find the target, perform bounding box detection on the target, and then find the key points of the target in the bounding box. From bottom to top, first find the key points, then summarize, first find the key points of all targets in the image, and then classify the key points belonging to the same target into one category. Based on the consideration of detection speed and accuracy, the present invention provides a method based on YOLOv8-pose (such as Fig.30 ) and RTMPose (such as Fig.31 ) Two cucumber keypoint detection models. YOLOv8 is a popular real-time object detection algorithm known for its speed and accuracy. YOLOv8-pose is an extension of YOLOv8, designed for keypoint detection, and can detect object keypoints in real time. RTMPose adopts a top-down approach, using detectors to obtain the object bounding box, and then calculates the keypoints of each object separately. Due to the additional use of detectors and the calculation of keypoints for each object separately, the cucumber keypoint detection model based on RTMPose is more accurate but slower. For example, on the same device with Intel Core i9 CPU and NVIDIA GeForce RTX 2080Ti GPU, the processing speed of a single image on the YOLOv8-pose model is within 20ms, while the RTMPose model takes about 130ms. In actual use, you can choose according to hardware conditions and specific business needs.

[0139] The third step is to calculate the cucumber grading index.

[0140] After the cucumber key point detection model training is completed, the cucumber images collected by the camera during the actual operation of the turnout fruit and vegetable sorting device are input into the model, and the key points of the cucumber in the image can be predicted. The prediction results are as follows: Fig.32 As mentioned above, the real size of the image pixel can be calculated based on the reference object (such as a pallet) in the image. In addition, when the distance between the camera and the pallet at the shooting position is always kept fixed, the real size of the pixel can also be obtained. According to the prediction results ( Fig.32 In order to calculate each index, the following definitions are given first:

[0141] D(a,b) represents the distance between point a and point b.

[0142] D(a,b_c) represents the vertical distance from point a to line bc, (x a ,y a) represents the image coordinates of point a. The coordinates of point 9 are: (x9, y9).

[0143] The length of the cucumber can be approximated by the sum of the distances from the maximum bend of the cucumber to the top of the cucumber and to the handle of the cucumber. The calculation formula is as follows (1):

[0144] D(melon length)=D(1,9')+D(2,9') (1)

[0145] In formula (1), 9' is the distance between point 5 and point 6, which is the point 9 at the maximum bend of the cucumber moved to the middle of the cucumber. Fig.34 , the coordinates of point 9' are:

[0146]

[0147] The transverse diameter difference is calculated by the following formula (2):

[0148] D(diameter difference) = max(|D(3,4)-D(5,6)|,|D(7,8)-D(5,6)|) (2)

[0149] The melon-to-handle ratio is calculated using the following formula (3):

[0150]

[0151] The bow height per 10 cm is calculated using the following formula (4):

[0152]

[0153] The indicator distances indicated above are all pixel distances. The real distance of each indicator can be obtained through the real size of the pixel.

[0154] For example, the distance between point 1 and point 9' is line segment B, the distance between point 2 and point 9' is line segment A, the distance between point 2 and point 10 is line segment C, the distance between point 1 and point 2 is line segment D, the straight line between point 1 and point 2 is 1_2, the perpendicular distance from point 9' to line 1_2 is line segment E, the distance from point 3 to point 4 is line segment F, the distance from point 5 to point 6 is line segment G, and the distance from point 7 to point 8 is line segment H, then:

[0155] D(melon length)=A+B

[0156] D (melon-to-handle ratio) = C / (A+B)

[0157] D (arch height) = E × 10 / D

[0158] D (transverse diameter difference) = max(|HG|,|FG|).

[0159] The fourth step is to determine the level.

[0160] The grade of the cucumbers in the image is determined according to the provisions of the national grading standards. Those with bow height D (bow height) ≤ 0.5 cm, transverse diameter difference D (transverse diameter difference) ≤ 0.5 cm, and handle ratio D (handle ratio) ≤ 1 / 8 are special grade.

[0161] The first grade is 0.5cm<D (bow height) ≤1cm, the transverse diameter difference D (transverse diameter difference) ≤1cm, and the melon-handle ratio D (melon-handle ratio) ≤1 / 7.

[0162] The second level is 1cm<D (bow height) ≤2cm, and the transverse diameter difference D (transverse diameter difference) ≤2cm, and the melon-handle ratio D (melon-handle ratio) ≤1 / 6.

[0163] Finally, the visual system sends the cucumber grade information to the controller for controlling the mechanical action through serial communication and other methods.

[0164] It can be seen that the above cucumber grading method is more accurate, faster, more efficient, more stable and not easily affected by ambient light intensity and image background.

[0165] It should be noted that the fourth step of judging the grade can also be combined with the cucumber color and damage degree for grade judgment. The cucumber color and damage degree can be determined by creating a deep learning classification network. The creation of a deep learning classification network can be completed in the second step above, using the pre-built cucumber appearance classification data set to train the model. The cucumber appearance classification data set is divided into a training set and a validation set in an 8:2 ratio. The training set is used to train the cucumber appearance classification model based on the convolutional neural network, and the validation set is used to evaluate the performance of the trained model.

[0166] The cucumber appearance classification dataset can be constructed in the first step above. The specific process is to let the camera look at the cucumber from a 45° angle on the left and right sides. Fig.35 The display shows a 45° viewing angle on the right side of the cucumber) to collect images of the cucumbers on the tray, and then use standard tools to annotate the collected cucumber images. The cucumbers with good glossy color and no damage are marked as special grade, the cucumbers with glossy color and slight damage are marked as first grade, and the cucumbers with dull color and a little damage are marked as second grade, thereby obtaining the cucumber appearance classification dataset.

[0167] Therefore, the following processing is added to the third step: the cucumber image collected by the camera during the actual operation of the turnout fruit and vegetable sorting device is input into the cucumber appearance classification model based on the convolutional neural network. Fig.36 ,The cucumber appearance classification model based on convolutional neural network outputs the appearance category levels: special grade, first grade, and second grade.

[0168] Therefore, the specific process of the fourth step of judging the grade based on the color and damage of the cucumber is:

[0169] Step S401, determine the grade of the cucumber in the image according to the provisions of the national grading standards, and the cucumbers with bow height D (bow height) ≤ 0.5cm, transverse diameter difference D (transverse diameter difference) ≤ 0.5cm, and handle ratio D (handle ratio) ≤ 1 / 8 are special grade.

[0170] The first grade is 0.5cm<D (bow height) ≤1cm, the transverse diameter difference D (transverse diameter difference) ≤1cm, and the melon-handle ratio D (melon-handle ratio) ≤1 / 7.

[0171] The second level is 1cm<D (bow height) ≤2cm, and the transverse diameter difference D (transverse diameter difference) ≤2cm, and the melon-handle ratio D (melon-handle ratio) ≤1 / 6.

[0172] Step S402, combining the appearance category level output by the cucumber appearance classification model based on the convolutional neural network with the level of the quantitative index in step S401, and determining the lower level or the same level as the final level. For example: if the appearance category level of a cucumber is special grade and the level of the quantitative index is level one, the final level is level one. For another example: if the appearance category level of a cucumber is level one and the level of the quantitative index is level two, the final level is level two. For another example, if the appearance category level of a cucumber is level one and the level of the quantitative index is special grade, the final level is level one. For another example. If the appearance category level of a cucumber is level one and the level of the quantitative index is level one, the final level is level one.

[0173] The above description is only for the preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes.

Claims

1. A switch type sorting device, characterized in that: It includes a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt, a switch-type diverting mechanism and a plurality of trays, wherein the driving mechanism is connected to the supporting frame; The stainless steel tube conveyor belt comprises a left chain, a right chain and a plurality of stainless steel tubes, wherein the plurality of stainless steel tubes are connected between the left chain and the right chain; the driving mechanism is used to drive the right chain and the left chain to move; The switch-type diversion mechanism includes a top-level track support tube, a main track, a segmented main track 1, a segmented main track 2, a diversion track 1, a diversion track 2, a graded track 1, a graded track 2, a bottom-level track support tube, a merging track, a return track 1, a return track 2, a reversing finger 1, an electromagnet connection seat 1, an electromagnet 1, a reversing finger 2, an electromagnet connection seat 2 and an electromagnet 2, a plurality of top-level track support tubes are fixedly connected to the support frame, a plurality of bottom-level track support tubes are fixedly connected to the support frame, a plurality of bottom-level track support tubes are located on a horizontal plane B, a plurality of top-level track support tubes are located on a horizontal plane A, and the horizontal plane B is located below the horizontal plane A; the main track, the segmented main track 1 and the segmented main track 2 are respectively connected to a plurality of top-level track support tubes. The support pipe is fixedly connected, the main track, the segmented main track one and the segmented main track two are located on the same straight line and close to one side of the support frame, a distance is provided between the front end of the segmented main track one and the rear end of the main track, a distance is provided between the front end of the segmented main track two and the rear end of the segmented main track one, the main track, the segmented main track one and the segmented main track two are distributed along the longitudinal direction, the shunt track one is fixedly connected to multiple top-level track support pipes, the shunt track one is tilted with respect to the straight line where the main track is located, the front end of the shunt track one is close to the rear end of the main track, the electromagnet connecting seat one is connected to the top-level track support pipe, the electromagnet one is connected to the electromagnet connecting seat one, the reversing finger one is connected to the electromagnet The first iron shaft is connected, and the reversing finger one is located between the rear end of the main track and the front end of the shunt track one; the shunt track two is fixedly connected to a plurality of top-level track support tubes, and the shunt track two is tilted with the straight line on which the segmented main track one is located as a reference, and the front end of the shunt track two is close to the rear end of the segmented main track one, and the shunt track two is located between the shunt track one and the segmented main track two, the second electromagnet connecting seat is connected to the top-level track support tube, the second electromagnet is connected to the second electromagnet connecting seat, the second reversing finger two is connected to the rotating shaft of the second electromagnet, and the second reversing finger two is located between the front end of the shunt track two and the rear end of the segmented main track one; the first graded track is fixedly connected to a plurality of top-level track support tubes, and the graded track one is close to the Near the rear end of the shunt track 1, the graded track 1 is located outside the rear end of the shunt track 1, and the graded track 1 is parallel to the main track; the graded track 2 is fixedly connected to a plurality of top track support tubes, the graded track 2 is close to the rear end of the shunt track 2, the graded track 2 is located outside the rear end of the shunt track 2, and the graded track 2 is parallel to the segmented main track 1; the merging track is fixedly connected to a plurality of bottom track support tubes, along the vertical direction, the merging track is located directly below the main track, the return track 1 is fixedly connected to a plurality of bottom track support tubes, the return track 1 is arranged obliquely with the straight line where the merging track is located, the rear end of the return track 1 is close to the merging track, and the front end of the return track 1 is located below the rear end of the graded track 1;The return track 2 is fixedly connected to a plurality of bottom track support tubes, and the return track 2 is arranged obliquely with respect to the straight line where the merging track is located, and the rear end of the return track 2 is close to the merging track, and the front end of the return track 2 is located below the rear end of the grading track 2; the main track, the segmented main track 1, the segmented main track 2, the diversion track 1, the diversion track 2, the grading track 1, the grading track 2, the merging track, the return track 1 and the return track 2 are all located in the space surrounded by the stainless steel tubular conveyor belt; The tray comprises a base, a lifting component and a guide shaft, wherein the lifting component is connected to the top of the base, and the guide shaft is connected to the bottom of the base; the base of the tray is slidably connected to the two stainless steel tubes of the stainless steel tube conveyor belt; The guide shaft can move along the main track, the guide shaft can move along the segmented main track one, the guide shaft can move along the segmented main track two, the guide shaft can move along the diversion track one, the guide shaft can move along the diversion track two, the guide shaft can move along the graded track one, the guide shaft can move along the graded track two, the guide shaft can move along the regression track one, the guide shaft can move along the regression track two, and the guide shaft can move along the merging track.

2. The switch type sorting device according to claim 1, characterized in that: The base of the tray is provided with two slide grooves, and the two slide grooves are respectively slidably connected to the two stainless steel pipes.

3. The switch type sorting device according to claim 2, characterized in that: The lifting components are a plurality of contoured lifting plates, and the plurality of contoured lifting plates are arranged in a line.

4. The switch type sorting device according to claim 1, characterized in that: The driving mechanism includes a driving motor, a driving shaft, a driven shaft, a first driving sprocket, a second driving sprocket, a first driven sprocket and a second driven sprocket, the driving motor is connected to a supporting frame, the driving shaft is rotatably connected to the supporting frame, the driven shaft is rotatably connected to the supporting frame, the output shaft of the driving motor is connected to the driving shaft, the first driving sprocket is connected to the right end of the driving shaft, the second driving sprocket is connected to the left end of the driving shaft, the first driven sprocket is connected to the right end of the driven shaft, and the second driven sprocket is connected to the left end of the driven shaft; the left chain of the stainless steel tubular conveyor belt is connected between the second driving sprocket and the second driven sprocket, and the right chain is connected between the first driving sprocket and the first driven sprocket.

5. The switch type sorting device according to claim 1, characterized in that: The main track is provided with a vertical portion, the segmented main track one is provided with a vertical portion, the segmented main track two is provided with a vertical portion, the diversion track one is provided with a vertical portion, the graded track one is provided with a vertical portion, the diversion track two is provided with a vertical portion, the graded track two is provided with a vertical portion, the merging track is provided with a vertical portion, the return track one is provided with a vertical portion, and the return track two is provided with a vertical portion.

6. A switch type sorting device, characterized in that: It includes a supporting frame, a driving mechanism, a stainless steel tubular conveyor belt, a switch-type diverting mechanism and a plurality of trays, wherein the driving mechanism is connected to the supporting frame; The stainless steel tube conveyor belt comprises a left chain, a right chain and a plurality of stainless steel tubes, wherein the plurality of stainless steel tubes are connected between the left chain and the right chain; the driving mechanism is used to drive the right chain and the left chain to move; The switch-type diversion mechanism includes a top-level rail support tube, a main track, a segmented main track, a diversion track, a graded track, a bottom-level rail support tube, a merging track, a return track, a reversing finger, an electromagnet connection seat and an electromagnet. A plurality of top-level rail support tubes are fixedly connected to the support frame, a plurality of bottom-level rail support tubes are fixedly connected to the support frame, a plurality of bottom-level rail support tubes are located on a horizontal plane B, a plurality of top-level rail support tubes are located on a horizontal plane A, and the horizontal plane B is located below the horizontal plane A; the main track and the segmented main track are respectively fixedly connected to a plurality of top-level rail support tubes, the main track and the segmented main track are located on the same straight line and close to one side of the support frame, a spacing is provided between the front end of the segmented main track and the rear end of the main track, the main track and the segmented main track are distributed along the longitudinal direction, the diversion track is fixedly connected to a plurality of top-level rail support tubes, the diversion track is arranged tilted with respect to the straight line where the main track is located, and the front end of the diversion track is close to the main track. The rear end of the track, the electromagnet connecting seat 1 is connected to the top track support tube, the electromagnet 1 is connected to the electromagnet connecting seat 1, the reversing finger 1 is connected to the rotating shaft of the electromagnet 1, and the reversing finger 1 is located between the rear end of the main track and the front end of the diversion track 1; the grading track 1 is fixedly connected to multiple top track support tubes, the grading track 1 is close to the rear end of the diversion track 1, the grading track 1 is located outside the rear end of the diversion track 1, and the grading track 1 is parallel to the main track; the merging track is fixedly connected to multiple bottom track support tubes, along the vertical direction, the merging track is located directly below the main track, the return track 1 is fixedly connected to multiple bottom track support tubes, the return track 1 is arranged obliquely with the straight line where the merging track is located, the rear end of the return track 1 is close to the merging track, and the front end of the return track 1 is located below the rear end of the grading track 1; the main track, the segmented main track 1, the diversion track 1, the grading track 1, the merging track and the return track 1 are all located in the space surrounded by the stainless steel tubular conveyor belt; The tray comprises a base, a lifting component and a guide shaft, wherein the lifting component is connected to the top of the base, and the guide shaft is connected to the bottom of the base; the base of the tray is slidably connected to the two stainless steel tubes of the stainless steel tube conveyor belt; The guide shaft can move along the main track, the guide shaft can move along the segmented main track, the guide shaft can move along the diversion track, the guide shaft can move along the graded track, the guide shaft can move along the return track, and the guide shaft can move along the merging track.

7. The switch type sorting device according to claim 6, characterized in that: The main track is provided with a vertical portion, the segmented main track 1 is provided with a vertical portion, the diversion track 1 is provided with a vertical portion, the graded track 1 is provided with a vertical portion, the merging track is provided with a vertical portion, and the return track 1 is provided with a vertical portion.

8. A sorting method using the switch type sorting device according to claim 1, characterized in that: After the turnout sorting device starts working, the fruits and vegetables on the tray at the input stage are graded to obtain grade data, and then the electromagnet 1 or electromagnet 2 is controlled to operate according to the grade data to sort the fruits and vegetables; The process of obtaining grade data by grading is carried out according to the following steps: The first step is to establish a fruit and vegetable key point detection image dataset; Step S101, image acquisition, using a camera to shoot fruits and vegetables on a tray directly below it from a bird's-eye view, to acquire a large number of fruit and vegetable images; Step S102, image annotation, annotating key points in the collected fruit and vegetable image data to construct a fruit and vegetable key point detection image data set; The second step is to use the fruit and vegetable key point detection image dataset to train the YOLOv8-pose model or RTMPose model to obtain the fruit and vegetable key point detection model; The third step is to calculate the fruit and vegetable grading index; Input the images of fruits and vegetables to be graded collected by the camera into the key point detection model of fruits and vegetables to obtain the key point recognition results; Calculate the quantitative size index based on the obtained key points; The fourth step is to determine the grade based on the quantitative size index.

9. A sorting method according to claim 8 using the switch type sorting device according to claim 1, characterized in that: The grading to obtain grade data is performed on cucumbers; in step S102, there are 10 key points, point 1 is the top of the cucumber, point 2 is the handle end of the cucumber, points 3 and 4 are edge points 3 cm away from the top of the cucumber, points 5 and 6 are edge points in the middle of the cucumber, points 7 and 8 are edge points 3 cm away from the handle end of the cucumber, point 9 is the maximum bend of the cucumber, and point 10 is the midpoint of the boundary between the handle and the body of the cucumber; In the third step, the key point recognition result has 10 key points, and the quantitative size index is calculated according to the 10 key points. The quantitative size index includes D (bow height), D (transverse diameter difference), and D (melon handle ratio); D (arch height) is calculated by the following formula (1): D (diameter difference) is calculated by the following formula (2): D(diameter difference)=max(|D(3,4)-D(5,6)|,|D(7,8)-D(5,6)|) (2); D (melon-to-grain ratio) is calculated by the following formula (3): In formula (3), D (melon length) is calculated by the following formula (4): D(melon length) = D(1,9') + D(2,9') (4); In the above formulas (1), (2) and (3), D(1,2) represents the distance between point 1 and point 2, D(2,10) represents the distance between point 2 and point 10, D(3,4) represents the distance between point 3 and point 4, D(5,6) represents the distance between point 5 and point 6, D(7,8) represents the distance between point 7 and point 8, 9' is the distance between point 5 and point 6, moving point 9 at the maximum bending point of the cucumber to the middle of the cucumber, D(1,9') represents the distance between point 1 and point 9', D(2,9') represents the distance between point 2 and point 9', and D(9',1_2) represents the perpendicular distance from point 9' to the straight line between point 1 and point 2.

10. A sorting method using the switch type sorting device according to claim 1, characterized in that: After the turnout sorting device starts working, the fruits and vegetables on the tray at the input stage are graded to obtain grade data, and then the electromagnet 1 and / or electromagnet 2 are controlled to operate according to the grade data to sort the fruits and vegetables; The process of obtaining grade data by grading is carried out according to the following steps: The first step is to establish a fruit and vegetable key point detection image dataset; Step S101, image acquisition, using a camera to shoot fruits and vegetables on a tray directly below it from a bird's-eye view, to acquire a large number of fruit and vegetable images; Step S102, image annotation, annotating key points in the collected fruit and vegetable image data to construct a fruit and vegetable key point detection image data set; The second step is to establish an appearance classification dataset. The fruits and vegetables on the tray are photographed by a camera from a top and side view to collect a large number of fruit and vegetable images, and then the fruits and vegetables are labeled to obtain an appearance classification dataset. The third step is to train the neural network model using the fruit and vegetable key point detection image dataset to obtain the fruit and vegetable key point detection model, and use the appearance classification dataset to train the convolutional neural network model to obtain the appearance classification model based on the convolutional neural network; The fourth step is (1) calculating the fruit and vegetable grading index; inputting the image of the fruit and vegetable to be graded collected by the camera into the fruit and vegetable key point detection model to obtain the key point recognition result, and calculating the quantitative size index based on the obtained key points; (2) inputting the images of fruits and vegetables to be graded collected by the camera into an appearance classification model based on a convolutional neural network, and the appearance classification model based on the convolutional neural network outputs the appearance category level; Step 5: Determine the grade based on the quantitative size index; Step 6. Combine the appearance category grade with the grade determined in step 5 and determine the lower grade or the same grade as the final grade.

Images