Slipper sorting and counting method and system based on AI

Abstract

The invention provides a slipper sorting and counting method and system based on AI in the technical field of slipper sorting management. The method comprises the steps that S1, a large number of historical slipper images are collected to construct a data set; s2, creating a slipper identification model; s3, training a slipper identification model by using the data set, and deploying the slipper identification model to a sorting robot; s4, the industrial personal computer obtains the production instruction list and analyzes the production instruction list to obtain SKU codes and the sorting number, corresponding induction counting tables are started, all the SKU codes, the sorting number and the number of the counting tables are bound to generate a binding relation, and the binding relation is sent to a sorting robot; s5, the sorting robot collects real-time slipper images and inputs the real-time slipper images into the slipper recognition model to obtain SKU codes, and the slippers are sorted to the corresponding induction counting tables based on the SKU codes and the binding relation; and S6, counting the slippers by the sensing counting table through a photoelectric sensor, and checking a counting result through a visual counting camera. The slipper sorting and counting device has the advantage that the efficiency and quality of slipper sorting and counting are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of slipper sorting management, and in particular to an AI-based slipper sorting and counting method and system. Background Art

[0002] With the improvement of people's living standards and the change of consumption concepts, the demand for slippers has not only increased in quantity, but also become more diversified in style, material, function, etc., which makes the types and models of slippers increasingly rich, and the difficulty and complexity of sorting and counting work also increase accordingly.

[0003] Traditionally, the sorting and counting of slippers requires manual sorting of the slippers flowing out of the injection molding machine, and classification and stacking according to specifications, types, colors, etc. After stacking, manual counting is performed based on the order to transfer the required quantity to the next production link. This has the following disadvantages: manual sorting is inefficient, and due to long-term fatigue work, errors in sorting and counting are prone to occur.

[0004] Therefore, how to provide an AI-based slipper sorting and counting method and system to improve the efficiency and quality of slipper sorting and counting has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide an AI-based slipper sorting and counting method and system to improve the efficiency and quality of slipper sorting and counting.

[0006] In a first aspect, the present invention provides an AI-based slipper sorting and counting method, comprising the following steps:

[0007] Step S1, collecting a large number of historical slipper images, and constructing a data set after preprocessing and annotating each of the historical slipper images;

[0008] Step S2: creating a slipper recognition model based on a neural network, and setting a loss function of the slipper recognition model;

[0009] Step S3, dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model by using the training set, testing the trained slipper recognition model by using the test set, verifying the tested slipper recognition model by using the verification set, and deploying the verified slipper recognition model on the sorting robot;

[0010] Step S4: The industrial computer obtains the input production instruction sheet, parses the production instruction sheet to obtain the SKU code and the sorting quantity, starts the corresponding induction counting station based on each SKU code, binds each SKU code, the sorting quantity and the counting station number of the induction counting station to generate a binding relationship, and sends the binding relationship to the sorting robot;

[0011] Step S5: The sorting robot collects real-time images of the slippers to be sorted through a camera, inputs the real-time images into a slipper recognition model to obtain a SKU code, and sorts the slippers to corresponding induction counting stations based on the recognized SKU code and the binding relationship;

[0012] Step S6, the induction counting station counts the sorted slippers through the photoelectric sensor, displays the counting result through the display, and verifies the counting result through the visual counting camera;

[0013] Step S7: After the industrial computer determines that the classification frame in the induction counting table is full based on the counting result, it controls the label printer to print slipper labels and stick them on the classification frame;

[0014] Step S8: The industrial computer circulates the full classification frames through the AGV trolley, and replaces the empty classification frames to continue sorting until the sorting quantity is met.

[0015] Furthermore, the step S1 is specifically as follows:

[0016] A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

[0017] Furthermore, the step S3 is specifically as follows:

[0018] The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then:

[0019] The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

[0020] Furthermore, the step S5 is specifically as follows:

[0021] The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

[0022] Furthermore, the step S6 is specifically as follows:

[0023] The induction counting table counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; if the counting results are incorrect, an alarm of counting error is prompted through the display.

[0024] In a second aspect, the present invention provides an AI-based slipper sorting and counting system, comprising the following modules:

[0025] A data set construction module, used to collect a large number of historical slipper images, and construct a data set after preprocessing and annotating each of the historical slipper images;

[0026] A slipper recognition model creation module, used to create a slipper recognition model based on a neural network and set a loss function of the slipper recognition model;

[0027] a slipper recognition model training module, for dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model through the training set, testing the trained slipper recognition model through the test set, verifying the tested slipper recognition model through the verification set, and deploying the verified slipper recognition model on the sorting robot;

[0028] The production instruction sheet parsing module is used for the industrial computer to obtain the input production instruction sheet, parse the production instruction sheet to obtain the SKU code and the sorting quantity, start the corresponding induction counting table based on each SKU code, bind each SKU code, the sorting quantity and the counting table number of the induction counting table to generate a binding relationship, and send the binding relationship to the sorting robot;

[0029] A slipper sorting module is used for the sorting robot to collect real-time slipper images of slippers to be sorted through a camera, input the real-time slipper images into a slipper recognition model to obtain a SKU code, and sort the slippers to a corresponding induction counting station based on the recognized SKU code and the binding relationship;

[0030] The slipper counting module is used for the sensing counting platform to count the sorted slippers through the photoelectric sensor, display the counting result through the display, and verify the counting result through the visual counting camera;

[0031] A slipper label printing module, which is used for the industrial computer to control the label printer to print slipper labels and stick them on the classification frame after judging that the classification frame in the induction counting table is full based on the counting result;

[0032] The classification frame circulation module is used for the industrial computer to circulate the full classification frames through the AGV trolley, and replace the empty classification frames to continue sorting until the sorting quantity is met.

[0033] Furthermore, the data set construction module is specifically used for:

[0034] A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

[0035] Furthermore, the slipper recognition model training module is specifically used for:

[0036] The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then:

[0037] The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

[0038] Furthermore, the slippers sorting module is specifically used for:

[0039] The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

[0040] Furthermore, the slippers counting module is specifically used for:

[0041] The induction counting table counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; if the counting results are incorrect, an alarm of counting error is prompted through the display.

[0042] The advantages of the present invention are:

[0043] 1. Build a data set by collecting a large number of historical slipper images and preprocessing and annotating them; divide the data set into training set, test set and verification set based on the preset ratio; train the slipper recognition model created based on the neural network through the training set, test the trained slipper recognition model through the test set, verify the tested slipper recognition model through the verification set, and deploy the verified slipper recognition model to the sorting robot; then the industrial computer obtains the input production instruction sheet and parses it to obtain the SKU code and the sorting quantity, starts the corresponding induction counting table based on each SKU code, binds each SKU code, the sorting quantity and the counting table number of the induction counting table to generate a binding relationship and sends it to the sorting robot; the sorting robot collects the real-time slipper image of the slippers to be sorted through the camera and inputs it into the slipper recognition model to obtain the SKU code, based on the recognized SKU The slippers are sorted to the corresponding induction counting station based on the coding and binding relationship; then the induction counting station counts the sorted slippers through the photoelectric sensor, displays the counting result through the display, verifies the counting result through the visual counting camera, and judges that the classification box in the induction counting station is full based on the counting result, controls the label printer to print the slipper label and paste it on the classification box, circulates the full classification box through the AGV car, and replaces the empty classification box to continue sorting until the sorting quantity is met; that is, the slippers are automatically identified and sorted by the sorting robot deployed with the pre-trained slipper recognition model, the sorted slippers are automatically counted by the photoelectric sensor of the induction counting station, and the counting result is verified by the visual counting camera, that is, the automatic sorting and counting of slippers is realized to replace the traditional manual sorting, thereby greatly improving the efficiency and quality of slipper sorting and counting.

[0044] 2. By collecting a large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles, each historical slipper image is preprocessed with sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each historical slipper image after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set, which effectively improves the diversity of the data set. The data set is then used to train the slipper recognition model, which effectively improves the recognition accuracy and generalization ability of the slipper recognition model.

[0045] 3. By dividing the data set into training set, test set and validation set, the slipper recognition model is trained with the training set until the loss value of the loss function is less than the preset loss threshold, the trained slipper recognition model is tested with the test set until the recognition accuracy is greater than the preset accuracy threshold, the tested slipper recognition model is verified with the validation set until the confidence is greater than the preset confidence threshold, and then the verified slipper recognition model is deployed on the sorting robot. That is, during the training process of the slipper recognition model, continuous testing and verification are performed until the requirements are met, thereby greatly improving the recognition accuracy of the slipper recognition model.

[0046] 4. By collecting real-time slipper images of the slippers to be sorted, the real-time slipper images are pre-processed including at least noise reduction, contrast enhancement, and edge enhancement, and then input into the slipper recognition model to recognize the slippers, that is, recognition is performed based on high-quality real-time slipper images, which effectively improves the recognition accuracy of the slipper recognition model.

[0047] 5. The sorted slippers are counted through several photoelectric sensors, and the perceptual signals within the preset time threshold are combined and counted to avoid repeated counting of the same pair of shoes and to avoid individual photoelectric sensors being blocked and unable to count. The blanking port is photographed and identified by a visual counting camera to verify the counting results. That is, the photoelectric sensor and the visual counting camera are combined for counting and verification, which greatly improves the accuracy of slipper counting. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention will be further described below in conjunction with embodiments with reference to the accompanying drawings.

[0049] Figure 1 The present invention is a flow chart of an AI-based slipper sorting and counting method.

[0050] Figure 2 It is a structural schematic diagram of an AI-based slipper sorting and counting system of the present invention. DETAILED DESCRIPTION

[0051] The technical solution in the embodiments of the present application has the following overall idea: the slippers are automatically identified and sorted by a sorting robot deployed with a pre-trained slipper recognition model, the sorted slippers are automatically counted by the photoelectric sensor of the induction counting table, and the counting results are verified by the visual counting camera, thereby realizing automatic sorting and counting of slippers to improve the efficiency and quality of slipper sorting and counting.

[0052] Please refer to Figure 1 to Figure 2 As shown, a preferred embodiment of the slippers sorting and counting method based on AI of the present invention comprises the following steps:

[0053] Step S1, collecting a large number of historical slipper images, and constructing a data set after preprocessing and annotating each of the historical slipper images;

[0054] Step S2: creating a slipper recognition model based on a neural network, and setting a loss function of the slipper recognition model;

[0055] Step S3, dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model by using the training set, testing the trained slipper recognition model by using the test set, verifying the tested slipper recognition model by using the verification set, and deploying the verified slipper recognition model on the sorting robot;

[0056] Step S4: The industrial computer obtains the input production instruction sheet, parses the production instruction sheet to obtain the SKU code and the sorting quantity, starts the corresponding induction counting station based on each SKU code, binds each SKU code, the sorting quantity and the counting station number of the induction counting station to generate a binding relationship, and sends the binding relationship to the sorting robot;

[0057] Step S5, the sorting robot collects real-time slipper images of the slippers to be sorted through a camera, inputs the real-time slipper images into a slipper recognition model to obtain a SKU code, and sorts the slippers to a corresponding induction counting table based on the recognized SKU code and the binding relationship; the blanking port of the induction counting table is made of stainless steel, which has the advantages of wear resistance and drop resistance, and each of the induction counting tables can be flexibly combined and spliced;

[0058] Step S6, the induction counting station counts the sorted slippers through the photoelectric sensor, displays the counting result through the display, and verifies the counting result through the visual counting camera;

[0059] Step S7: After the industrial computer determines that the classification box in the induction counting table is full based on the counting result, it controls the label printer to print slipper labels and stick them on the classification box; after the classification box is full, the full frame indicator light provided on the induction counting table lights up to remind the user that the classification box is full.

[0060] Step S8: The industrial computer circulates the full classification frames through the AGV trolley, and replaces the empty classification frames to continue sorting until the sorting quantity is met.

[0061] The step S1 is specifically as follows:

[0062] A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

[0063] By collecting a large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles, each historical slipper image is preprocessed with sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each historical slipper image after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set, which effectively improves the diversity of the data set. The data set is then used to train a slipper recognition model, which effectively improves the recognition accuracy and generalization ability of the slipper recognition model.

[0064] The step S3 is specifically as follows:

[0065] The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then:

[0066] The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

[0067] By dividing the data set into a training set, a test set and a validation set, the slipper recognition model is trained with the training set until the loss value of the loss function is less than the preset loss threshold, the trained slipper recognition model is tested with the test set until the recognition accuracy is greater than the preset accuracy threshold, the tested slipper recognition model is verified with the validation set until the confidence is greater than the preset confidence threshold, and the verified slipper recognition model is then deployed on the sorting robot. That is, during the training process of the slipper recognition model, testing and verification are continuously performed until the requirements are met, thereby greatly improving the recognition accuracy of the slipper recognition model.

[0068] The step S5 is specifically as follows:

[0069] The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

[0070] By collecting real-time slipper images of the slippers to be sorted, the real-time slipper images are pre-processed with at least image enhancement including noise reduction, contrast enhancement, and edge enhancement, and then input into a slipper recognition model to recognize the slippers, that is, recognition is performed based on high-quality real-time slipper images, thereby effectively improving the recognition accuracy of the slipper recognition model.

[0071] The step S6 is specifically as follows:

[0072] The induction counting table counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; if the counting results are incorrect, an alarm of counting error is prompted through the display.

[0073] The sorted slippers are counted through several photoelectric sensors, and the perceptual signals within the preset time threshold are combined and counted to avoid the same pair of shoes being counted repeatedly and to avoid individual photoelectric sensors being blocked and unable to count. The blanking port is photographed and identified by a visual counting camera to verify the counting results. That is, the photoelectric sensor and the visual counting camera are combined for counting and verification, thereby greatly improving the accuracy of slipper counting.

[0074] A preferred embodiment of the slipper sorting and counting system based on AI of the present invention includes the following modules:

[0075] A data set construction module, used to collect a large number of historical slipper images, and construct a data set after preprocessing and annotating each of the historical slipper images;

[0076] A slipper recognition model creation module, used to create a slipper recognition model based on a neural network and set a loss function of the slipper recognition model;

[0077] a slipper recognition model training module, for dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model through the training set, testing the trained slipper recognition model through the test set, verifying the tested slipper recognition model through the verification set, and deploying the verified slipper recognition model on the sorting robot;

[0078] The production instruction sheet parsing module is used for the industrial computer to obtain the input production instruction sheet, parse the production instruction sheet to obtain the SKU code and the sorting quantity, start the corresponding induction counting table based on each SKU code, bind each SKU code, the sorting quantity and the counting table number of the induction counting table to generate a binding relationship, and send the binding relationship to the sorting robot;

[0079] The slipper sorting module is used for the sorting robot to collect real-time slipper images of the slippers to be sorted through a camera, input the real-time slipper images into the slipper recognition model to obtain the SKU code, and sort the slippers to the corresponding induction counting table based on the recognized SKU code and the binding relationship; the blanking port of the induction counting table is made of stainless steel, which has the advantages of wear resistance and fall resistance, and each of the induction counting tables can be flexibly combined and spliced;

[0080] The slipper counting module is used for the sensing counting platform to count the sorted slippers through the photoelectric sensor, display the counting result through the display, and verify the counting result through the visual counting camera;

[0081] The slipper label printing module is used for the industrial computer to control the label printer to print slipper labels and paste them on the classification box after judging that the classification box in the induction counting table is full based on the counting result; after the classification box is full, the full frame indicator light provided on the induction counting table lights up to remind the full frame;

[0082] The classification frame circulation module is used for the industrial computer to circulate the full classification frames through the AGV car, and replace the empty classification frames to continue sorting until the sorting quantity is met.

[0083] The data set construction module is specifically used for:

[0084] A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

[0085] By collecting a large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles, each historical slipper image is preprocessed with sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each historical slipper image after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set, which effectively improves the diversity of the data set. The data set is then used to train a slipper recognition model, which effectively improves the recognition accuracy and generalization ability of the slipper recognition model.

[0086] The slipper recognition model training module is specifically used for:

[0087] The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then:

[0088] The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

[0089] By dividing the data set into a training set, a test set and a validation set, the slipper recognition model is trained with the training set until the loss value of the loss function is less than the preset loss threshold, the trained slipper recognition model is tested with the test set until the recognition accuracy is greater than the preset accuracy threshold, the tested slipper recognition model is verified with the validation set until the confidence is greater than the preset confidence threshold, and the verified slipper recognition model is then deployed on the sorting robot. That is, during the training process of the slipper recognition model, testing and verification are continuously performed until the requirements are met, thereby greatly improving the recognition accuracy of the slipper recognition model.

[0090] The slippers sorting module is specifically used for:

[0091] The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

[0092] By collecting real-time slipper images of the slippers to be sorted, the real-time slipper images are pre-processed with at least image enhancement including noise reduction, contrast enhancement, and edge enhancement, and then input into a slipper recognition model to recognize the slippers, that is, recognition is performed based on high-quality real-time slipper images, thereby effectively improving the recognition accuracy of the slipper recognition model.

[0093] The slippers counting module is specifically used for:

[0094] The induction counting table counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; if the counting results are incorrect, an alarm of counting error is prompted through the display.

[0095] The sorted slippers are counted through several photoelectric sensors, and the perceptual signals within the preset time threshold are combined and counted to avoid the same pair of shoes being counted repeatedly and to avoid individual photoelectric sensors being blocked and unable to count. The blanking port is photographed and identified by a visual counting camera to verify the counting results. That is, the photoelectric sensor and the visual counting camera are combined for counting and verification, thereby greatly improving the accuracy of slipper counting.

[0096] In summary, the advantages of the present invention are:

[0097] 1. Build a data set by collecting a large number of historical slipper images and preprocessing and annotating them; divide the data set into training set, test set and verification set based on the preset ratio; train the slipper recognition model created based on the neural network through the training set, test the trained slipper recognition model through the test set, verify the tested slipper recognition model through the verification set, and deploy the verified slipper recognition model to the sorting robot; then the industrial computer obtains the input production instruction sheet and parses it to obtain the SKU code and the sorting quantity, starts the corresponding induction counting table based on each SKU code, binds each SKU code, the sorting quantity and the counting table number of the induction counting table to generate a binding relationship and sends it to the sorting robot; the sorting robot collects the real-time slipper image of the slippers to be sorted through the camera and inputs it into the slipper recognition model to obtain the SKU code, based on the recognized SKU The slippers are sorted to the corresponding induction counting station based on the coding and binding relationship; then the induction counting station counts the sorted slippers through the photoelectric sensor, displays the counting result through the display, verifies the counting result through the visual counting camera, and judges that the classification box in the induction counting station is full based on the counting result, controls the label printer to print the slipper label and paste it on the classification box, circulates the full classification box through the AGV car, and replaces the empty classification box to continue sorting until the sorting quantity is met; that is, the slippers are automatically identified and sorted by the sorting robot deployed with the pre-trained slipper recognition model, the sorted slippers are automatically counted by the photoelectric sensor of the induction counting station, and the counting result is verified by the visual counting camera, that is, the automatic sorting and counting of slippers is realized to replace the traditional manual sorting, thereby greatly improving the efficiency and quality of slipper sorting and counting.

[0098] 2. By collecting a large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles, each historical slipper image is preprocessed with sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each historical slipper image after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set, which effectively improves the diversity of the data set. The data set is then used to train the slipper recognition model, which effectively improves the recognition accuracy and generalization ability of the slipper recognition model.

[0099] 3. By dividing the data set into training set, test set and validation set, the slipper recognition model is trained with the training set until the loss value of the loss function is less than the preset loss threshold, the trained slipper recognition model is tested with the test set until the recognition accuracy is greater than the preset accuracy threshold, the tested slipper recognition model is verified with the validation set until the confidence is greater than the preset confidence threshold, and then the verified slipper recognition model is deployed on the sorting robot. That is, during the training process of the slipper recognition model, continuous testing and verification are performed until the requirements are met, thereby greatly improving the recognition accuracy of the slipper recognition model.

[0100] 4. By collecting real-time slipper images of the slippers to be sorted, the real-time slipper images are pre-processed including at least noise reduction, contrast enhancement, and edge enhancement, and then input into the slipper recognition model to recognize the slippers, that is, recognition is performed based on high-quality real-time slipper images, which effectively improves the recognition accuracy of the slipper recognition model.

[0101] 5. The sorted slippers are counted through several photoelectric sensors, and the perceptual signals within the preset time threshold are combined and counted to avoid repeated counting of the same pair of shoes and to avoid individual photoelectric sensors being blocked and unable to count. The blanking port is photographed and identified by a visual counting camera to verify the counting results. That is, the photoelectric sensor and the visual counting camera are combined for counting and verification, which greatly improves the accuracy of slipper counting.

[0102] Although the specific implementation modes of the present invention are described above, those skilled in the art should understand that the specific implementation modes described are only illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A slipper sorting and counting method based on AI, characterized by: The steps include: Step S1, collecting a large number of historical slipper images, and constructing a data set after preprocessing and annotating each of the historical slipper images; Step S2: creating a slipper recognition model based on a neural network, and setting a loss function of the slipper recognition model; Step S3, dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model by using the training set, testing the trained slipper recognition model by using the test set, verifying the tested slipper recognition model by using the verification set, and deploying the verified slipper recognition model on the sorting robot; Step S4: The industrial computer obtains the input production instruction sheet, parses the production instruction sheet to obtain the SKU code and the sorting quantity, starts the corresponding induction counting station based on each SKU code, binds each SKU code, the sorting quantity and the counting station number of the induction counting station to generate a binding relationship, and sends the binding relationship to the sorting robot; Step S5: The sorting robot collects real-time images of the slippers to be sorted through a camera, inputs the real-time images into a slipper recognition model to obtain a SKU code, and sorts the slippers to corresponding induction counting stations based on the recognized SKU code and the binding relationship; Step S6, the induction counting station counts the sorted slippers through the photoelectric sensor, displays the counting result through the display, and verifies the counting result through the visual counting camera; Step S7: After the industrial computer determines that the classification frame in the induction counting table is full based on the counting result, it controls the label printer to print slipper labels and stick them on the classification frame; Step S8: The industrial computer circulates the full classification frames through the AGV trolley, and replaces the empty classification frames to continue sorting until the sorting quantity is met.

2. The AI-based slipper sorting and counting method according to claim 1, characterized in that: The step S1 is specifically as follows: A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

3. The AI-based slipper sorting and counting method according to claim 1, characterized in that: The step S3 is specifically as follows: The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:

1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then: The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

4. The AI-based slipper sorting and counting method according to claim 1, characterized in that: The step S5 is specifically as follows: The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

5. The AI-based slipper sorting and counting method according to claim 1, characterized in that: The step S6 is specifically as follows: The induction counting station counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; If the counting result is wrong, an error alarm will be prompted on the display.

6. An AI-based slipper sorting and counting system, characterized by: Includes the following modules: A data set construction module, used to collect a large number of historical slipper images, and construct a data set after preprocessing and annotating each of the historical slipper images; A slipper recognition model creation module, used to create a slipper recognition model based on a neural network and set a loss function of the slipper recognition model; a slipper recognition model training module, for dividing the data set into a training set, a test set and a verification set based on a preset ratio, training the slipper recognition model through the training set, testing the trained slipper recognition model through the test set, verifying the tested slipper recognition model through the verification set, and deploying the verified slipper recognition model on the sorting robot; The production instruction sheet parsing module is used for the industrial computer to obtain the input production instruction sheet, parse the production instruction sheet to obtain the SKU code and the sorting quantity, start the corresponding induction counting table based on each SKU code, bind each SKU code, the sorting quantity and the counting table number of the induction counting table to generate a binding relationship, and send the binding relationship to the sorting robot; A slipper sorting module is used for the sorting robot to collect real-time slipper images of slippers to be sorted through a camera, input the real-time slipper images into a slipper recognition model to obtain a SKU code, and sort the slippers to a corresponding induction counting station based on the recognized SKU code and the binding relationship; The slipper counting module is used for the sensing counting platform to count the sorted slippers through the photoelectric sensor, display the counting result through the display, and verify the counting result through the visual counting camera; A slipper label printing module, which is used for the industrial computer to control the label printer to print slipper labels and stick them on the classification frame after judging that the classification frame in the induction counting table is full based on the counting result; The classification frame circulation module is used for the industrial computer to circulate the full classification frames through the AGV trolley, and replace the empty classification frames to continue sorting until the sorting quantity is met.

7. The AI-based slipper sorting and counting system according to claim 6, characterized in that: The data set construction module is specifically used for: A large number of historical slipper images of different styles, types, materials, colors, sizes and shooting angles are collected, and each of the historical slipper images is preprocessed by sample expansion including at least rotation, scaling, flipping, distortion and occlusion, and each of the historical slipper images after sample expansion is labeled with at least style, type, material, color, size and SKU code to construct a data set.

8. The AI-based slipper sorting and counting system according to claim 6, characterized in that: The slipper recognition model training module is specifically used for: The data set is divided into a training set, a test set and a validation set based on a ratio of 7:2:

1. The slipper recognition model is trained by the training set until the loss value of the loss function is less than a preset loss threshold. The trained slipper recognition model is tested by the test set to determine whether the recognition accuracy is greater than the preset accuracy threshold. If not, the training set is expanded to continue training. If yes, then: The tested slipper recognition model is verified by the verification set to determine whether the confidence is greater than a preset confidence threshold. If not, the training set is expanded to continue training; if so, the verified slipper recognition model is deployed on the sorting robot.

9. The AI-based slipper sorting and counting system according to claim 6, characterized in that: The slippers sorting module is specifically used for: The sorting robot collects real-time slipper images of the slippers to be sorted through a camera, and after preprocessing the real-time slipper images including at least noise reduction, contrast enhancement, and edge enhancement, the real-time slipper images are input into a slipper recognition model to identify the slippers and obtain matching SKU codes. Based on the identified SKU codes and binding relationships, the corresponding slippers are grabbed and sorted to the corresponding induction counting table.

10. The AI-based slipper sorting and counting system according to claim 6, characterized in that: The slippers counting module is specifically used for: The induction counting station counts the sorted slippers through several photoelectric sensors installed at the blanking port, combines and counts the inductive signals within the preset time threshold, displays the counting results on the display, and uses the visual counting camera to photograph and identify the blanking port to verify the counting results. If the counting results are correct, the counting results are synchronized to the ERP order system; If the counting result is wrong, an error alarm will be prompted on the display.

Images