Ceramic tube sorting method and device based on image recognition

Through the combination of image recognition technology and dual-spectral information, the problems of slow sorting speed and low accuracy of traditional ceramic tubes are solved, and high-speed, accurate sorting and automated production of ceramic tubes are achieved.

CN118904759AActive Publication Date: 2024-11-08SHENZHEN POLYTECHNIC
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411182552.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-08
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

The sorting of traditional ceramic tubes relies on manual visual inspection, which leads to slow sorting speed, low accuracy and inconsistent sorting, making it difficult to meet the needs of large-scale production.

Method used

Using an image recognition method, by placing the ceramic tube on the sorting turntable, the front end, back end and length images of the ceramic tube are obtained, combined with visible light and near-infrared spectral information, the pre-trained image recognition model is used for identification, and according to Identify the results and control the air valve for classification collection.

Benefits of technology

It realizes high-speed and accurate sorting of ceramic tubes, improves sorting efficiency and consistency, reduces manual intervention and resource waste, and is suitable for large-scale production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118904759A_ABST
    Figure CN118904759A_ABST
Patent Text Reader

Abstract

The invention provides a ceramic tube sorting method and device based on image recognition, and relates to the technical field of computers. The method comprises the steps that ceramic tubes to be sorted are placed on a sorting rotary table and stably placed at a fixed angle, a ceramic tube image set of the ceramic tubes is obtained along with rotation of the sorting rotary table, and the ceramic tube image set comprises a front-end image, a rear-end image and a length image; a pre-trained image recognition model is called, the ceramic tube image set is input into the image recognition model for recognition, a corresponding recognition result is obtained, and the recognition result comprises good products and defective products; and when the sorting rotary table conveys the ceramic tubes to a discharging area, different air valves are controlled to be opened based on the recognition result, and the ceramic tubes are conveyed to a preset area to be classified and collected. The problem that in the prior art, the accuracy of ceramic tube sorting is not high is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular, to a method and device for sorting ceramic tubes based on image recognition. Background Art

[0002] Electronic ceramic tubes usually refer to ceramic components used in electronic devices. These components can be insulators or ceramic parts with specific electrical properties. Electronic ceramic tubes are widely used in the electronics industry, such as insulators, high-frequency circuit components, sensor components, etc. In the process of ceramic tube sorting, it is necessary to ensure that these components have good electrical and mechanical properties.

[0003] Traditional ceramic tube sorting mainly relies on manual visual inspection, which has many limitations, including but not limited to: since manual inspection is required one by one, the sorting speed is slow and it is difficult to meet the needs of large-scale production. In addition, different inspectors have different experiences and technical levels, resulting in different sorting standards and poor consistency.

[0004] From the above, it can be seen that how to improve the efficiency of ceramic tube sorting based on image recognition remains to be solved. Summary of the invention

[0005] The present application provides a ceramic tube sorting method, device, electronic device and storage medium based on image recognition, which can solve the problem of low accuracy of ceramic tube sorting in related technologies. The technical solution is as follows:

[0006] According to one aspect of the present application, a ceramic tube sorting method based on image recognition includes: placing the ceramic tubes to be sorted on a sorting turntable and stably placing them at a fixed angle, and obtaining a ceramic tube image set of the ceramic tubes as the sorting turntable rotates, wherein the ceramic tube image set includes a front end image, a rear end image and a length image; calling a pre-trained image recognition model, inputting the ceramic tube image set into the image recognition model for recognition, and obtaining corresponding recognition results, wherein the recognition results include good products and defective products; when the sorting turntable is transported to the unloading area, different air valves are controlled to open based on the recognition results, and the ceramic tubes are sent to a predetermined area for classification and collection.

[0007] In an exemplary embodiment,

[0008] In an exemplary embodiment, in the process of obtaining a set of ceramic tube images of ceramic tubes, the method also includes: controlling the emission of dual light of visible light and near-infrared spectrum to the ceramic tubes to be sorted; obtaining dual-spectrum images of the ceramic tubes; wherein the image recognition model includes a Transformer-based deep learning hierarchical detection algorithm.

[0009] In an exemplary embodiment, in the process of inputting the ceramic tube image set into the image recognition model for identification and obtaining the corresponding identification results, the method also includes: inputting all the front-end images, rear-end images and length images corresponding to the dual-spectrum images into the image recognition model for identification; or, inputting the front-end images, rear-end images and length images corresponding to the dual-spectrum images into the image recognition model one by one for identification, and obtaining the front-end identification results corresponding to the front-end images and the rear-end identification results corresponding to the rear-end images; if there are defective products in the front-end identification results or the rear-end identification results, and defective ceramic tubes are determined, then the subsequent identification of the defective ceramic tubes is stopped, and the defective ceramic tubes are sent to a predetermined area for classification and collection.

[0010] In an exemplary embodiment, the method further includes: after identifying a defective ceramic tube, determining a corresponding defective position based on the defective ceramic tube, wherein the defective position includes the front end of the ceramic tube, the rear end of the ceramic tube, and the length of the ceramic tube; obtaining a defective frequency corresponding to each of the defective positions, and sorting the three defective frequencies; determining a vulnerable position of the ceramic tube based on the largest defective frequency, and switching the identification order of the vulnerable position of the ceramic tube to priority identification.

[0011] In an exemplary embodiment, the dual-spectrum image, visible light image and near-infrared light image, includes a method further comprising: simultaneously or alternately emitting visible light and emitting near-infrared light to illuminate the ceramic tubes to be sorted, to obtain corresponding visible light images and near-infrared light images; extracting corresponding visible light features based on the visible light image, extracting corresponding near-infrared light features based on the near-infrared light image, and fusing the visible light features with the near-infrared light features to obtain corresponding fused features; and inputting the fused features into the image recognition model for recognition.

[0012] In an exemplary embodiment, in the process of fusing the visible light feature with the near infrared light feature to obtain a corresponding fused feature, the method further includes: acquiring a visible light image I vis and near infrared image I NIR , for the visible light image I vis and visible light image I NIR Perform preprocessing and obtain the processed I′ vis and I′ NIR ; to I′ vis and I′ NIR Feature extraction is performed on both to obtain the corresponding visible light feature F vis With near infrared light characteristics F NIR ; Retrieve weight data, where the weight of visible light is α, the weight of near-infrared light is β, and α+β=1, and calculate the fusion feature F fused , and F fused=α×F vis +β×F NIR .

[0013] In an exemplary embodiment, the method further includes: controlling the front sensor to obtain the reflection intensity of the surface of the ceramic tube; determining whether the reflection intensity matches a preset light intensity range, and if not, adjusting the light intensity.

[0014] According to one aspect of the present application, a ceramic tube sorting device based on image recognition includes: a ceramic tube image set acquisition module, which places the ceramic tube to be sorted on a sorting turntable and stably places it at a fixed angle, and as the sorting turntable rotates, it is used to acquire a ceramic tube image set of the ceramic tube, wherein the ceramic tube image set includes a front end image, a rear end image and a length image; a recognition result acquisition module, which calls a pre-trained image recognition model and inputs the ceramic tube image set into the image recognition model for recognition, so as to obtain a corresponding recognition result, wherein the recognition result includes good products and defective products; a classification and collection module, which controls different gas valves to open based on the recognition result when the sorting turntable is transported to the unloading area, so as to transport the ceramic tubes to a predetermined area for classification and collection.

[0015] According to one aspect of the present application, an electronic device includes at least one processor and at least one memory, wherein the memory stores computer-readable instructions; the computer-readable instructions are executed by one or more of the processors, so that the electronic device implements the ceramic tube sorting method based on image recognition as described above.

[0016] According to one aspect of the present application, a storage medium stores computer-readable instructions thereon, and the computer-readable instructions are executed by one or more processors to implement the ceramic tube sorting method based on image recognition as described above.

[0017] According to one aspect of the present application, a computer program product includes computer-readable instructions, which are stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, so that the electronic device implements the ceramic tube sorting method based on image recognition as described above.

[0018] The beneficial effects of the technical solution provided by this application are:

[0019] During the rotation of the sorting turntable, images of the ceramic tube are obtained from different angles (front end, rear end and length direction), so that the appearance quality of the ceramic tube can be comprehensively checked. Then the image recognition model can quickly process a large number of images in a short time, thereby realizing a high-speed sorting process. When the sorting turntable moves to the unloading area, the system controls the opening of the corresponding air valve according to the recognition result to accurately send the ceramic tube to the preset collection area. On the one hand, since the whole process is highly automated, continuous and uninterrupted production can be achieved, reducing waiting time and manual intervention time, which can avoid errors in manual operation and improve sorting accuracy and efficiency. On the other hand, by timely identifying and separating defective products, resource waste in subsequent processing steps is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0021] Figure 1 It is a schematic diagram of the implementation environment involved in this application;

[0022] Figure 2 is a flow chart of a ceramic tube sorting method based on image recognition according to an exemplary embodiment;

[0023] Figure 3 is a flow chart of S103 to S105 in a ceramic tube sorting method based on image recognition according to an exemplary embodiment;

[0024] Figure 4 is a flow chart of S106 to S108 in a ceramic tube sorting method based on image recognition according to an exemplary embodiment;

[0025] Figure 5 This is a structural block diagram of a ceramic tube sorting device based on image recognition according to an exemplary embodiment. DETAILED DESCRIPTION

[0026] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0027] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present disclosure refers to the presence of the features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0028] The ceramic tube sorting method based on image recognition provided in the present application can effectively improve the accuracy of ceramic tube sorting. Accordingly, the ceramic tube sorting method based on image recognition is suitable for a ceramic tube sorting device based on image recognition, and the ceramic tube sorting device based on image recognition can be deployed in electronic equipment.

[0029] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0030] Figure 1 The schematic diagram of the implementation environment involved in a ceramic tube sorting method based on image recognition is shown in FIG. The implementation environment includes a collection end and a service end.

[0031] See also Figure 2 The present application embodiment provides a ceramic tube sorting method based on image recognition, which is applicable to electronic equipment, which may be Figure 1 Shown are servers in an implementation environment.

[0032] In the following method embodiments, for ease of description, the execution subject of each step of the method is taken as an electronic device as an example for illustration, but this does not constitute a specific limitation.

[0033] like Figure 2 As shown, the method may include the following steps:

[0034] S100, placing the ceramic tubes to be sorted on a sorting turntable and stably placing them at a fixed angle, and acquiring a set of ceramic tube images of the ceramic tubes as the sorting turntable rotates.

[0035] It should be pointed out here that the ceramic tubes to be sorted can be placed on the vibrating plate for loading first, so that the ceramic tubes can be placed on the sorting turntable at a fixed angle, and the placement position and angle of the ceramic tubes on the sorting turntable are fixed, ensuring that each image is taken under consistent conditions, which helps to improve the quality and consistency of the image.

[0036] Among them, the ceramic tube image set includes front end image, rear end image and length image; during the rotation of the sorting turntable, the images of the ceramic tube are obtained from different angles (front end, rear end and length direction), so that the appearance quality of the ceramic tube can be comprehensively checked.

[0037] S110, calling a pre-trained image recognition model, inputting the ceramic tube image set into the image recognition model for recognition, and obtaining a corresponding recognition result.

[0038] Among them, the recognition results include good products and defective products. In addition, the image recognition model used here is trained in advance. The specific training process is as follows: collect a large number of ceramic tube images as training sets. These images should include various types of good products and defective products. For each image, its category (good product or defective product) needs to be marked, and the specific defect location or type may also need to be marked. Then remove duplicate, unclear or non-compliant images to ensure the quality of the data set. It is necessary to adjust the image size, unify the color space, etc. to make the image consistent. The diversity and richness of the data set can also be increased through transformations such as rotation, scaling, and flipping to help the model generalize better; select appropriate deep learning according to task requirements Frameworks, such as convolutional neural networks, divide the data set into training set, validation set and test set, where the training set is used to train the model, the validation set is used to adjust the hyperparameters, and the test set is used to evaluate the final model performance. The data in the training set is used for iterative training, and the weights and biases are adjusted through the back-propagation algorithm to minimize the loss function. During this period, the validation set is used to monitor the performance of the training process. If problems such as overfitting are found, the model structure or regularization parameters are adjusted. Necessary adjustments are made based on the evaluation results, such as adjusting the learning rate, adding more training data, or modifying the model architecture. Finally, the trained model is saved as an image recognition model for easy deployment, and the model is deployed to the actual sorting system for real-time image recognition and sorting tasks.

[0039] S120, when the sorting turntable is transported to the unloading area, different air valves are controlled to open based on the recognition results, and the ceramic tubes are sent to the predetermined area for classification and collection.

[0040] When the sorting turntable moves to the unloading area, the system controls the corresponding air valve to open according to the recognition result, and accurately delivers the ceramic tube to the preset collection area. This can avoid errors in manual operation and improve the accuracy of sorting.

[0041] It should be pointed out here that when the sorting turntable moves to the unloading area, after passing the unloading area corresponding to good products and the unloading area corresponding to defective products, if the ceramic tubes to be sorted on the sorting turntable are not blown off, then another blowing valve on the system control device is opened to collect the residual ceramic tubes on the sorting turntable to avoid repeated inspections and avoid affecting the subsequent ceramic tube sorting on the sorting turntable.

[0042] In order to further improve the quality of the ceramic tubes to be sorted, in the process of acquiring a set of ceramic tube images of the ceramic tubes, the method further includes:

[0043] S101, controlling the emission of dual light of visible light and near infrared spectrum to illuminate the ceramic tubes to be sorted.

[0044] It should be pointed out here that visible light can provide conventional visual information on the surface of ceramic tubes, making it easy to capture surface features such as color and shape; while near-infrared spectroscopy can reveal some characteristics inside the material, such as internal structure, composition differences, etc., which is particularly useful for detecting some subtle defects under the surface.

[0045] S102, acquiring a dual-spectrum image of the ceramic tube.

[0046] Among them, combining the image information of visible light and near-infrared spectra can obtain a more comprehensive characterization of ceramic tubes, which helps to improve the accuracy and reliability of detection; near-infrared spectral images can detect some internal or subsurface defects that cannot be found by visible light alone, such as microcracks, pores, etc., so dual-spectrum images can reduce the impact of ambient light changes on image quality and improve image consistency and stability.

[0047] Among them, in an embodiment of the present application, the image recognition model includes a Transformer-based deep learning hierarchical detection algorithm. Transformer is a powerful deep learning architecture that is particularly suitable for processing sequence data. In image recognition tasks, it can be used to process local and global context information in image blocks or feature maps, and the Transformer-based model can effectively capture the details and patterns in the image, especially for complex defect detection tasks, and can provide more accurate recognition results.

[0048] By adopting dual illumination and Transformer-based deep learning hierarchical detection algorithms, not only can more information about the surface and interior of the ceramic tube be obtained, but good and defective products can also be identified and classified more accurately.

[0049] After acquiring the dual-spectrum image, as Figure 3 As shown, the method also includes:

[0050] S103, inputting the front-end image, the rear-end image and the length image corresponding to the dual-spectrum image into the image recognition model for recognition.

[0051] Among them, all images (front-end images, back-end images and length images) are simultaneously input into the image recognition model for recognition. The image recognition model can comprehensively consider information from multiple perspectives to make more accurate judgments. It can also reduce the number of model calls and improve the overall processing speed.

[0052] S104, or, inputting the front-end image, the rear-end image and the length image corresponding to the dual-spectrum image into the image recognition model one by one for recognition, and obtaining the front-end recognition result corresponding to the front-end image and the rear-end recognition result corresponding to the rear-end image;

[0053] S105, if there are defective products in the front-end recognition result or the back-end recognition result, and the defective ceramic tubes are determined, the subsequent recognition of the defective ceramic tubes is stopped, and the defective ceramic tubes are sent to a predetermined area for classification and collection.

[0054] The front-end image, back-end image and length image are input into the image recognition model for recognition one by one. If a defective product is found in the front-end or back-end recognition results, the subsequent recognition steps of the ceramic tube are immediately stopped and it is directly classified as a defective product. Once a ceramic tube is determined to be a defective product, it can be processed immediately without additional recognition steps, thus saving computing resources and time.

[0055] By using the information of dual-spectrum images and adjusting the corresponding recognition strategy, the efficiency and accuracy of the ceramic tube sorting process are improved, while unnecessary consumption of computing resources is reduced.

[0056] In addition, it should be pointed out that in the recognition process, the front-end image, the back-end image and the length image are recognized respectively, and in the above-mentioned strategy of recognizing one by one, in order to further improve the recognition efficiency, such as Figure 4 As shown, the following steps are also performed:

[0057] S106, after identifying a defective ceramic tube, determining a corresponding defective position based on the defective ceramic tube.

[0058] Among them, the defective positions include the front end of the ceramic tube, the rear end of the ceramic tube and the length of the ceramic tube; by determining the defective positions, it is possible to more accurately understand which parts are prone to problems.

[0059] S107, obtaining the defect frequency corresponding to each defect position, and sorting the three defect frequencies.

[0060] S108, determining the vulnerable position of the ceramic tube based on the maximum defect frequency, and switching the identification order of the vulnerable position of the ceramic tube to priority identification.

[0061] Among them, the defective frequencies of the three defective positions are sorted, and according to the sorting results, the position with the highest defective frequency, that is, the vulnerable position, is found, and the identification order of the vulnerable position is adjusted to the priority identification position. By giving priority to identifying the vulnerable position, defective products can be found more quickly, thereby reducing subsequent unnecessary identification steps and improving sorting efficiency.

[0062] In addition, in the embodiment of the present application, the process of identifying the dual-spectrum image requires simultaneously or alternately emitting visible light and emitting near-infrared light to illuminate the ceramic tube to be sorted, and then obtaining the corresponding visible light image I vis and near infrared image I NIR , for the visible light image I vis and visible light image I NIR Perform preprocessing and obtain the processed I′ vis and I′ NIR , then for I′ vis and I′ NIR Feature extraction is performed on both to obtain the corresponding visible light feature F vis With near infrared light characteristics F NIR .

[0063] It should be pointed out here that for the visible light feature F vis With near infrared light characteristics F NIR , the corresponding weight data is pre-set in the system, and the weight data can be modified, and then the weight data is retrieved, where the weight of visible light is α, the weight of near-infrared light is β, and α+β=1, and the fusion feature F is calculated fused , and F fused =α×F vis +β×F NIR , and finally the fusion feature F fused By inputting the image recognition model for recognition, the corresponding recognition results can be obtained.

[0064] By fusing the features of visible light and near-infrared light, the internal and external characteristics of ceramic tubes can be comprehensively analyzed from multiple angles to improve recognition accuracy. The adjustability of weight data enables the image recognition model to be personalized according to different types of ceramic tubes or specific needs, thereby improving the applicability of the image recognition model. The use of fused features reduces the amount of calculation required to process each spectral image individually, thereby speeding up recognition.

[0065] Finally, in the process of irradiating the ceramic tubes to be sorted, the light intensity can be adjusted, which specifically includes the following steps: controlling the front sensor to obtain the reflection intensity of the ceramic tube surface, and then determining whether the reflection intensity matches the pre-set light intensity range. If not, adjusting the light intensity.

[0066] By controlling the front sensor to obtain the reflective intensity of the ceramic tube surface and adjusting the light intensity according to whether the reflective intensity matches the pre-set light intensity range, the quality of image acquisition can be ensured, the accuracy and reliability of image recognition can be improved, and the performance and efficiency of the entire sorting system can be improved. This method is particularly suitable for automated production lines and can significantly improve the accuracy and speed of ceramic tube sorting.

[0067] The following is an embodiment of the device of the present application, which can be used to execute the ceramic tube sorting method based on image recognition involved in the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the method embodiment of the ceramic tube sorting method based on image recognition involved in the present application.

[0068] See also Figure 5 In the embodiment of the present application, a ceramic tube sorting device based on image recognition is provided, including but not limited to:

[0069] The ceramic tube image set acquisition module 200 places the ceramic tube to be sorted on the sorting turntable and stably places it at a fixed angle. As the sorting turntable rotates, the ceramic tube image set of the ceramic tube is acquired, wherein the ceramic tube image set includes a front end image, a rear end image and a length image;

[0070] The recognition result acquisition module 210 retrieves a pre-trained image recognition model, inputs a set of ceramic tube images into the image recognition model for recognition, and obtains corresponding recognition results, wherein the recognition results include good products and defective products;

[0071] The classification and collection module 220 controls different air valves to open based on the recognition results when the sorting turntable is transported to the unloading area, so as to transport the ceramic tubes to the predetermined area for classification and collection.

[0072] In an exemplary embodiment, the device further includes but is not limited to:

[0073] A light control module, used to control the dual light of visible light and near infrared spectrum to be irradiated to the ceramic tubes to be sorted;

[0074] A dual-spectrum image acquisition module, used for acquiring a dual-spectrum image of the ceramic tube;

[0075] Among them, the image recognition model includes a Transformer-based deep learning hierarchical detection algorithm.

[0076] In an exemplary embodiment, the device further includes but is not limited to:

[0077] An image recognition module is used to input the front-end image, the back-end image and the length image corresponding to the dual-spectrum image into the image recognition model for recognition;

[0078] A recognition module is used to input the front-end image, the rear-end image and the length image corresponding to the dual-spectrum image into the image recognition model one by one for recognition, and obtain the front-end recognition result corresponding to the front-end image and the rear-end recognition result corresponding to the rear-end image;

[0079] If there are defective products in the front-end recognition results or the back-end recognition results, and the defective ceramic tubes are determined to be defective, the subsequent identification of the defective ceramic tubes will be stopped, and the defective ceramic tubes will be sent to the predetermined area for classification and collection.

[0080] In an exemplary embodiment, the device further includes but is not limited to:

[0081] A defective position determination module, after identifying a defective ceramic tube, is used to determine a corresponding defective position based on the defective ceramic tube, wherein the defective position includes the front end of the ceramic tube, the rear end of the ceramic tube and the length of the ceramic tube;

[0082] A sorting module obtains the defect frequency corresponding to each defect position and is used to sort the three defect frequencies;

[0083] The identification sequence switching module determines the vulnerable position of the ceramic tube based on the maximum defective frequency and is used to switch the identification sequence of the vulnerable position of the ceramic tube to priority identification.

[0084] In an exemplary embodiment, the device further includes but is not limited to:

[0085] An image acquisition module, which simultaneously or alternately emits visible light and near-infrared light to illuminate the ceramic tubes to be sorted, so as to acquire corresponding visible light images and near-infrared light images;

[0086] A fusion module, which extracts corresponding visible light features based on the visible light image, and extracts corresponding near-infrared light features based on the near-infrared light image, and is used to fuse the visible light features with the near-infrared light features to obtain corresponding fusion features;

[0087] The feature recognition module is used to input the fused features into the image recognition model for recognition.

[0088] In an exemplary embodiment, the device further includes but is not limited to:

[0089] Obtain visible light image I vis and near infrared image I NIR , for the visible light image I vis and visible light image I NIR Perform preprocessing and obtain the processed I′ vis and I′ NIR ;

[0090] to I' vis and I′ NIRFeature extraction is performed on both to obtain the corresponding visible light feature F vis With near infrared light characteristics F NIR ;

[0091] Retrieve weight data, where the weight of visible light is α, the weight of near-infrared light is β, and α+β=1, and calculate the fusion feature F fused , and F fused =α×F vis +β×F NIR .

[0092] In an exemplary embodiment, the device further includes but is not limited to:

[0093] A reflection intensity acquisition module controls the front sensor to obtain the reflection intensity of the ceramic tube surface;

[0094] The judgment module is used to judge whether the reflected light intensity matches the preset light intensity range. If not, the light intensity is adjusted.

[0095] It should be noted that the ceramic tube sorting device based on image recognition provided in the above embodiment only uses the division of the above-mentioned functional modules as an example when performing ceramic tube sorting based on image recognition. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the ceramic tube sorting device based on image recognition will be divided into different functional modules to complete all or part of the functions described above.

[0096] In addition, the ceramic tube sorting device based on image recognition provided in the above embodiments and the ceramic tube sorting method based on image recognition belong to the same concept, and the specific way in which each module performs the operation has been described in detail in the method embodiment and will not be repeated here.

[0097] According to one aspect of the present application, an electronic device includes at least one processor and at least one memory, wherein the memory stores computer-readable instructions; the computer-readable instructions are executed by one or more processors, so that the electronic device implements the above-mentioned ceramic tube sorting method based on image recognition.

[0098] In addition, a storage medium is provided in an embodiment of the present application, on which computer-readable instructions are stored, and the computer-readable instructions are executed by one or more processors to implement the above-mentioned ceramic tube sorting method based on image recognition.

[0099] A computer program product is provided in an embodiment of the present application. The computer program product includes computer-readable instructions, which are stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, so that the electronic device implements the above-mentioned ceramic tube sorting method based on image recognition.

[0100] In summary, during the rotation of the sorting turntable, images of the ceramic tube are obtained from different angles (front end, rear end and length direction), so that the appearance quality of the ceramic tube can be comprehensively checked. Then the image recognition model can quickly process a large number of images in a short time, thereby realizing a high-speed sorting process. When the sorting turntable moves to the unloading area, the system controls the opening of the corresponding air valve according to the recognition result to accurately send the ceramic tube to the preset collection area. On the one hand, since the whole process is highly automated, continuous and uninterrupted production can be achieved, reducing waiting time and manual intervention time, which can avoid errors in manual operation and improve sorting accuracy and efficiency. On the other hand, by timely identifying and separating defective products, resource waste in subsequent processing steps is reduced.

[0101] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0102] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A ceramic tube sorting method based on image recognition, characterized in that: include: The ceramic tube to be sorted is placed on a sorting turntable and stably placed at a fixed angle. As the sorting turntable rotates, a ceramic tube image set of the ceramic tube is obtained, wherein the ceramic tube image set includes a front end image, a rear end image and a length image; Retrieving a pre-trained image recognition model, inputting the ceramic tube image set into the image recognition model for recognition, and obtaining corresponding recognition results, wherein the recognition results include good products and defective products; When the sorting turntable is transported to the unloading area, different air valves are controlled to open based on the recognition result, and the ceramic tubes are transported to the predetermined area for classification and collection.

2. The method according to claim 1, characterized in that In the process of acquiring the ceramic tube image set of the ceramic tube, the method further includes: Controlling the dual light of visible light and near infrared spectrum to irradiate the ceramic tube to be sorted; Obtain dual-spectral images of ceramic tubes; Among them, the image recognition model includes a Transformer-based deep learning hierarchical detection algorithm.

3. The method according to claim 2, characterized in that In the process of inputting the ceramic tube image set into the image recognition model for recognition and obtaining the corresponding recognition result, the method further includes: Inputting the front-end image, the rear-end image and the length image corresponding to the dual-spectrum image into the image recognition model for recognition; Or, the front-end image, the rear-end image and the length image corresponding to the dual-spectrum image are input into the image recognition model one by one for recognition, and the front-end recognition result corresponding to the front-end image and the rear-end recognition result corresponding to the rear-end image are obtained; If there are defective products in the front-end recognition results or the back-end recognition results, and the defective ceramic tubes are determined to be defective, the subsequent identification of the defective ceramic tubes will be stopped, and the defective ceramic tubes will be sent to the predetermined area for classification and collection.

4. The method according to claim 3, characterized in that The method further comprises: After identifying a defective ceramic tube, determining a corresponding defective position based on the defective ceramic tube, wherein the defective position includes a front end of the ceramic tube, a rear end of the ceramic tube, and a length of the ceramic tube; Obtain the defect frequency corresponding to each defect position, and sort the three defect frequencies; The vulnerable position of the ceramic tube is determined based on the maximum defect frequency, and the identification order of the vulnerable position of the ceramic tube is switched to priority identification.

5. The method according to claim 1, characterized in that The dual-spectrum image, a visible light image and a near-infrared light image, includes a method that further includes: Simultaneously or alternately emitting visible light and near-infrared light to illuminate the ceramic tubes to be sorted, and obtaining corresponding visible light images and near-infrared light images; Extracting corresponding visible light features based on the visible light image, extracting corresponding near-infrared light features based on the near-infrared light image, and fusing the visible light features with the near-infrared light features to obtain corresponding fused features; The fused features are input into the image recognition model for recognition.

6. The method according to claim 5, characterized in that In the process of fusing the visible light feature with the near infrared light feature to obtain a corresponding fused feature, the method further includes: Obtain visible light image I vis and near infrared image I NIR , for the visible light image I vis and visible light image I NIR Perform preprocessing and obtain the processed I′ vis and I′ NIR ; to I' vis and I′ NIR The features are extracted to obtain the corresponding visible light features F vis With near infrared light characteristics F NIR ; Retrieve weight data, where the weight of visible light is α, the weight of near-infrared light is β, and α+β=1, and calculate the fusion feature F fused , and F fused =α×F vis +β×F NIR .

7. The method according to claim 1, characterized in that The method further comprises: Control the front sensor to obtain the reflection intensity of the ceramic tube surface; Determine whether the reflected light intensity matches the preset light intensity range. If not, adjust the light intensity.

8. A ceramic tube sorting device based on image recognition, characterized in that: include: The ceramic tube image set acquisition module places the ceramic tube to be sorted on the sorting turntable and stably places it at a fixed angle. As the sorting turntable rotates, the ceramic tube image set of the ceramic tube is acquired, wherein the ceramic tube image set includes a front end image, a rear end image and a length image; A recognition result acquisition module, which calls a pre-trained image recognition model, inputs the ceramic tube image set into the image recognition model for recognition, and is used to obtain corresponding recognition results, wherein the recognition results include good products and defective products; The classification and collection module controls different air valves to open based on the recognition result when the sorting turntable is transported to the unloading area, so as to transport the ceramic tubes to the predetermined area for classification and collection.

9. An electronic device, characterized in that: include: at least one processor and at least one memory, wherein: The memory has computer-readable instructions stored thereon; The computer-readable instructions are executed by one or more of the processors, so that the electronic device implements the ceramic tube sorting method based on image recognition as described in any one of claims 1 to 7.

10. A storage medium having computer-readable instructions stored thereon, characterized in that: The computer-readable instructions are executed by one or more processors to implement the ceramic tube sorting method based on image recognition as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Multispectral mineral sorting method and device, computer equipment and storage medium

    CN115797758A

  • Bursting bead sorting equipment

    CN115921350A

  • Ceramic bushing detection system, method, device and equipment and storage medium

    CN117589793A

  • Glass sorting method and device based on combination of image color recognition and deep learning

    CN118279652A

  • Method and device of getting information

    US20190205699A1