Sample data processing method, apparatus, electronic device, storage medium, and program

The method automates defect detection in spinning boxes by comparing working images with normal images to generate a defect detection model, improving efficiency and accuracy in identifying abnormalities in spinning boxes.

JP7588268B1Active Publication Date: 2024-11-21ZHEJIANG HENGYI PETROCHEMICAL CO LTD +1
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
JP2024147260
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Priority Date
2023-09-11
Filing Date
2024-08-29
Publication Date
2024-11-21
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

Conventional manual patrol inspection of spinning boxes in the chemical fiber industry is inefficient and costly, and existing AI-based image detection methods face challenges in collecting sample images due to the low frequency of abnormalities and minute thread-like characteristics of yarn defects.

Method used

A method and apparatus for processing sample data to automatically identify defects in spinning boxes by comparing working images with normal images, generating a defect detection model using a first sample data set, and employing a defect detection model to identify defects in subsequent images.

Benefits of technology

Improves the efficiency and accuracy of sample data collection and model training, enhancing the detection accuracy of spinning box abnormalities, thereby reducing manual inspection time and costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007588268000001_ABST
    Figure 0007588268000001_ABST
Patent Text Reader

Abstract

The present invention provides a method, apparatus, electronic device, storage medium and program for processing sample data, and relates to the field of artificial intelligence technology, particularly to the fields of image processing and computer vision. The method includes: obtaining a plurality of first work images of the spinning box; for each of the plurality of first work images, if there is a difference between the first work image and a normal image of the spinning box that satisfies a first condition, determining the first work image as a defect image of the spinning box; and obtaining a first sample data set based on each defect image in the plurality of first work images. According to the solution of the present disclosure, the defect image of the spinning box can be obtained by the first work images of the spinning box, and a first sample data set for training a defect detection model can be obtained from the defect images, so that the detection accuracy of the work image of the spinning box by the defect detection model can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the fields of image processing and computer vision. [Background technology]

[0002] In the chemical fiber industry, in the process of producing chemical fiber yarn, the spinning box has a certain viscosity because the spinning raw material is in the form of a non-Newtonian fluid, and the temperature of the machine is too high when the spinning box is operated for a long time, so that abnormalities such as yarn breakage, floating yarn, yarn hooking, tilt of the oil nozzle, tilt of the yarn guide hook, and yarn not entering the yarn guide hook may occur in the spinning box, so the spinning box needs to be inspected patrol. The traditional method of patrol inspection is manual patrol inspection, which has the disadvantages of low efficiency and high cost. In the related technology, an image detection method using an AI model is proposed to automatically identify abnormalities in the spinning box. However, a large number of samples are required for model training, and sample images need to be collected artificially. Meanwhile, the occurrence frequency of abnormalities in the spinning box is low, and due to the filamentous characteristics of the yarn itself, abnormal areas such as floating yarn and yarn hooking are very fine, making them difficult to identify. Therefore, the efficiency of manually collecting sample images is very low and errors are likely to occur. Summary of the Invention [Problem to be solved by the invention]

[0003] The present disclosure provides a method, apparatus, electronic device, storage medium and program for processing sample data to solve or alleviate one or more technical problems in the prior art. [Means for solving the problem]

[0004] In a first aspect, the present disclosure provides a method for processing sample data, the method comprising: acquiring a plurality of first working images of the spinning box; For each of the plurality of first work images, when there is a difference between the first work image and the normal image of the spinning box that satisfies a first condition, determining the first work image as a defective image of the spinning box; Obtaining a first sample data set based on each defect image in the plurality of first work images, the first sample data set being used to train and obtain a defect detection model of the spinning box, and the defect detection model being used to detect whether a defect exists in a second work image of the spinning box.

[0005] In a second aspect, the present disclosure provides an apparatus for processing sample data, the apparatus comprising: an image acquisition module for acquiring a plurality of first working images of the spinning box; a defect image determination module for determining, for each of the plurality of first work images, the first work image as a defect image of the spinning box when there is a difference between the first work image and the normal image of the spinning box that satisfies a first condition; and a first sample processing module for obtaining a first sample data set based on each defect image in a plurality of first work images, the first sample data set being used to train and obtain a defect detection model of the spinning box, and the defect detection model being used to detect whether a defect exists in a second work image of the spinning box.

[0006] In a third aspect, the present disclosure provides an electronic device, the device comprising: At least one processor; a memory in communication with the at least one processor; The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, cause the implementation of any one of the methods in the embodiments of the present disclosure.

[0007] In a fourth aspect, there is provided a non-transitory computer-readable storage medium having stored thereon computer instructions for causing a computer to perform any one of the methods in the embodiments of the present disclosure.

[0008] In a fifth aspect, there is provided a program for implementing any one of the methods in the embodiments of the present disclosure when executed by a processor.

[0009] The beneficial effects of the technical solution provided by the present disclosure at least include: by comparing the first working image of the spinning box with the normal image, a defect image of the spinning box can be obtained, and a first sample data set for training a defect detection model can be obtained from the defect image, which can improve the collection efficiency and accuracy of the sample data, and accordingly improve the training efficiency and model accuracy of the defect detection model, and improve the accuracy of abnormality detection of the spinning box.

[0010] It should be understood that the contents described herein are not intended to describe key or important features of the embodiments of the present disclosure, nor are they used to limit the scope of the present disclosure, other features of the present disclosure are facilitated by the following specification. [Brief description of the drawings]

[0011] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the several accompanying drawings indicate the same or similar components or elements. The accompanying drawings are not necessarily drawn to scale. It should be understood that the drawings depict only some examples provided by the present disclosure and should not be considered as limiting the scope of the present disclosure. [Figure 1] FIG. 2 is a schematic diagram illustrating an application scenario of the sample data processing method according to an embodiment of the present disclosure. [Diagram 2] 1 is a schematic flow chart of a method for processing sample data according to an embodiment of the present disclosure. [Diagram 3]FIG. 1 is a schematic block diagram illustrating a processing device for sample data according to an embodiment of the present disclosure. [Figure 4] FIG. 13 is a schematic block diagram illustrating a processing device for sample data according to another embodiment of the present disclosure. [Diagram 5] FIG. 13 is a schematic block diagram illustrating a processing device for sample data according to another embodiment of the present disclosure. [Figure 6] FIG. 1 is a block diagram of an electronic device for performing a method for processing sample data according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] The present disclosure will now be described in more detail with reference to the accompanying drawings, in which like reference numbers represent the same or similar elements, and in which various aspects of the embodiments are illustrated, and which, unless otherwise noted, are not necessarily drawn to scale.

[0013] Furthermore, in order to better explain the present disclosure, many specific details are described in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can be similarly implemented and obtained without some details. In some embodiments, methods, means, components, circuits, etc. that are well known to those skilled in the art are not described in detail so that the gist of the present disclosure is clear.

[0014] In order to facilitate understanding of the sample data processing method according to the embodiment of the present disclosure, the following first describes an exemplary application scenario of the method. Figure 1 is a schematic diagram showing an exemplary application scenario of the sample data processing method according to the embodiment of the present disclosure. The application scenario includes a patrol inspection robot 10 arranged in a workshop, and the patrol inspection robot 10 can include an image collection device 11 (e.g., a photographing device, a camera, etc.) for collecting operation images of a spinning box 20.

[0015] The image collecting device 11 may be connected to the processing device 12, and sends a first working image of the spinning box 20 to the processing device 12. The processing device 12 is used to process the first working image of the spinning box 20 to obtain a first sample data set. The first sample data set is used to train and obtain a defect detection model of the spinning box 20. After obtaining the defect detection model, the image collecting device 11 may be connected to the defect detection model, and sends a second working image of the spinning box 20 to the defect detection model. The defect detection model is used to detect defect-related information of the second working image of the spinning box 20, and the defect-related information may include the presence or absence of defects, the detected defect types, and the reliability of each defect type. When the defect detection model detects that the second working image has defects, the mechanical arm of the patrol robot 10 can trigger the power supply of the spinning box 20 corresponding to the second working image, so as to stop the spinning box 20 to wait for maintenance. The mechanical arm of the patrol robot 10 can also trigger an alarm device on the spinning box 20 corresponding to the second work image so that the spinning box 20 sends alarm information to a user device, where the alarm information can include the location, model number, etc. of the spinning box 20.

[0016] According to the method according to the embodiment of the present disclosure, in this application scenario, an image collecting device 11 and a processing device 12 of the patrol robot 10 are installed. The patrol robot 10 can obtain defect-related information of the spinning box 20. The patrol robot 10 can interact with a user device to send the defect-related information of the spinning box 20 to the user device, or perform various management functions related to the first sample data set.

[0017] Alternatively, the processing device 12 may be installed on the traveling robot 10 or on a remote device. For example, the image collecting device 11 and the processing device 12 may be installed on the traveling robot 10 as shown in FIG. 1, or the processing device 12 may not be installed on the traveling robot 10, that is, the traveling robot 10 may only be responsible for collecting images of the spinning box 20 and transmit the images of the spinning box 20 to the processing device 12.

[0018] FIG. 2 is a schematic diagram illustrating a method for processing sample data according to an embodiment of the present disclosure. The method can be applied to a processing device for sample data that can be deployed in an electronic device. The electronic device can be, for example, a standalone device or a plurality of online terminals, a server, or other processing devices. Here, the terminal can be a user device (User Equipment, UE), such as a mobile device, a personal digital assistant (PDA), a handheld device, a computing device, an in-vehicle device, a wearable device, etc. In some possible embodiments, the method can also be performed by a processor invoking computer-readable instructions stored in a memory. As shown in FIG. 1, the method can include the following steps S210 to S230.

[0019] In step S210, a plurality of first working images of the spinning box are obtained.

[0020] In an embodiment of the present disclosure, the first working image includes an image of one or more positions in the spinning box in an actual working state. Illustratively, the first working image can include an image of any one or more positions of the spinneret, the thread, the thread guide hook, and the oil nozzle in the spinning box.

[0021] Exemplarily, the first working image may be obtained by photographing the spinning box in real time by a patrol robot or manually, or may be a historical image of the spinning box in a database.

[0022] Exemplarily, the first working image may include a normal image when the spinning box is in a normal working state, and may also include a defective image when the spinning box is in an abnormal working state.

[0023] In step S220, for each of the multiple first work images, if there is a difference between the first work image and the normal image of the spinning box that satisfies a first condition, the first work image is determined to be a defective image of the spinning box.

[0024] In the embodiment of the present disclosure, the normal image includes an image of one or more positions in the spinning box during normal operation, and the position of the spinning box appearing in the normal image is the same as or corresponds to the position of the spinning box appearing in the first operation image. For example, if the first operation image includes an image of the spinneret in the spinning box when it is working (it is necessary to determine whether it is working normally or abnormally), the normal image includes an image of the spinneret in the spinning box when it is in a normal operation state. Illustratively, the quality of the chemical fiber yarn produced from the spinning box can determine whether the image of the spinning box is a normal image, and if the quality of the chemical fiber yarn produced from the spinning box is high, the image of the spinning box is a normal image. In practical application, normal images can be collected in advance to be applied to the above-mentioned method.

[0025] Exemplarily, the pixel points of the first working image are compared with the pixel points of the normal image, and the difference between the pixel points of the first working image and the pixel points of the normal image is obtained by comparing the pixel points of the first working image and the pixel points of the normal image. Here, the pixel points may be pixel points at all positions in the image, or may be pixel points at one or more positions in the image. For example, the pixel points of the spinneret of the first working image and the spinneret of the normal image are compared to obtain the difference between the spinneret of the first working image and the spinneret of the normal image.

[0026] In an embodiment of the present disclosure, the first condition can be used to determine whether a difference between the first working image and the normal image is obvious or whether the difference is a defect in the first working image. Illustratively, the first condition can be related to the size, number, location, etc. of the difference locations. For example, the first condition can include that the number of pixels with differences between the first working image and the normal image is greater than a threshold (e.g., greater than 0, greater than 10, greater than 100, etc.), the area is greater than a threshold, etc.

[0027] Exemplarily, the first condition may be preset by a user, and the user may set a plurality of first conditions for a plurality of positions, such as a waiting area for arrangement in a spinneret, a floating yarn, a position of a yarn guide hook, a position of an oil nozzle, etc.

[0028] In one embodiment, the patrol inspection robot can move to a preset position to take pictures of multiple spinning boxes, and obtain a first working image and a normal image. If the preset position of each spinning box to which the patrol inspection robot moves is the same as the distance between each spinning box, the patrol inspection robot can take pictures of the first working image and the normal image at the same shooting angle and distance, and can compare the difference between the first working image and the normal image more clearly.

[0029] In step S230, a first sample data set is obtained based on each defect image in the plurality of first working images, where the first sample data set is used to train and obtain a defect detection model of the spinning box, and the defect detection model is used to detect whether a defect exists in the second working image of the spinning box.

[0030] Illustratively, each defect image in the first sample data set is used as an input image for a preset model and labeled as defective, and the input image and the corresponding labeling information are used to train the preset model, and when the preset model converges, the preset model can be determined as a defect detection model applicable to an actual defect detection process.

[0031] Optionally, the labeling information for the input image may include information such as defect type, defect location, etc. The specific labeling process may be performed manually or using an automated labeling tool.

[0032] In an embodiment of the present disclosure, a defect detection model for the spinning box can be trained and obtained using the first sample data set, and the defect detection model can be used to identify the second working image to determine that the second working image is a normal image or a defective image. Optionally, a confidence level of the normal image or the defective image can also be obtained.

[0033] In the embodiment of the present disclosure, the second working image is used to indicate the working state of the spinning box, and when the defect detection model obtains that the second working image is a normal image, the spinning box is in a normal working state, while when the defect detection model obtains that the second working image is a defective image, the spinning box is in an abnormal working state. The user can repair the spinning box in the abnormal working state. Illustratively, the defect detection model may be a feature probability model, or a neural network model based on deep learning, etc. The second working image is obtained by photographing one or more spinning boxes in real time by a patrol inspection robot or artificially, so as to detect the working state of one or more spinning boxes in real time.

[0034] Exemplarily, the first sample data set may include each defect image in a plurality of first work images, a partial image in each defect image, and an image obtained by cropping out the background of the spinning box in each defect image or the partial image.

[0035] According to the solution of the embodiment of the present disclosure, a defect image of the spinning box can be obtained by comparing the first working image of the spinning box with the normal image. A first sample data set for training a defect detection model can be obtained from the defect image. The method of comparing images eliminates the need to manually observe the spinning box for a long time, and can improve the efficiency of sample collection. In addition, since the abnormal area in the spinning box is minute, by scanning and comparing the image, if a difference that meets a predetermined condition occurs with the normal image, it can be determined to be abnormal, which is more accurate than the method of artificial identification. Therefore, the above method can improve the efficiency and accuracy of collecting sample data, and accordingly improve the training efficiency and model accuracy of the defect detection model. Furthermore, the detection accuracy of the defect detection model for the second work image of the spinning box is improved, and whether the second work image is a normal image or a defective image can be obtained through the defect detection model, so that it can timely and accurately determine whether the spinning box corresponding to the second work image is in a normal work state or an abnormal work state, and when the spinning box is in an abnormal work state, the user can be notified to repair the spinning box, and the abnormal work of the spinning box can be prevented from affecting the production of the spinning box.

[0036] In some embodiments, obtaining the first sample data set based on each defect image in the plurality of first working images includes: performing a classification for each defect image based on at least one defect type to obtain at least one set of images in one-to-one correspondence with the at least one defect type; and obtaining a first sample data set based on the at least one image set.

[0037] Optionally, the at least one defect type includes at least one of defects of the spinneret, the yarn, the yarn guide hook, and the oil nozzle in the spinning box. Specifically, the at least one defect type includes abnormalities such as yarn breakage, floating yarn, yarn hook, tilt of the oil nozzle, tilt of the yarn guide hook, and the yarn not in the yarn guide hook in the spinning box. For example, it may include that there are many items waiting to be sorted on the spinneret, floating yarn exists in the yarn, the position of the yarn guide hook is tilted, and the position of the oil nozzle is tilted. Exemplarily, when floating yarn exists in the yarn in the defect image, the defect image is in the corresponding image set where floating yarn exists in the yarn, and when the position of the yarn guide hook is tilted in the defect image, the defect image is in the corresponding image set where the position of the yarn guide hook is tilted.

[0038] Exemplarily, the pixel points of the defective image are compared with the pixel points of the normal image, and the difference between the pixel points of the defective image and the pixel points of the normal image can be more accurately obtained by comparing the pixel points of the defective image and the pixel points of the normal image, and the pixel points can be pixel points at all positions in the image, or pixel points at one or more positions in the image. For example, the pixel points of the spinneret of the defective image and the spinneret of the normal image can be compared to obtain the difference between the spinneret of the defective image and the spinneret of the normal image, for example, if the spinneret of the defective image has more items waiting to be sorted than the spinneret of the normal image, the defective image is in the corresponding image set with more items waiting to be sorted on the spinneret.

[0039] In one embodiment, at least one defect detection model has a one-to-one correspondence with at least one image set. That is, a defect detection model corresponding to a specific defect type can be obtained by training a model using each image set, and the defect detection model detects the second work image of the spinning box. The defect detection model can obtain whether a defect of a specific defect type exists in the second work image, and the specific defect type is the corresponding defect type of the image set used for training. It should be understood that the training data of the defect detection model can also include images in which the specific defect type does not exist, such as normal images or images corresponding to other defect types. According to the above embodiment, a defect detection model can be obtained by selecting multiple defect types according to the needs of a user and training on the defect types selected by the user. For example, the first sample data set includes an image set corresponding to a defect type in which the position of the yarn guide hook is tilted and an image set corresponding to a defect type in which the position of the oil nozzle is tilted, and a user can select the image set corresponding to the defect type in which the position of the yarn guide hook is tilted and perform training to obtain a defect detection model, and by detecting the second work image using this defect detection model, it can be determined whether the yarn guide hook of the spinning box corresponding to the second work image is tilted or not.

[0040] In one embodiment, one defect detection model can correspond to all image sets, and the defect detection model can detect the second work image of the spinning box, and the defect detection model can obtain whether the second work image has a defect and, if so, the defect type of the defect. For example, the first sample data set includes an image set corresponding to a defect type in which the position of the yarn guide hook is tilted and an image set corresponding to a defect type in which the position of the oil nozzle is tilted, and these image sets can be used to perform model training to obtain the defect detection model. The defect detection model can be used to detect the second work image, and can obtain whether the yarn guide hook and the oil nozzle of the spinning box corresponding to the second work image are tilted.

[0041] In one embodiment, if the defect detection model determines that the second work image is a defective image, the defect detection model can also determine a defect type of the second work image and a confidence level of the defect type. The confidence level of the defect type is used to indicate the probability that the defect type of the second work image is the defect type. For example, the defect confidence level of a spinneret can be the probability that a defect exists in the spinneret in the second work image.

[0042] According to the solution of the embodiment of the present disclosure, the first sample data set can include an image set of at least one defect type, and when the second working image is a defect image, the defect detection model can obtain the defect type of the second working image. The user can repair the spinning box corresponding to the second working image according to the defect type of the second working image, and can improve the repair efficiency of the spinning box in the abnormal working state.

[0043] In some embodiments, obtaining the first sample data set based on the at least one image set includes: For each image set of the at least one image set, crop each image in the image set based on a defect type corresponding to the image set to obtain a detection area image set corresponding to the image set; Obtaining a first sample data set based on at least one detection region image set in one-to-one correspondence with the at least one image set.

[0044] In an embodiment of the present disclosure, the detection area image set may include a set of images of the detection area after cropping each image in the image set, the detection area including a region of the defect type corresponding to the image set in each image in the image set. Here, the cropping may leave only a region of the specific defect type in the image and remove other parts. Exemplarily, if the defect type corresponding to the image set is a tilted position of the oil nozzle, the detection area may be the position of the oil nozzle in each image in the image set, and the cropped image of the position of the oil nozzle in each image in the image set may be the detection area image set.

[0045] Illustratively, the pixel points of each image detection region in the image set can be cropped to more accurately obtain a detection region image set that corresponds to the defect type in the image set.

[0046] According to the solution of an embodiment of the present disclosure, by performing cropping on each image in the image set based on the defect type corresponding to the image set, a defect detection model can be obtained by training only the detection area image set corresponding to the image set, thereby reducing the computational power required to obtain the defect detection model through training and reducing the training cost of the defect detection model.

[0047] In some embodiments, the method of processing the sample data comprises: determining a second sample data set from the first sample data set based on the feature amount of each image in the first sample data set when the reliability of the defect detection model does not satisfy the second condition; and updating the defect detection model based on the second sample data set.

[0048] In an embodiment of the present disclosure, the confidence level of the defect detection model may be determined based on the confidence level that the second working image obtained by the defect detection model is a normal image, a defect image, or the confidence level of the defect type of the defect image, etc. For example, if the model outputs information with low confidence level for one or more images, the model may be considered to have low confidence level.

[0049] Alternatively, the reliability of the defect detection model can be determined using the validation images, for example, by dividing the images of the first sample data set into training images and validation images, the training images are used to train and obtain the defect detection model, and the validation images are used to validate whether the defect detection model can accurately identify the defect conditions in the validation images, and the reliability of the defect detection model can be determined based on the proportion of accurate identification.

[0050] Exemplarily, when the reliability does not satisfy the second condition (e.g., when the reliability is below a threshold), the feature amount of the image can be determined based on the difference between the detection area image of each image and the corresponding area of ​​the normal image. For example, when the difference between the detection area image of an image in the image set and the corresponding area of ​​the normal image is large, it is determined that the feature amount of the image is relatively high. In this way, an image with a relatively high feature amount can be selected from the first sample data set to update the defect detection model.

[0051] According to the solution of the embodiment of the present disclosure, the defect detection model can be updated using the feature amounts of each image in the first sample data set, thereby improving the detection accuracy of the defect detection model.

[0052] In some embodiments, determining the second sample data set from the first sample data set based on the feature of each image in the first sample data set comprises: performing clustering on each image in the first sample data set to obtain a plurality of class clusters; determining, for each of the plurality of class clusters, a target image in the class cluster based on a feature amount of each image in the class cluster; The method can further include obtaining a second sample data set based on the target images in each class cluster.

[0053] According to the above embodiment, the difference in features between each class cluster is large, and by selecting a target image based on the image features in each class cluster, there is an obvious difference between the selected target images and the features are relatively high, so that the quality of the second sample data set can be further improved, and thus the accuracy of the defect detection model can be improved.

[0054] According to an embodiment of the present disclosure, the present disclosure further provides a sample data processing device, and Fig. 3 is a schematic block diagram showing a sample data processing device according to an embodiment of the present disclosure. As shown in Fig. 3, the sample data processing device includes: an image acquisition module 310 for acquiring a plurality of first operation images of the spinning box; a defective image determination module 320 for determining, for each of a plurality of first work images, the first work image as a defective image of the spinning box when there is a difference between the first work image and the normal image of the spinning box that satisfies a first condition; and a first sample processing module 330 for obtaining a first sample data set based on each defect image in a plurality of first work images, the first sample data set being used to train and obtain a defect detection model of the spinning box, and the defect detection model being used to detect whether a defect exists in a second work image of the spinning box.

[0055] In some embodiments, as shown in FIG. 4, the first sample processing module 330 includes: an image classification unit 410 for performing a classification for each defect image based on at least one defect type to obtain at least one image set in one-to-one correspondence with the at least one defect type; A sample processing unit 420 for obtaining a first sample data set based on the at least one image set.

[0056] In some embodiments, the at least one defect type includes at least one of defects in a spinneret, a yarn, a yarn guide hook, and an oil nozzle in a spinning box.

[0057] In some embodiments, the sample processing unit 420 specifically comprises: For each image set of the at least one image set, crop each image in the image set based on a defect type corresponding to the image set to obtain a detection area image set corresponding to the image set; and obtaining a first sample data set based on at least one detection region image set in one-to-one correspondence with the at least one image set.

[0058] In some embodiments, as shown in FIG. 5, the sample data processing device comprises: a second sample processing module 510 for determining a second sample data set from the first sample data set based on the feature amount of each image in the first sample data set when the reliability of the defect detection model does not satisfy the second condition; and a model updating module 520 for updating the defect detection model based on the second sample data set.

[0059] In some embodiments, the second sample processing module 510 specifically includes: performing clustering on each image in the first sample data set to obtain a plurality of class clusters; determining, for each of the plurality of class clusters, a target image in the class cluster based on a feature amount of each image in the class cluster; and obtaining a second sample data set based on the target images in each class cluster.

[0060] FIG. 6 is a structural block diagram of an electronic device according to an embodiment of the present disclosure. As shown in FIG. 6, the electronic device includes a memory 610 and a processor 620, and the memory 610 stores a computer program executable by the processor 620. The number of the memory 610 and the processor 620 may be one or more. The memory 610 may store one or more computer programs, which, when executed by the electronic device, cause the electronic device to perform the method provided by the above method embodiment. The electronic device may further include: The communication interface 630 is used for communicating with an external device and performing data interaction and transmission.

[0061] When the memory 610, the processor 620, and the communication interface 630 are implemented independently, the memory 610, the processor 620, and the communication interface 630 are connected to each other via a bus and can communicate with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus may be classified into an address bus, a data bus, a control bus, and the like. For ease of explanation, only one thick line is shown in FIG. 6, but this does not indicate only one bus or only one type of bus.

[0062] Optionally, in a specific implementation, when the memory 610, the processor 620, and the communication interface 630 are integrated on one chip, the memory 610, the processor 620, and the communication interface 630 can communicate with each other via an internal interface.

[0063] It should be understood that the processor may be a Central Processing Unit (CPU) or may be other general purpose processors, Digital Signal Processing (DSP), Application Specific Integrated Circuits (ASIC), Field Programmable Gate Arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware assemblies, etc. The general purpose processor may be a microprocessor or any conventional processor, etc. The processor may be a processor supporting the Advanced RISC Machines (ARM) architecture.

[0064] Additionally, optionally, the memory may include read-only memory and random access memory, and may further include non-volatile random access memory. The memory may be either volatile or non-volatile memory, or may optionally include both volatile and non-volatile memory, where non-volatile memory may include Read-Only Memory (ROM), Programmable ROM (PROM), Erasable PROM (EPROM), Electrically EPROM (EEPROM), or flash memory. Volatile memory may include Random Access Memory (RAM), which acts as an external cache. By way of example and not limitation, many forms of RAM are available. For example, static random access memory (Static RAM, SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (Enhanced SDRAM, ESDRAM), synchlink DRAM (Synchlink DRAM, SLDRAM), and direct RAMBUS RAM (DR RAM).

[0065] The above-mentioned embodiments may be implemented or acquired, in whole or in part, in software, hardware, firmware, or any combination thereof. When implemented in software, the whole or in part may be implemented or acquired in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, a process or function according to the embodiments of the present disclosure is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via wire (e.g., coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, Bluetooth, microwave, etc.). The computer-readable storage medium is any available medium that can be accessed and acquired by a computer, or a data storage device including a server, data center, etc. integrated with one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a digital versatile disk (DVD)), or a semiconductor medium (e.g., a solid state disk (SSD)). Note that the computer-readable storage medium referred to in this disclosure is a non-volatile storage medium, in other words, a non-transitory storage medium.

[0066] Those skilled in the art can understand that all or part of the steps of executing the above embodiments may be implemented by hardware, or may be implemented by instructing related hardware by a program, and the program may be stored in a computer-readable storage medium, and the storage medium may be a read-only memory, a magnetic disk, an optical disk, etc.

[0067] In the description of the embodiments of the present disclosure, the reference words "one embodiment", "several embodiments", "examples", "specific examples", or "several examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. And, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. Furthermore, a person skilled in the art may combine different embodiments or examples and features of different embodiments or examples described in the present disclosure to the extent that they are not mutually inconsistent.

[0068] In the description of the embodiments of the present disclosure, unless otherwise specified, " / " means "or", for example, A / B means and obtains either A or B. "And / or" in the present disclosure is merely for describing the related relationship of related objects, and indicates that three kinds of relationships may exist, for example, A and / or B can indicate the following three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0069] In the description of the embodiments of the present disclosure, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance, nor should they be construed as implying the number of technical features depicted. Thus, features defined as "first" and "second" explicitly or implicitly include and capture one or more of such features. In the description of the embodiments of the present disclosure, "plurality" means two or more, unless otherwise specified.

[0070] The above are merely illustrative examples of the present disclosure, and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure should be included within the scope of protection of the present disclosure.

Claims

1. 1. A method for processing sample data, comprising the steps of: acquiring a plurality of first working images of the spinning box; For each of the plurality of first work images, when there is a difference between the first work image and a normal image of the spinning box that satisfies a first condition, determining the first work image as a defective image of the spinning box; obtaining a first sample data set based on each defect image in the plurality of first working images, the first sample data set being used to train and obtain a defect detection model of the spinning box, and the defect detection model being used to detect whether a defect exists in a second working image of the spinning box; determining a second sample data set from the first sample data set based on a feature amount of each image in the first sample data set when the reliability of the defect detection model does not satisfy a second condition; updating the defect detection model based on the second sample data set. How to process sample data.

2. Obtaining a first sample data set based on each defect image in the plurality of first work images includes: performing a classification for each of the defect images based on at least one defect type to obtain at least one image set in one-to-one correspondence with the at least one defect type; obtaining the first sample data set based on the at least one image set. The method for processing sample data according to claim 1 .

3. The at least one defect type includes at least one of defects of a spinneret, a yarn, a yarn guide hook, and an oil nozzle in the spinning box; The method for processing sample data according to claim 2.

4. Obtaining the first sample data set based on the at least one image set includes: for each image set of the at least one image set, cropping each image in the image set based on a defect type corresponding to the image set to obtain a detection area image set corresponding to the image set; obtaining the first sample data set based on at least one detection area image set in one-to-one correspondence with the at least one image set; The method for processing sample data according to claim 2.

5. Determining a second sample data set from the first sample data set based on a feature amount of each image in the first sample data set includes: performing clustering on each image in the first sample data set to obtain a plurality of class clusters; determining, for each class cluster among the plurality of class clusters, a target image in the class cluster based on a feature amount of each image in the class cluster; The second sample data set is calculated based on the target images in each of the class clusters. and The method for processing sample data according to claim 1 .

6. A sample data processing device, comprising: an image capture module for capturing a first plurality of working images of the spinning box; a defect image determination module for determining, for each of the plurality of first work images, when there is a difference between the first work image and a normal image of the spinning box that satisfies a first condition, the first work image as a defect image of the spinning box; a first sample processing module for obtaining a first sample data set according to each defect image in the plurality of first working images, the first sample data set being used to train and obtain a defect detection model of the spinning box, and the defect detection model being used to detect whether a defect exists in a second working image of the spinning box; a second sample processing module for determining a second sample data set from the first sample data set based on a feature amount of each image in the first sample data set when the reliability of the defect detection model does not satisfy a second condition; a model updating module for updating the defect detection model based on the second sample data set. A device for processing sample data.

7. At least one processor; a memory in communication with the at least one processor; The memory stores instructions executable by the at least one processor, the instructions, when executed by the at least one processor, causing the at least one processor to perform a method according to any one of claims 1 to 5. Electronic devices.

8. A non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform the method of any one of claims 1 to 5.

9. A program for implementing the method according to any one of claims 1 to 5 when executed by a processor in a computer.

Citation Information

Patent Citations

  • Spinning machine fault monitoring methods

    CN110175659B

  • Abnormality detection method and device, electronic equipment and computer readable storage medium

    CN113688889A

  • Intelligent inspection device for chemical fiber spinning process yarn path

    CN210775245U

  • Autonomous robot for spinning workshop

    CN215281920U

  • Image classification device and image classification method

    JP2017054239A