Fiber atlas retrieval method and device based on polarized light image, equipment and medium
Through the fiber map retrieval method based on polarized light images, the convolutional neural network and similarity measurement algorithm are used to solve the problem of low classification efficiency of artificial fibers in the prior art, and efficient and accurate classification of fiber maps is achieved.
Patent Information
- Application Number
- CN202411470247.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-21
- Publication Date
- 2025-02-11
AI Technical Summary
In the existing technology, artificial fiber classification methods no longer meet the requirements of efficient identification, and appraisers' demand for library construction management and classification of fiber maps is becoming increasingly strong.
Provide a fiber map search method based on polarized light images. By constructing a fiber map data set that sets feature labels, training a convolutional neural network and/or image feature extraction network, combining a similarity metric algorithm to realize the retrieval of fiber images, and The fiber map is stored in the form of image feature vectors + text feature vectors to form a fiber map library.
It realizes efficient classification of fiber maps, and more accurate retrieval through model, reduces manual intervention, and improves retrieval efficiency and accuracy.
Smart Images
Figure CN119418109A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of fiber identification technology, and in particular to a fiber atlas retrieval method, device, equipment and medium based on polarized light images. Background Art
[0002] The polarization interference characteristics of the fiber are related to the material and surface morphology of the fiber. The type of the fiber can be judged based on the polarization interference characteristics of the fiber.
[0003] During the inspection, the inspector will use a stereo microscope to observe the color, thickness and shape of the fiber. Then the fiber sample is placed on a glass slide with glycerol drops and covered with a coverslip. The prepared glass slide is placed under a polarized light microscope for observation. In order to compare the uniformity of the fiber morphology, the fiber will be placed on the glass slide at a 45° angle to archive the polarized light photo. The polarized light microscope will output a polarized light image (or atlas) in jpg format; different types of fibers will present different morphologies in their polarized light photos.
[0004] In the current daily inspection process, identification personnel will judge the type of fiber based on the morphological characteristics of polarized light photographs and compare them with known fiber morphological patterns. However, with the increasing number of fiber types and classification tasks, manual classification can no longer meet the requirements of efficient identification. Identification personnel have an increasing demand for the database management and classification of fiber atlases. Summary of the invention
[0005] In order to overcome the problems existing in the related art, the present disclosure provides a fiber atlas retrieval method, device, equipment and medium based on polarized light images to solve the technical problem in the related art that manual fiber classification methods can no longer meet the requirements of efficient identification.
[0006] One or more embodiments of this specification provide a fiber atlas retrieval method based on polarized light images, comprising the following steps:
[0007] Dataset construction: constructing a fiber atlas dataset with feature labels, wherein the feature labels include atlas type labels and morphological feature labels including fiber types;
[0008] Model training: inputting the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and a similarity measurement algorithm is combined to realize the retrieval of fiber images;
[0009] Atlas storage: the data set is input into the fiber detection model training model to store the fiber atlas with feature labels in the form of image feature vector + text feature vector to obtain the fiber atlas library;
[0010] Image retrieval process: input the target fiber atlas with feature labels into the fiber detection model, calculate the similarity between the image feature vector extracted according to the feature labels and each fiber atlas in the fiber atlas library to obtain the top m target image feature vectors, thereby obtaining the image retrieval results, where m is a natural number greater than 1; and
[0011] Category determination: The appraiser determines the category of the target fiber atlas based on the image retrieval results.
[0012] Furthermore, the construction of the fiber atlas dataset with feature labels specifically includes the following steps:
[0013] Based on the obtained fiber sample, a fiber atlas is collected by a polarized light image collection device;
[0014] Add atlas type labels and morphological feature labels to the collected fiber atlases, where atlas types include standard samples and test samples, and morphological feature labels include type, color, overall morphology, surface morphology, internal morphology, cross-section, and other special morphological information labels.
[0015] Furthermore, the method also includes the steps of: classifying the images in the fiber atlas library by a clustering algorithm to obtain various sub-datasets, and calculating the average value of all image feature vectors in the sub-datasets as the first image feature vector of the sub-dataset, and creating an index file for each sub-dataset and storing it in the fiber atlas library, and using the first image feature vector as the comparison object for similarity calculation in the image retrieval process.
[0016] Furthermore, the fiber detection model is a convolutional neural network and an image feature extraction network, which respectively extract the first image feature vector and the second image feature vector in the fiber atlas according to the feature labels through the convolutional neural network and the image feature extraction network, and then calculate the similarity between the first image feature vector and the second image feature vector and each fiber atlas in the fiber atlas library, and obtain the top m first target image feature vectors and the second target image feature vectors, respectively, and calculate and calculate the first target image feature vector and the second target image feature vector to obtain the target atlas and display it.
[0017] Furthermore, it also includes image preprocessing of the fiber atlas in the data set to obtain a preprocessed image, wherein the preprocessing includes denoising, image enhancement and normalization preprocessing operations.
[0018] One or more embodiments of this specification provide a fiber atlas retrieval device based on polarized light images, including:
[0019] A data set construction module is used to construct a fiber atlas data set with feature labels, wherein the feature labels include atlas type labels and morphological feature labels including fiber types;
[0020] A model training module is used to input the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and a similarity measurement algorithm is combined to realize the retrieval of fiber images;
[0021] A library construction module is used to input the data set into the fiber detection model training model to store the fiber atlas with feature labels in the form of image feature vector + text feature vector to obtain a fiber atlas library;
[0022] A graph retrieval module is used to input a target fiber graph with a feature label into a fiber detection model, and perform similarity calculation between an image feature vector extracted according to the feature label and each fiber graph in a fiber graph library to obtain the top m target image feature vectors, thereby obtaining a search result of an image, wherein m is a natural number greater than 1; and
[0023] The category confirmation module is used by identification personnel to determine the category of the target fiber atlas based on the image retrieval results.
[0024] Furthermore, the data set construction module includes a fiber atlas acquisition submodule and a label adding submodule, which are specifically configured to perform the following steps:
[0025] A fiber atlas acquisition submodule is used to acquire a fiber atlas through a polarized light image acquisition device based on the acquired fiber sample;
[0026] The label adding submodule is used to add atlas type labels and morphological feature labels to the collected fiber atlases, where the atlas types include standard samples and test samples, and the morphological feature labels include type, color, overall morphology, surface morphology, internal morphology, cross-section and other special morphological information labels.
[0027] Furthermore, the fiber detection model is a convolutional neural network and an image feature extraction network. The atlas retrieval module extracts the first image feature vector and the second image feature vector in the fiber atlas according to the feature labels through the convolutional neural network and the image feature extraction network, and then calculates the similarity between the first image feature vector and the second image feature vector and each fiber atlas in the data set, and obtains the top m first target image feature vectors and the second target image feature vectors, and obtains the target atlas after calculating the first target image feature vector and the second target image feature vector.
[0028] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the fiber atlas retrieval method based on polarized light images as described in any one of the above when executing the computer program.
[0029] One or more embodiments of the present specification provide a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the fiber atlas retrieval method based on polarized light images as described in any one of the above items is implemented.
[0030] The present disclosure provides a fiber atlas retrieval method, device, equipment and medium based on polarized light images, which have the advantages of adding text label data representing the properties of the fiber atlas to various collected fiber atlases to form a data set, and the data set is used to train the fiber detection model. The model can learn the connection between the image and the text concept, making the model retrieval more accurate, and the data set is used as an object for comparing the similarity of the target fiber atlas, thereby obtaining a preset number of target images with the highest similarity to the target fiber atlas. The identification personnel can refer to the target image to quickly determine the category of the target fiber atlas, thereby realizing efficient classification of the fiber atlas, and the fiber atlas classification realized by the model is more accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0032] Figure 1 A flowchart of a fiber atlas retrieval method based on polarized light images provided for one or more embodiments of this specification;
[0033] Figure 2 A block diagram of a fiber atlas retrieval device based on polarized light images provided for one or more embodiments of this specification;
[0034] as well as
[0035] Figure 3 A schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. DETAILED DESCRIPTION
[0036] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0037] The present invention is described in detail below in conjunction with specific implementation methods and the accompanying drawings.
[0038] Method Embodiment
[0039] According to an embodiment of the present invention, a fiber atlas retrieval method based on polarized light images is provided. Figure 1 FIG. 1 is a flow chart of a fiber atlas retrieval method based on polarized light images provided in this embodiment. The fiber atlas retrieval method based on polarized light images according to an embodiment of the present invention comprises the following steps:
[0040] Step S1, constructing a fiber atlas dataset with feature labels, wherein the feature labels include atlas type labels and morphological feature labels including fiber types.
[0041] Step S2, inputting the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and a similarity measurement algorithm is combined to realize the retrieval of fiber images.
[0042] Step S3, inputting the data set into the fiber detection model training model to store the fiber atlas with feature labels in the form of image feature vector + text feature vector to obtain a fiber atlas library.
[0043] Step S4, input the target fiber atlas with feature labels into the fiber detection model, calculate the similarity between the image feature vector extracted according to the feature labels and each fiber atlas in the data set to obtain the top m target image feature vectors, thereby obtaining the image retrieval results, where m is a natural number greater than 1.
[0044] Step S5: the identification personnel determine the category of the target fiber atlas based on the image retrieval results.
[0045] The fiber atlas retrieval method based on polarized light images provided in this embodiment adds feature label data representing the properties of the fiber atlas to various collected fiber atlases to construct a data set. The data set is used to train the fiber detection model. The model can learn the connection between the image and the text concept, making the model retrieval more accurate. A fiber atlas library is constructed by storing the data set in a vector format. The target fiber atlas is used as an object for comparing similarity. Based on the comparison of image feature vectors, a preset number of target images with the highest similarity to the target fiber atlas are obtained. The identification personnel can refer to the target image to quickly determine the category of the target fiber atlas, thereby realizing efficient classification of the fiber atlas, and the fiber atlas classification realized by the model is more accurate.
[0046] In this embodiment, step S1 constructs a fiber atlas dataset with feature labels and specifically includes the following steps:
[0047] Step S11, based on the acquired fiber sample, collecting a fiber atlas through a polarized light image collection device;
[0048] In this embodiment, since fibers of different types and structures, processing techniques and orientations have different polarized light interference characteristics, when collecting the atlas of various types of fibers through step S1, it is necessary to collect atlases of different structures, processing techniques and placement positions for each fiber, which will be used for subsequent steps to build a fiber atlas library and model training. Because the sample information presented by different structures and different processing techniques is different, in order to make the trained model applicable, the training of the model not only needs to let it learn the fiber morphological characteristics under different conditions. Therefore, step S11 is specifically as follows:
[0049] Obtain fiber samples of various structures and processing techniques, and use a polarized light image acquisition device to collect fiber maps based on multiple placement positions.
[0050] Step S12, adding a map type label and a morphological feature label to the collected fiber map; the map type includes a standard sample and a test sample, and the morphological feature label includes type, color, overall morphology, surface morphology, internal morphology, cross section and other special morphological information labels.
[0051] In this embodiment, the morphological feature label describes the appearance of the fiber under a polarized light microscope from various dimensions. The feature label can be set through the feature area selection tool. The feature area selection box can be 100*100px in size, and the corresponding feature setting corresponding serial number is set. Multiple feature areas can be selected and saved. These saved feature areas will be compared as key comparison areas. In this embodiment, other information of the atlas can be added according to needs, such as inspection instruments and equipment, inspectors, atlas numbers and other information. After supplementation, the fiber atlas storage process can be completed.
[0052] In one embodiment, the fiber detection model may be a convolutional neural network or an image feature extraction network, and combined with a similarity measurement algorithm to realize the retrieval of fiber images. The specific retrieval process may be to extract the image feature vector corresponding to the feature label through a convolutional neural network or an image feature extraction network, and then calculate the similarity between the image feature vector and each fiber atlas in the data set to obtain the top m target image feature vectors, and obtain the m target images and the corresponding feature label information according to the index relationship.
[0053] In this embodiment, the convolutional neural network can be any type of neural network, such as a Visual Geometry Group (VGG) network, a residual network, etc. In a specific example, a VGG16 network is used, and a training data set containing various fiber images is first used to train the VGG16 network to obtain a trained convolutional neural network. The image feature extraction network can be a YOLO model, a HOG feature extraction method, such as using opencv to extract hog features, and then using the feature descriptor of the hog feature and the similarity between each image feature vector in the fiber atlas library to calculate, thereby obtaining the m target images with the highest similarity and displaying them, which is used for identification personnel to determine the category of the target fiber atlas according to the image retrieval results, wherein the target image is displayed in order from high to low similarity, and supports tools for zooming in / out, rotating angles, and displaying feature labels, while viewing the target atlas and the candidate atlas, and the fiber categories include common fiber types such as cotton, linen, wool, rabbit hair, polyester, nylon, polypropylene, vinylon, acrylic, and chlorofiber.
[0054] In this embodiment, in order to improve the retrieval accuracy, the fiber detection model can be a convolutional neural network and an image feature extraction network, which respectively extract the first image feature vector and the second image feature vector in the fiber atlas according to the text label through the convolutional neural network and the image feature extraction network, and then calculate the similarity between the first image feature vector and the second image feature vector and each fiber atlas in the fiber atlas library, respectively, to obtain the top m first target image feature vectors and the second target image feature vectors, respectively, and calculate and calculate the first target image feature vector and the second target image feature vector to obtain the target atlas and display it.
[0055] In this embodiment, in order to improve the retrieval efficiency, the images in the data set are classified by a clustering algorithm to obtain various sub-data sets, the average value of all image feature vectors in the sub-data set is calculated as the first image feature vector of the sub-data set, and an index file is created for each sub-data set and stored in the fiber atlas library, and the first image feature vector is used as the comparison object for similarity calculation in the image retrieval process, that is, in the similarity calculation process, the image feature vector of the target fiber atlas is obtained according to the fiber detection model, and the image feature vector is compared with the first image feature vector for similarity to obtain the target sub-data set with the highest similarity, and then the similarity between the image feature vector and each image feature vector in the target sub-data set is calculated, and according to a preset threshold, the top m target image feature vectors ranked in the calculation result are selected, and finally the images corresponding to the m target image feature vectors and the text information representing the fiber atlas characteristics are obtained. In the comparison retrieval process, the retrieval efficiency can be greatly improved through index retrieval.
[0056] In this embodiment, in order to enable the fiber detection model to realize the retrieval of fiber atlases without feature labels, the fiber detection model can use the Clip (Contrastive language-image pre-training) model + classifier combination as the basic model, use the data set of text labels for setting fiber atlas attributes as input for training, and use the Clip visual language large model as the encoding model of the data pair to obtain a data set consisting of fiber atlases and feature labels characterizing the characteristics of the fiber atlas. The trained Clip model can flexibly meet the image retrieval text, text retrieval image or image and text fusion feature mixed retrieval, so that the fiber retrieval can more flexibly meet the needs of users and the model retrieval is more accurate.
[0057] When using the Clip visual language large model to construct a fiber atlas library, first obtain the first image feature vector corresponding to the fiber image feature and the second feature vector corresponding to each feature label, and calculate the weighted sum of the first image feature vector and the second feature vector according to a preset vector weight ratio to obtain a fused feature vector. The fused feature vector is clustered based on a clustering algorithm to obtain a sub-dataset, and the average of all the fused feature vectors in the sub-dataset is taken as the first image feature vector of the sub-dataset for retrieval; based on the retrieval results, the feature label information of the corresponding image can be determined and displayed according to the corresponding second feature vector for identification personnel as a reference for judgment. The training process of the Clip model + classifier is a conventional technology and will not be elaborated here.
[0058] In this embodiment, it also includes image preprocessing of the fiber atlas in the data set to obtain a preprocessed image, including: denoising the fiber atlas to obtain a denoised image, such as by using denoising algorithms such as mean filtering and Gaussian filtering; performing image enhancement on the denoised image to obtain an enhanced image, such as by using image enhancement algorithms such as histogram equalization and Retinex algorithm.
[0059] The fiber atlas can also be normalized. Due to the influence of light in the atlas acquisition process, the local surface exposure contributes a large proportion to the texture intensity of the atlas. Therefore, normalization can effectively reduce the local shadows and lighting changes in the image. Because color information is not very useful, it is usually converted into a grayscale image first.
[0060] This embodiment provides a fiber atlas comparison method based on text labels to solve the different interference patterns presented by various types of fibers under a polarized light fiberscope. By screening and labeling feature areas, and integrating deep learning of known fiber patterns with traditional comparison algorithms, similarity calculation between multiple fiber atlases is achieved, thereby providing comparison results for inspectors to assist in the identification of fiber types.
[0061] Device Embodiment
[0062] According to an embodiment of the present invention, a fiber atlas retrieval device based on polarized light images is provided. Figure 2 FIG. 1 is a block diagram of a fiber atlas retrieval device based on polarized light images provided in this embodiment. The fiber atlas retrieval device based on polarized light images according to an embodiment of the present invention comprises:
[0063] The data set construction module 10 is used to construct a fiber atlas data set with characteristic labels, wherein the characteristic labels include atlas type labels and morphological characteristic labels including fiber types.
[0064] The model training module 20 is used to input the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and is combined with a similarity measurement algorithm to realize the retrieval of fiber images.
[0065] The library construction module 30 is used to input the data set into the fiber detection model training model to store the fiber atlas with set feature labels in the form of image feature vector + text feature vector to obtain the fiber atlas library.
[0066] The atlas retrieval module 40 is used to input the target fiber atlas with feature labels into the fiber detection model, and calculate the similarity between the image feature vector extracted according to the text label and each fiber atlas in the fiber atlas library to obtain the top m target image feature vectors, thereby obtaining the image retrieval results, wherein m is a natural number greater than 1.
[0067] The category confirmation module 50 is used by the identification personnel to determine the category of the target fiber atlas according to the image retrieval results.
[0068] The fiber atlas retrieval device based on polarized light images provided in this embodiment adds text label data representing the properties of the fiber atlas to various collected fiber atlases through the data set construction module 10 to build a data set. The model training module 20 trains the fiber detection model through the data set so that the model can learn the connection between the image and the text concept, and the model retrieval is more accurate. The atlas retrieval module 40 uses the fiber atlas library as the object for comparing the similarity of the target fiber atlas, thereby obtaining a preset number of target images with the highest similarity to the target fiber atlas. Finally, the identification personnel can refer to the target image to quickly determine the category of the target fiber atlas, thereby realizing efficient classification of the fiber atlas, and the fiber atlas classification realized by the model is more accurate.
[0069] In this embodiment, the data set construction module 10 includes a fiber atlas acquisition submodule 110 and a label adding submodule 120, which are specifically configured to perform the following steps:
[0070] The fiber atlas acquisition submodule 110 is used to acquire a fiber atlas through a polarized light image acquisition device based on the acquired fiber sample;
[0071] The label adding submodule 120 is used to add the atlas type label and the morphological feature label to the collected fiber atlas, wherein the atlas type includes the standard sample and the inspection material sample, and the morphological feature label includes the type, color, overall morphology, surface morphology, internal morphology, cross section and other special morphological information.
[0072] In this embodiment, the morphological feature label describes the appearance of the fiber under a polarized light microscope from various dimensions. The feature label can be set through the feature area selection tool. The feature area selection box can be 100*100px in size, and the corresponding feature setting corresponding serial number is set. Multiple feature areas can be selected and saved. These saved feature areas will be compared as key comparison areas. In this embodiment, other information of the atlas can be added according to needs, such as inspection instruments and equipment, inspectors, atlas numbers and other information. After supplementation, the fiber atlas storage process can be completed.
[0073] In this embodiment, the fiber detection model can be a convolutional neural network or an image feature extraction network, and can be combined with a similarity measurement algorithm to achieve retrieval of fiber images.
[0074] In another embodiment, in order to improve the retrieval accuracy, the fiber detection model can be a convolutional neural network and an image feature extraction network, which respectively extract the first image feature vector and the second image feature vector in the fiber atlas according to the text label through the convolutional neural network and the image feature extraction network, and then calculate the similarity of the first image feature vector and the second image feature vector with each fiber atlas in the fiber atlas library, respectively, to obtain the top m first target image feature vectors and the second target image feature vectors, respectively, and then obtain the target atlas after merging the first target image feature vector and the second target image feature vector and display it.
[0075] In this embodiment, in order to improve the retrieval efficiency, a feature classification module 60 is further provided, which is used to classify the images in the data set constructed by the data set construction module 10 through a clustering algorithm to obtain various sub-data sets, and to calculate the average value of all image feature vectors in the sub-data set as the first image feature vector of the sub-data set, and to create an index file for each sub-data set and store it in the fiber atlas library, so that the atlas retrieval module 40 can greatly improve the retrieval efficiency through index retrieval during the comparison retrieval process;
[0076] This embodiment further provides a preprocessing module 70 for performing image preprocessing on the fiber atlas collected by the data set building module 10 to obtain a preprocessed image, wherein the image preprocessing includes preprocessing operations such as denoising, image enhancement and normalization.
[0077] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.
[0078] like Figure 3 As shown, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the fiber atlas retrieval method based on polarized light images in the above-mentioned embodiment is implemented.
[0079] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fiber atlas retrieval method based on polarized light images in the above-mentioned embodiment, or, when executed by a processor, implements the fiber atlas retrieval method based on polarized light images in the above-mentioned embodiment.
[0080] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0081] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.
Claims
1. A fiber atlas retrieval method based on polarized light images, characterized in that The following steps are involved: Dataset construction: constructing a dataset of fiber atlases with feature labels, wherein the feature labels include atlas type labels and morphological feature labels including fiber types; Model training: inputting the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and a similarity measurement algorithm is combined to realize the retrieval of fiber images; Atlas storage: the data set is input into the fiber detection model training model to store the fiber atlas with feature labels in the form of image feature vector + text feature vector to obtain the fiber atlas library; Image retrieval process: input the target fiber atlas with feature labels into the fiber detection model, calculate the similarity between the image feature vector extracted according to the feature labels and each fiber atlas in the fiber atlas library to obtain the top m target image feature vectors, thereby obtaining the image retrieval results, where m is a natural number greater than 1; and Category determination: The appraiser determines the category of the target fiber atlas based on the image retrieval results.
2. The fiber atlas retrieval method based on polarized light images according to claim 1, characterized in that: The construction of the fiber atlas dataset with feature labels specifically includes the following steps: Based on the obtained fiber sample, a fiber atlas is collected by a polarized light image collection device; Add atlas type labels and morphological feature labels to the collected fiber atlases, where atlas types include standard samples and test samples, and morphological feature labels include type, color, overall morphology, surface morphology, internal morphology, cross-section, and other special morphological information labels.
3. The fiber atlas retrieval method based on polarized light images according to claim 1, characterized in that: Also includes the steps: The images in the fiber atlas library are classified by clustering algorithm to obtain various sub-datasets, and the average value of all image feature vectors in the sub-dataset is taken as the first image feature vector of the sub-dataset. An index file is created for each sub-dataset and stored in the fiber atlas library. The first image feature vector is used as the comparison object for similarity calculation in the image retrieval process.
4. The fiber atlas retrieval method based on polarized light images according to claim 1, characterized in that: The fiber detection model is a convolutional neural network and an image feature extraction network, which respectively extracts the first image feature vector and the second image feature vector in the fiber atlas according to the feature label through the convolutional neural network and the image feature extraction network, and then calculates the similarity between the first image feature vector and the second image feature vector and each fiber atlas in the fiber atlas library, respectively, and obtains the first target image feature vector and the second target image feature vector that rank first m, respectively, and calculates and calculates the first target image feature vector and the second target image feature vector to obtain the target atlas and display it.
5. The fiber atlas retrieval method based on polarized light images according to claim 1, characterized in that: The method also includes image preprocessing of the fiber atlas in the data set to obtain a preprocessed image, wherein the preprocessing includes denoising, image enhancement and normalization preprocessing operations.
6. A fiber atlas retrieval device based on polarized light images, characterized in that include: A data set construction module is used to construct a fiber atlas data set with feature labels, wherein the feature labels include atlas type labels and morphological feature labels including fiber types; A model training module is used to input the data set into a pre-built fiber detection model training model to obtain a trained fiber detection model, wherein the fiber detection model is a convolutional neural network and / or an image feature extraction network, and a similarity measurement algorithm is combined to realize the retrieval of fiber images; A library construction module is used to input the data set into the fiber detection model training model to store the fiber atlas with feature labels in the form of image feature vector + text feature vector to obtain a fiber atlas library; A graph retrieval module is used to input a target fiber graph with a feature label into a fiber detection model, and perform similarity calculation between an image feature vector extracted according to the feature label and each fiber graph in a fiber graph library to obtain the top m target image feature vectors, thereby obtaining a search result of an image, wherein m is a natural number greater than 1; and The category confirmation module is used by identification personnel to determine the category of the target fiber atlas based on the image retrieval results.
7. The fiber atlas retrieval device based on polarized light images according to claim 6, characterized in that: The data set construction module includes a fiber atlas acquisition submodule and a label addition submodule, which are specifically configured to perform the following steps: A fiber atlas acquisition submodule is used to acquire a fiber atlas through a polarized light image acquisition device based on the acquired fiber sample; The label adding submodule is used to add atlas type labels and morphological feature labels to the collected fiber atlases, where the atlas types include standard samples and test samples, and the morphological feature labels include type, color, overall morphology, surface morphology, internal morphology, cross-section and other special morphological information labels.
8. The fiber atlas retrieval device based on polarized light images according to claim 6, characterized in that: The fiber detection model is a convolutional neural network and an image feature extraction network. The atlas retrieval module extracts the first image feature vector and the second image feature vector in the fiber atlas according to the feature labels through the convolutional neural network and the image feature extraction network, and then calculates the similarity between the first image feature vector and the second image feature vector and each fiber atlas in the data set, respectively, to obtain the top m first target image feature vectors and the second target image feature vectors, and obtains the target atlas after calculating the first target image feature vector and the second target image feature vector.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the fiber atlas retrieval method based on polarized light images as described in any one of claims 1 to 5 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the fiber atlas retrieval method based on polarized light images as described in any one of claims 1 to 5 is implemented.