Hydroelectric equipment magnetic particle detection method and system based on AI image recognition

Abstract

The invention belongs to the technical field of hydroelectric equipment magnetic particle detection, and particularly relates to a hydroelectric equipment magnetic particle detection method and system based on AI image recognition. The method comprises the steps of obtaining an image of hydroelectric equipment magnetic particle detection and preprocessing the image to obtain an enhanced image; unobvious features are highlighted, interference of environmental factors is reduced, and information in the image is easier to identify; and detail information, especially defect information, cannot be lost. Inputting the enhanced image into a preset neural network model for network training to obtain network parameters; according to the network parameters, defect detection of the enhanced image is carried out, and an area where defects possibly exist in the image, the probability of the defects existing in the area, and the type and size information of the defects are obtained. According to the defect identification method, part of network parameters are randomly sampled through the preset neural network model to perform network iteration optimization training, and training data are regenerated from a detected result, so that the problem of insufficient training data is solved, dependence on data volume is reduced, and the defect identification accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of magnetic particle testing for hydropower equipment, and particularly relates to a magnetic particle testing method and system for hydropower equipment based on AI image recognition. Background Art

[0002] Hydropower equipment needs to be regularly subjected to magnetic particle testing during initial installation and its life cycle to check for surface or near-surface defects near the welds, and its main manifestation is defects.

[0003] Magnetic particle testing is to apply an external magnetic field to the ferromagnetic material to be tested to make it magnetic, and then apply magnetic powder or magnetic suspension fluid, and observe the distribution of surface magnetic marks to judge whether there are defects. If there are defects, holes and other defects on the surface and near the surface of the material, these defects will interrupt the magnetic force lines inside the material and form magnetic leakage of the magnetic field. The magnetic powder will be affected by the magnetic leakage field and accumulate at the defect or defect position, forming obvious magnetic powder aggregation traces for easy observation and analysis.

[0004] Traditional magnetic particle testing technology completely relies on manual operation and visual judgment, which is time-consuming and laborious and has many limitations. It is very difficult to erect a scaffolding inside the equipment, and there are major safety accidents such as high-altitude falls for the testing personnel; the ventilation inside the equipment is poor, and the toxic and harmful gases contained in the equipment will cause damage to personnel or even cause death; the working environment is harsh, and there are phenomena of human missed inspections and misjudgments in manual testing.

[0005] AI image recognition is a technology in which a computer understands, analyzes and processes images through artificial intelligence algorithms, and can imitate human visual ability to perform the testing work of hydropower equipment.

[0006] AI image recognition technology mainly includes four processes: image acquisition, image processing, feature extraction, and object recognition. Except for image acquisition, the other three processes are implemented through one or several neural network algorithms, and finally the computer obtains and recognizes the image information.

[0007] At present, the AI image recognition technology has a relatively poor recognition effect on objects with insignificant features in a complex environment, and there is a lack of sufficient data sets for training. The images of magnetic particle testing for hydropower equipment are exactly those with a complex environment, many pseudo-defects and difficult to uniformly characterize defects, so the testing difficulty is very high.

[0008] At this stage, it is necessary to develop an AI image recognition method and a supporting hardware system to solve the above problems. Summary of the Invention

[0009] The object of the present invention is to provide a magnetic particle detection method and system for hydropower equipment based on AI image recognition, so as to solve the technical problems that the prior art has poor recognition effect on objects with insignificant features in complex environments and low accuracy due to lack of sufficient data sets for training.

[0010] To achieve the above object, the present invention adopts the following technical solutions: In the first aspect, the present application discloses a magnetic particle detection method for hydropower equipment based on AI image recognition, including: Obtaining an image of magnetic particle detection of hydropower equipment and performing preprocessing to obtain an enhanced image; Inputting the enhanced image into a preset neural network model for network training to obtain network parameters; According to the network parameters, defect detection of the enhanced image is performed to obtain the area where defects may exist in the image, the probability of defects existing in this area, the type and size information of the defects.

[0011] Preferably, the step of inputting the enhanced image into a preset neural network model for network training specifically includes: S201: Manually annotating the enhanced image to mark the areas with defects and the types of defects; S202: Extracting features from the annotated enhanced image to obtain the areas with defects and the types of defects; S203: Classifying the areas with defects according to the types of defects to obtain classified defect areas and their corresponding defect types; S204: Performing object detection on the classified defect areas to obtain the network parameters of this classified defect area; S205: Randomly sampling and feeding back the network parameters in S204 to S201 for iterative optimization training.

[0012] Preferably, the network parameters include: the area where defects exist, the type and size information of the defects.

[0013] Preferably, the step of performing defect detection on the enhanced image according to the network parameters specifically includes: S301: The convolutional neural network extracts features from the enhanced image to obtain enhanced image features; S302: Based on the network parameters, judge the image features to obtain the area where defects may exist; S303: For the area where defects may exist, based on the network parameters, judge the probability of defects existing in this area, the type and size information of the defects.

[0014] Preferably, the preprocessing is image feature enhancement, specifically: Extract features from the acquired image to obtain image features; The image features are respectively processed by histogram equalization, contrast stretching, gamma transformation, Laplacian operator, high-pass filtering, sharpening, and low-pass filtering to obtain corresponding processed features; The processed features are respectively subjected to weighted fusion and stitching fusion to obtain a weighted fusion feature and a stitching fusion feature; The weighted fusion feature and the stitching fusion feature are superimposed and then image decoding is performed to obtain an enhanced image.

[0015] Preferably, under the illumination of a light source, an image for magnetic particle inspection of a hydroelectric equipment is captured by a camera.

[0016] In a second aspect, the present application discloses a magnetic particle inspection system for hydroelectric equipment based on AI image recognition, including: An image acquisition and preprocessing unit, configured to acquire an image for magnetic particle inspection of a hydroelectric equipment and perform preprocessing to obtain an enhanced image; A network training unit, configured to input the enhanced image into a preset neural network model for network training to obtain network parameters; A defect detection unit, configured to perform defect detection on the enhanced image according to the network parameters to obtain the area where defects may exist in the image, the probability of defects existing in this area, the type and size information of the defects.

[0017] Preferably, in the defect detection unit, according to the network parameters, defect detection of the enhanced image is specifically as follows: A convolutional neural network extracts features from the enhanced image to obtain enhanced image features; Based on the network parameters, a judgment is made on the image features to obtain the area where defects may exist; For the area where defects may exist, based on the network parameters, the probability of defects existing in this area, the type and size information of the defects are judged.

[0018] In a third aspect, the present application discloses an electronic 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 steps of the magnetic particle inspection method for hydroelectric equipment based on AI image recognition described in any one of the above are implemented.

[0019] In a fourth aspect, the present application discloses a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the magnetic particle inspection method for hydroelectric equipment based on AI image recognition described in any one of the above are implemented.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present application discloses a magnetic particle detection method for hydropower equipment based on AI image recognition. Through image preprocessing, an enhanced image is obtained, which highlights the unobvious features, reduces the interference of environmental factors, and makes the information in the image easier to identify; no detailed information, especially defect information, will be lost. By randomly sampling some network parameters through a preset neural network model for network iterative optimization training, the detection results are regenerated into training data to solve the problem of insufficient training data and reduce the dependence on the data volume. Moreover, according to the network parameters output by the iteratively optimized neural network model, the image features are identified and judged, improving the accuracy of defect recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a schematic flow chart of the method of the present invention; Figure 2 It is a schematic flow chart of an embodiment of the present invention; Figure 3 It is a schematic preprocessing flow chart of an embodiment of the present invention; Figure 4 It is a schematic processing flow chart of the defect detection unit of an embodiment of the present invention; Figure 5 It is a schematic random sampling processing flow chart of an embodiment of the present invention; Figure 6 It is a schematic principle diagram of image acquisition implementation of an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0024] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0025] It should be noted that like reference numerals and letters refer to like items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0026] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figures, or the orientation or positional relationship in which the product of the present invention is usually placed during use, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation on the present invention. In addition, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0027] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0028] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, if terms such as "set", "installed", "connected", "connected" are understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0029] The present invention will be further described in detail below with reference to the accompanying drawings: See Figure 1 , a magnetic particle detection method for hydropower equipment based on AI image recognition, including: S1: Obtain the image of the magnetic particle detection of the hydropower equipment and perform preprocessing to obtain an enhanced image; S2: Input the enhanced image into a preset neural network model for network training to obtain network parameters; S3: According to the network parameters, perform defect detection on the enhanced image to obtain the area where defects may exist in the image, the probability of defects existing in this area, the type and size information of the defects.

[0030] The present application discloses a magnetic particle detection method for hydropower equipment based on AI image recognition. After image preprocessing, an enhanced image is obtained, which highlights the unobvious features, reduces the interference of environmental factors, and makes the information in the image easier to identify; no detailed information, especially defect information, will be lost. Through a preset neural network model, some network parameters are randomly sampled for network iterative optimization training, and the detection results are regenerated into training data to solve the problem of insufficient training data and reduce the dependence on the data volume. Moreover, according to the network parameters output by the iteratively optimized neural network model, the image features are identified and judged, improving the accuracy of defect recognition.

[0031] In some embodiments, inputting the enhanced image into a preset neural network model for network training specifically includes: S201: Manually annotate the enhanced image to mark the defective areas and defect types. S202: Extract features from the annotated enhanced image to obtain the defective areas and defect types. S203: Classify the defective areas according to the defect types to obtain the classified defective areas and their corresponding defect types. S204: Perform object detection on the classified defective areas to obtain the network parameters of the classified defective areas. S205: Randomly sample the network parameters in S204 and feedback them to S201 for iterative optimization training.

[0032] In some embodiments, the network parameters include: the defective areas, the types and size information of the defects.

[0033] In some embodiments, performing defect detection on the enhanced image according to the network parameters specifically includes: S301: The convolutional neural network extracts features from the enhanced image to obtain enhanced image features. S302: Based on the network parameters, judge the image features to obtain the areas where defects may exist. S303: For the areas where defects may exist, based on the network parameters, judge the probability of defects existing in the area, the types and size information of the defects.

[0034] In some embodiments, the basic principles of AI image recognition methods are similar, but for the recognition method for magnetic particle detection of hydropower equipment, three problems must be solved. One is image enhancement, the second is the sensitivity to defect features, and the third is that training can be completed with a smaller data set. Therefore, the following specific implementation method is designed. Figure 2 This method includes two parts, model training and defect detection.

[0035] The first step in model training and defect detection is image acquisition. That is, magnetic traces are generated on the surface of the object to be measured by a magnetic particle testing device. Under the illumination of a light source, the magnetic traces become more obvious, and a camera captures the magnetic trace images and transmits them into the neural network; The second step in model training and defect detection is image preprocessing. Through various methods, image enhancement is performed to make the information in the images easier to identify; The third step in model training is network training. The areas containing defects and defect types in the preprocessed images are labeled, and network learning is carried out to obtain network parameters; (including 5 sub-steps: manual annotation, manually annotating the areas and defect types with defects in the preprocessed images; feature extraction, a convolutional neural network extracts features from the images; classification and sorting, the convolutional neural network classifies the features; object detection, judging information such as the probability, type, and size of defects existing in the selected area; random sampling, converting the training results into new training data through the neural network) The third step in defect detection is defect identification. According to the network parameters obtained from model training, the areas that may have defects, the probability of having defects, and information such as defect types and sizes in the preprocessed images are automatically labeled. (including 3 sub-steps, feature extraction, a convolutional neural network extracts features from the images; classification and sorting, the convolutional neural network classifies the features; object detection, judging information such as the probability, type, and size of defects existing in the selected area).

[0036] In some embodiments, referring to Figure 3 , the preprocessing is image feature enhancement. The image preprocessing process is essentially image enhancement, highlighting the unobvious features and reducing the interference of environmental factors. In this embodiment, a neural network with functions including histogram equalization, contrast stretching, gamma transformation, Laplacian operator, high-pass filtering, sharpening, and low-pass filtering is used for image enhancement to facilitate subsequent manual annotation and object detection. Specifically: Feature extraction is performed on the acquired images to obtain image features; The image features are respectively processed through histogram equalization, contrast stretching, gamma transformation, Laplacian operator, high-pass filtering, sharpening, and low-pass filtering to obtain corresponding processed features; The processed features are respectively subjected to weighted fusion and stitching fusion to obtain weighted fusion features and stitching fusion features; The weighted fusion features and stitching fusion features are superimposed and then image decoding is performed to obtain enhanced images; the images are effectively enhanced without losing detail information, especially defect information.

[0037] In some embodiments, under the illumination of a light source, images of magnetic particle testing of hydroelectric equipment are obtained by a camera photographing magnetic traces.

[0038] In some embodiments, the application is in the field of magnetic particle testing for hydropower equipment, which is very different from the application scenarios of general AI image recognition technologies. In particular, the dataset available for model training is relatively small. For this reason, this case proposes a supervised learning method. Through a network algorithm responsible for random sampling, the detection results are regenerated into training data to solve the problem of insufficient training data. The specific implementation is as follows Figure 5 . Applied to Figure 2 the random sampling link in, and the result is to generate new data, which is sent to the manual annotation link for training data generation. Using the training results as data reduces the dependence on the amount of data.

[0039] This application also discloses a magnetic particle testing system for hydropower equipment based on AI image recognition, which is characterized by including: An image acquisition and preprocessing unit, which is used to acquire and preprocess the images of magnetic particle testing for hydropower equipment to obtain enhanced images; A network training unit, which is used to input the enhanced images into a preset neural network model for network training to obtain network parameters; A defect detection unit, which is used to perform defect detection on the enhanced images according to the network parameters to obtain the areas in the images where defects may exist, the probability of defects existing in these areas, the types and size information of the defects.

[0040] In some embodiments, referring to Figure 4 , in the defect detection unit, performing defect detection on the enhanced images according to the network parameters specifically includes: The convolutional neural network extracts features from the enhanced images to obtain enhanced image features; Based on the network parameters, judgments are made on the image features to obtain the areas where defects may exist; For the areas where defects may exist, based on the network parameters, judge the probability of defects existing in these areas, the types and size information of the defects.

[0041] To solve the sensitivity to defects, a convolutional neural network responsible for extracting defect features is specifically designed. Compared with general network models, the innovation point of this embodiment is to add a judgment layer to judge whether the extracted features are defects, obtain the areas where defects may exist, and improve the sensitivity to defects.

[0042] In some embodiments, the image acquisition and preprocessing unit includes a set of hardware systems for image acquisition and its control method. The hardware system is mainly composed of an industrial camera and an auxiliary lighting system. The industrial camera can adjust the aperture, shutter, focal length, and white balance to achieve high-quality imaging. The auxiliary lighting system can adjust the ambient brightness, light irradiation direction, color temperature, etc. The implementation principle is as follows Figure 6By controlling the camera and light source parameters, the quality of image input is improved.

[0043] This application also discloses an electronic 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 steps of the magnetic particle detection method for hydropower equipment based on AI image recognition described in any one of the above are implemented.

[0044] This application also discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the magnetic particle detection method for hydropower equipment based on AI image recognition described in any one of the above are implemented.

[0045] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0046] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one or more of the processes Figure 1 or multiple processes and / or blocks

[0047] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the specified functions in Figure 1 one or more of the processes Figure 1 or multiple processes and / or blocks

[0048] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to generate a computer-implemented process, thereby the instructions executed on the computer or other programmable apparatus provide steps for realizing the functions specified in one process or multiple processes and / or one block or multiple blocks. Figure 1 one process or multiple processes and / or blocks Figure 1 steps for realizing the functions specified in one block or multiple blocks.

[0049] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific embodiments of the present invention. Any modification or equivalent replacement without departing from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A magnetic particle inspection method for hydropower equipment based on AI image recognition, characterized in that, Including: Obtain the images of magnetic particle inspection of hydroelectric equipment and perform preprocessing to obtain enhanced images; Input the enhanced images into a preset neural network model for network training to obtain network parameters, and randomly sample some network parameters for network iterative optimization training; According to the network parameters, perform defect detection on the enhanced images to obtain the areas in the images where defects may exist, the probability of defects existing in these areas, the types and size information of the defects.

2. The magnetic particle inspection method for hydropower equipment based on AI image recognition according to claim 1, wherein, The step of inputting the enhanced images into a preset neural network model for network training specifically includes: S201: Manually annotate the enhanced images to mark the areas with defects and the defect types; S202: Extract features from the annotated enhanced images to obtain the areas with defects and the defect types; S203: Classify the areas with defects according to the defect types to obtain classified defect areas and their corresponding defect types; S204: Perform object detection on the classified defect areas to obtain the network parameters of these classified defect areas; S205: Randomly sample the network parameters in S204 and feedback them to S201 for iterative optimization training.

3. The magnetic particle inspection method for hydropower equipment based on AI image recognition according to claim 2, wherein The network parameters include: the areas with defects, the types and size information of the defects.

4. A magnetic particle inspection method for hydropower equipment based on AI image recognition according to claim 1, characterized in that, The step of performing defect detection on the enhanced images according to the network parameters specifically includes: S301: The convolutional neural network extracts features from the enhanced images to obtain enhanced image features; S302: Based on the network parameters, make judgments on the image features to obtain the areas where defects may exist; S303: For the areas where defects may exist, based on the network parameters, judge the probability of defects existing in these areas, the types and size information of the defects.

5. A magnetic particle inspection method for hydroelectric equipment based on AI image recognition according to claim 1, characterized in that, The preprocessing is image feature enhancement, specifically: Extract features from the obtained images to obtain image features; The image features are respectively processed by histogram equalization, contrast stretching, gamma transformation, Laplacian operator, high-pass filtering, sharpening and low-pass filtering to obtain the corresponding processed features; The processed features are respectively subjected to weighted fusion and stitching fusion to obtain weighted fusion features and stitching fusion features; The weighted fusion features and the stitching fusion features are superimposed and then image decoding is performed to obtain enhanced images.

6. The magnetic particle inspection method for hydropower equipment based on AI image recognition according to claim 1, wherein, Under the illumination of a light source, use a camera to capture magnetic traces to obtain the images of magnetic particle inspection of hydroelectric equipment.

7. A magnetic particle inspection system for hydropower equipment based on AI image recognition, characterized in that, Including: An image acquisition and preprocessing unit for obtaining the images of magnetic particle inspection of hydroelectric equipment and performing preprocessing to obtain enhanced images; A network training unit for inputting the enhanced images into a preset neural network model for network training to obtain network parameters; A defect detection unit for performing defect detection on the enhanced images according to the network parameters to obtain the areas in the images where defects may exist, the probability of defects existing in these areas, the types and size information of the defects.

8. An AI image recognition-based magnetic particle inspection system for hydroelectric equipment according to claim 7, characterized in that, In the defect detection unit, the defect detection of the enhanced images according to the network parameters is specifically: The convolutional neural network extracts features from the enhanced images to obtain enhanced image features; Based on the network parameters, make judgments on the image features to obtain the areas where defects may exist; For a region that may have defects, based on network parameters, determine the probability of defects, the type and size information of the defects within the region.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the magnetic particle inspection method for hydropower equipment based on AI image recognition according to any one of claims 1-6 are implemented.

10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of the magnetic particle inspection method for hydropower equipment based on AI image recognition according to any one of claims 1-6 are implemented.

Images