Power equipment label text identification method and device of power system and electronic equipment

Abstract

The invention discloses a power equipment label text identification method and device of a power system and electronic equipment. The method comprises the steps that label image data of power equipment in a power system are acquired, and the label image data are used for representing number information and identification information corresponding to the power equipment; detecting the label image data through an attention mechanism by using a label text detection model to obtain a text data area, the text data area being used for representing the area of the number information and the identification information in the label image data; and identifying the image data of the text data area by using a label text identification model to obtain label text data which comprises number information and identification information. According to the method and the device, the technical problem of relatively low identification accuracy of the tag text of the power equipment in related technologies is solved.

Description

Technical Field

[0001] The present invention relates to the field of power systems, and in particular to a method, a device and an electronic device for identifying text on power equipment labels in a power system. Background Art

[0002] In the operation, maintenance and management of power systems, secondary cabinets serve as a bridge between primary equipment and control equipment, and the correctness of their terminal blocks is directly related to the safe and stable operation of the power system. The traditional manual detection method, that is, comparing the correspondence between the on-site terminal blocks and the design drawings by naked eyes, has problems such as long detection time, low accuracy and poor safety.

[0003] In the related art, optical character recognition technology is used to automatically recognize text on terminal block labels, which improves the efficiency of information processing. However, since the placement environment of the secondary panel cabinet is dim and complex, and the characters on the terminal block labels of the secondary panel cabinet are small and dense, interference factors exist in the text recognition process, which in turn leads to a low accuracy rate in text recognition of power equipment labels in the related art.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method, a device and an electronic device for recognizing text on a label of electric equipment in an electric power system, so as to at least solve the technical problem of low accuracy in recognizing text on a label of electric equipment in the related art.

[0006] According to one aspect of an embodiment of the present invention, a method for identifying label text of electric power equipment in an electric power system is provided, comprising: obtaining label image data of electric power equipment in the electric power system, wherein the label image data is used to represent numbering information and identification information corresponding to the electric power equipment; using a label text detection model to detect the label image data through an attention mechanism to obtain a text data area, wherein the text data area is used to represent an area of ​​numbering information and identification information in the label image data; using a label text recognition model to recognize image data of the text data area to obtain label text data, wherein the label text data contains numbering information and identification information.

[0007] Furthermore, the label text detection model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, an attention processing layer and a feature detection layer; the label text detection model is used to detect the label image data through the attention mechanism to obtain the text data area, including: using the first feature extraction layer to extract features of the label image data to obtain the first label image feature; using the second feature extraction layer to extract features of the first label image feature to obtain the second label image feature, and the feature scale of the second label image is smaller than the feature scale of the first label image; using the feature fusion layer to fuse the first label image feature with the second label image feature to obtain a fused feature, and the fused feature is used to represent the feature obtained by splicing label image features of different scales; using the attention processing layer to perform attention processing on the fused feature to obtain the attention feature; using the attention feature of the feature detection layer to detect and obtain the text data area.

[0008] Furthermore, the method also includes: using a preprocessing model to preprocess the text data area to obtain a preprocessed text data area; using a label text recognition model to recognize the image data of the text data area to obtain label text data, including: using a label text recognition model to recognize the image data of the preprocessed text data area to obtain label text data.

[0009] Furthermore, the preprocessing model includes a color space conversion layer, a contour detection and correction layer, and a target area cropping layer. The preprocessing model is used to preprocess the text data area to obtain a preprocessed text data area, including: using the color space conversion layer to convert the input text data area from the original color space to the color brightness space; using the contour detection and correction layer to detect the boundary contour of the text data area to obtain the boundary contour coordinates, wherein the boundary contour coordinates are used to represent the specific coordinate position information of the boundary contour; using the target area cropping layer to crop the text data area based on the boundary contour coordinates to obtain the preprocessed text data area.

[0010] Furthermore, the boundary contour of the text data area is detected using a contour detection correction layer to obtain boundary contour coordinates, including: detecting the boundary contour of the text data area using a contour detection correction layer; calculating a deflection angle based on the boundary contour, wherein the deflection angle is used to represent the angle of the boundary contour relative to a horizontal plane; correcting the text data area based on the deflection angle to ensure that the text data area and the boundary contour are located on the same horizontal plane; and obtaining boundary contour coordinates based on the corrected text data area.

[0011] Furthermore, the label text recognition model is used to recognize the image data of the preprocessed text data area to obtain the label text data, including: optimizing the parameters of the label text recognition model based on the parameter optimization method to find the target parameters; applying the target parameters to the label text recognition model to obtain a new label text recognition model; and using the new label text recognition model to recognize the image data of the preprocessed text data area to obtain the label text data.

[0012] According to another aspect of an embodiment of the present invention, a device for recognizing label text of electric power equipment in an electric power system is also provided, including: an acquisition module, used to acquire label image data of electric power equipment in the electric power system, wherein the label image data is used to represent numbering information and identification information corresponding to the electric power equipment; a detection module, used to detect the label image data using an attention mechanism using a label text detection model to obtain a text data area, wherein the text data area is used to represent an area of ​​numbering information and identification information in the label image data; and a recognition module, used to recognize image data of the text data area using a label text recognition model to obtain label text data, wherein the label text data contains numbering information and identification information.

[0013] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0014] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0015] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0017] In an embodiment of the present invention, the label image data of the power equipment in the power system is first obtained, and then the label image data is detected by the label text detection model through the attention mechanism to obtain the text data area, and finally the image data of the text data area is recognized by the label text recognition model to obtain the label text data. It is easy to notice that the present application obtains the numbering information and identification information corresponding to the power equipment by obtaining the label image data, and then detects the label image data by combining the label text detection model with the attention mechanism to obtain the text data area containing the key features, and finally the label text recognition model is used to recognize the text data area, so as to obtain the numbering information and identification information of the power equipment, thereby realizing the accurate recognition of the label image data, and then solving the technical problem of low accuracy of the label text recognition of the power equipment in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0019] Figure 1 is a flow chart of a method for identifying text on a label of power equipment in a power system according to an embodiment of the present invention;

[0020] Figure 2 4 is a schematic diagram of a device for identifying text on a label of electric equipment in an electric power system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] According to an embodiment of the present invention, an embodiment of a method for identifying text of labels of power equipment in an electric power system is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] Figure 1 is a flow chart of a method for identifying text of a label of an electric power device in an electric power system according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0025] Step S102: acquiring label image data of the power equipment in the power system, wherein the label image data is used to represent the number information and identification information corresponding to the power equipment.

[0026] The power equipment in the above steps is used to describe various equipment used for power generation, transmission, transformation, distribution and consumption in the power system. The types of power equipment may include but are not limited to secondary panels, relays, control cabinets, monitoring equipment, distribution boards, switch cabinets, etc. The specific type of power equipment needs to be determined based on the actual identification task and is not limited here. The functions of power equipment may include but are not limited to remote monitoring, control and protection of the power system. The specific functions of power equipment need to be determined based on the type of power equipment and are not limited here.

[0027] The label image data in the above steps is used to describe the image of the identification label on the power equipment obtained by image acquisition means, including the number information and identification information of the power equipment. The above identification label refers to various labels usually affixed to the power equipment in order to identify the type, number, function and connection information of the power equipment, such as the terminal number label on the terminal block, the identification label on the cable, etc. The above number information is the unique identification of the equipment or component, which can be used to quickly locate and identify the equipment in the power system. The above identification information includes the type, function and safety warning of the equipment, etc., which can be used to clarify the purpose and operation guide of the equipment. The role of the label image data can include but is not limited to equipment management, wiring correction, fault diagnosis and maintenance, etc.

[0028] The methods for obtaining the data in the above steps may include but are not limited to the following methods:

[0029] The first method: Use fixed cameras to obtain tag image data. Fixed cameras are set up in the power equipment installation area to continuously monitor equipment tags.

[0030] The second method: obtain label image data through drone or robot inspection. Use drones or robots equipped with cameras and image acquisition systems to automatically inspect power equipment and collect equipment label images.

[0031] The third method: Obtain label image data through infrared or thermal imaging technology. In some cases, such as when the device is at high temperature or at night, infrared or thermal imaging cameras can be used to collect device label images under special conditions.

[0032] In an optional embodiment, the equipment label image data can be collected comprehensively and flexibly by deploying a fixed camera monitoring system, using drones or robots for inspection, or using handheld devices of staff to take photos. The acquisition of equipment label image data makes the label information of power equipment digital and structured, which is convenient for subsequent identification processing and data analysis, thereby ensuring the safe operation of the power system, improving the speed of fault diagnosis, optimizing equipment maintenance plans, and laying a solid foundation for the automated management of the power system.

[0033] Step S104, using the label text detection model to detect the label image data through the attention mechanism to obtain the text data area, wherein the text data area is used to represent the area of ​​the numbering information and the identification information in the label image data.

[0034] The label text detection model in the above steps is used to describe a computer vision technology that detects and locates text from an image through a deep learning algorithm. The types of label text detection models may include but are not limited to target detection systems (You Only Look Once v8, referred to as YOLOv8), improved version of YOLOv8, Single Shot MultiBox Detector (SSD), fast regional convolutional neural network (Region-based Convolutional Neural Network, referred to as Faster R-CNN), etc. The specific label text detection model needs to be determined according to the model design, which is not limited here. The label text detection model can be used to accurately detect the numbering information and identification information on the power equipment.

[0035] The attention mechanism in the above steps is used to describe a technology that enables the model to focus on the key parts of the input image. The types of attention mechanisms may include but are not limited to Multi-Head SelfAttention (MHSA), Squeeze-and-Excitation Network (SENet), and Convolutional Block Attention Module (CBAM). The specific type of attention mechanism should be determined according to the model design requirements and is not limited here. The role of the attention mechanism may include but is not limited to helping the model identify which parts of the image features are crucial for text detection, thereby improving the efficiency and effectiveness of detection.

[0036] The text data area in the above steps is used to describe the specific location range containing numbering information and identification information detected by the label text detection model in the label image data, the type of power equipment and the design of the label in the specific text data area. Each type of text data area may differ in size, shape and text content. The role of the text data area may include but is not limited to clarifying the specific location of the text information in the image to facilitate subsequent recognition processing.

[0037] In an optional embodiment, the label text detection model is combined with the attention mechanism to detect the label image data in the power system, so as to accurately lock the position of the number information and identification information in the image. Specifically, attention mechanisms such as SENet, CBAM or MHSA can be integrated into efficient detection models such as YOLOv8. The integrated label text detection model can focus on the key parts of the image, that is, the area containing important text, so that the model can quickly identify the label position of the power equipment, and then accurately extract the number information and identification information on the label, reduce false detection and missed detection, and improve the accuracy and reliability of detection.

[0038] Step S106, using the label text recognition model to recognize the image data of the text data area to obtain label text data, wherein the label text data includes number information and identification information.

[0039] The label text recognition model in the above steps is a deep learning model used to describe the recognition of text information from images. The types of label text recognition models may include but are not limited to the PaddleOptical Character Recognition model (PaddleOCR for short), the EasyOptical Character Recognition model (EasyOCR for short), etc. The specific label text recognition model needs to be determined according to actual needs and is not limited here. The functions of the label text recognition model may include but are not limited to automatic text extraction, intelligent data analysis, etc.

[0040] The label text data in the above steps is used to describe the text information identified and extracted from the image data, including the number information and identification information of the power equipment. The label text data can be used to help power workers quickly and accurately locate and identify power equipment.

[0041] In an optional embodiment, the image data of the text data area is recognized by the label text recognition model to obtain the label text data, which not only improves the accuracy and efficiency of recognition, reduces the possibility of human misreading and omission, but also provides reliable data support for the digital management of the power system. This step is a key link between computer vision and intelligent power operation and maintenance, realizes the conversion from image to structured text data, provides direct text information for subsequent data analysis, equipment tracking and fault diagnosis, and further improves the automation level of power system operation and maintenance.

[0042] In an embodiment of the present invention, the label image data of the power equipment in the power system is first obtained, and then the label image data is detected by the label text detection model through the attention mechanism to obtain the text data area, and finally the image data of the text data area is recognized by the label text recognition model to obtain the label text data. It is easy to notice that the present application obtains the numbering information and identification information corresponding to the power equipment by obtaining the label image data, and then detects the label image data by combining the label text detection model with the attention mechanism to obtain the text data area containing the key features, and finally the label text recognition model is used to recognize the text data area, so as to obtain the numbering information and identification information of the power equipment, thereby realizing the accurate recognition of the label image data, and then solving the technical problem of low accuracy of the label text recognition of the power equipment in the related technology.

[0043] Optionally, the label text detection model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, an attention processing layer and a feature detection layer; the label text detection model is used to detect the label image data through the attention mechanism to obtain the text data area, including: using the first feature extraction layer to extract features of the label image data to obtain the first label image feature; using the second feature extraction layer to extract features of the first label image feature to obtain the second label image feature, and the scale of the second label image feature is smaller than the scale of the first label image feature; using the feature fusion layer to fuse the first label image feature with the second label image feature to obtain a fused feature, and the fused feature is used to represent the feature obtained by splicing label image features of different scales; using the attention processing layer to perform attention processing on the fused feature to obtain the attention feature; using the attention feature of the feature detection layer to detect and obtain the text data area.

[0044] The first feature extraction layer mentioned above is used to describe the first convolution layer in the label text detection model, which is used to perform preliminary feature extraction on the input label image data to obtain the first label image feature.

[0045] The first label image features mentioned above are used to describe image features extracted from the label image data after being processed by the first feature extraction layer. These features usually have a higher resolution and can retain detailed information in the image.

[0046] The second feature extraction layer mentioned above is used to describe another feature extraction layer in the label text detection model, which is usually located after the first feature extraction layer and is used to further extract more advanced features from the image after preliminary feature extraction.

[0047] The second label image feature mentioned above is used to describe the feature vector or feature map extracted from the image after being processed by the second feature extraction layer. Compared with the first label image feature, the scale of the second label image feature is usually smaller, but the semantic information contained is richer.

[0048] The above-mentioned feature fusion layer is used to describe the network layer that combines different scales into a more complete feature representation by splicing, weighted summation, etc. The types of feature fusion layers may include but are not limited to cross-stage feature fusion network (CrossConcatenate Fusion, referred to as C2F), feature pyramid network (Feature Pyramid Network, referred to as FPN), etc. The specific feature fusion layer can be determined according to the model design requirements and is not limited here. The feature fusion layer can be used to integrate the advantages of features of different scales, so that the model can consider details and macro structures at the same time when processing complex images, thereby improving the accuracy of detection.

[0049] The above-mentioned attention processing layer is used to describe the network layer used to emphasize the important parts of the input features in the label text detection model. The types of attention processing layers may include but are not limited to multi-head self-attention mechanism (Multi-Head Self Attention, referred to as MHSA), squeeze-excitation network (Squeeze-and-Excitation Network, referred to as SENet), convolutional block attention module (Convolutional Block Attention Module, referred to as CBAM), etc. The specific attention processing layer can be determined according to the model design requirements and is not limited here. The role of the attention processing layer may include but is not limited to helping the model to more effectively process the target information in the image, improve the accuracy of detection and recognition, and reduce unnecessary computing resource consumption.

[0050] The above-mentioned attention features are used to describe the feature representation obtained after the attention mechanism is applied to the fusion features in the attention processing layer. The attention features are given higher weights in the decision-making process of the model, reflecting more critical text information in the image. The role of the attention features can include but is not limited to helping the model more accurately locate and identify text areas by emphasizing important parts of the image, thereby improving the overall detection performance.

[0051] The above-mentioned feature detection layer is used to describe the network layer in the model that is used to ultimately detect the attention features in the image. The feature detection layer locates and selects the text area through a specific detection mechanism. The above-mentioned detection mechanism may include but is not limited to an anchor mechanism, a sliding window, etc. The feature detection layer can be used to select the text data area based on the attention features.

[0052] In an optional embodiment, a label text detection model including an attention mechanism is designed to detect label image data. The label text detection model includes a first feature extraction layer, a second feature extraction layer, a feature fusion layer, an attention processing layer and a feature detection layer. First, the first feature extraction layer processes the original label image data and extracts a preliminary feature representation, namely, a first label image feature. Subsequently, the second feature extraction layer further extracts higher-level features from the first label image feature to form a second label image feature, which has a smaller scale and focuses more on the structure and semantic information of the image. Then, through the feature fusion layer, the first label image feature is integrated with the second label image feature to obtain a fused feature. This fusion of features of different scales enables the model to consider both the details and macroscopic structure of the image, thereby improving the comprehensiveness and accuracy of the detection. Then, the attention processing layer processes the fused feature, highlights the key information of the text data area through the attention mechanism, and obtains the attention feature. Finally, the feature detection layer uses the attention feature for detection to accurately locate and select the text data area in the image. The above process, by combining multi-scale feature extraction and attention mechanism, realizes the rapid and accurate recognition of power equipment labels such as secondary panel cabinet terminal blocks, greatly improving the efficiency and safety of power system management.

[0053] Optionally, the method also includes: using a preprocessing model to preprocess the text data area to obtain a preprocessed text data area; using a label text recognition model to recognize the image data of the text data area to obtain label text data, including: using a label text recognition model to recognize the image data of the preprocessed text data area to obtain label text data.

[0054] The above-mentioned preprocessing model is used to describe a model that preprocesses the text data area between the label text detection model and the label text recognition model, wherein the types of preprocessing models may include but are not limited to image filtering layer, image enhancement layer, image correction layer, geometric transformation layer and character segmentation layer, etc. The specific preprocessing model needs to be determined according to actual needs and is not limited here. The role of the preprocessing model may include but is not limited to removing background noise, standardizing image data, and improving the accuracy and efficiency of subsequent text recognition.

[0055] In an optional embodiment, due to the complexity of the power site environment, such as light changes, background interference, angle deviation and other factors, the image quality can be improved through the preprocessing model to ensure that the text data area meets the recognition requirements. For example, by using image enhancement and perspective correction technology, the recognition accuracy under different lighting conditions and angle deviations can be improved; through size standardization and grayscale processing, it can be ensured that the model is not affected by the input image size and color space during processing. After obtaining the preprocessed text data area, the label text recognition model is used to identify the image data of the preprocessed text data area to obtain the label text data. The above steps combine the preprocessing model with the label text recognition model, reduce the model's misrecognition and missed recognition, and achieve efficient and accurate recognition of the label text of power equipment.

[0056] Optionally, the preprocessing model includes a color space conversion layer, a contour detection and correction layer, and a target area cropping layer. The preprocessing model is used to preprocess the text data area to obtain a preprocessed text data area, including: using the color space conversion layer to convert the input text data area from the original color space to the color brightness space; using the contour detection and correction layer to detect the boundary contour of the text data area to obtain the boundary contour coordinates, wherein the boundary contour coordinates are used to represent the specific coordinate position information of the boundary contour; using the target area cropping layer to crop the text data area based on the boundary contour coordinates to obtain the preprocessed text data area.

[0057] The above-mentioned color space conversion layer is used to describe the network layer that converts the text data area from one color space to another color space. The color space conversion layer may include but is not limited to converting the red green blue (Red Green Blue, referred to as RGB) color space to hue, saturation, brightness (Hue Saturation Value, referred to as HSV) and other color spaces. The specific color space conversion of the color space conversion layer needs to be determined according to actual needs and is not limited here. The color space conversion can be used to highlight certain color features, thereby facilitating subsequent image processing. For example, in the RGB space, color information is intertwined, while in the HSV space, color information is decomposed into three independent dimensions of hue, saturation and brightness, which makes it easier to identify and segment colors.

[0058] The above-mentioned original color space is a natural representation form used to describe the text data area, and the original color space may be an RGB space.

[0059] The above-mentioned color brightness space is used to describe a color space used to emphasize image brightness information. The color brightness space may include but is not limited to HSV space and other color spaces that emphasize image brightness information. The specific color brightness space can be determined according to actual needs and is not limited here. The role of the color brightness space may include but is not limited to enhancing the contrast of the image, improving the visibility of the text area, and making subsequent contour detection and correction more accurate.

[0060] The above-mentioned contour detection and correction layer is used to describe the network layer used to detect and correct the boundary contour of the text data area. The function of the contour detection and correction layer can include but is not limited to accurately locating the text area. Even if the text is tilted or deformed, it can be corrected through coordinate correction to improve the accuracy of text recognition.

[0061] The above-mentioned boundary contour is used to describe the contour line formed around the edge of the text data area. The shape of the boundary contour may include but is not limited to a rectangle, a square, etc. The specific boundary contour needs to be determined according to the text data area and is not limited here.

[0062] The above-mentioned boundary contour coordinates are used to describe a set of position information representing each point on the boundary contour in the image coordinate system. The boundary contour coordinates may include but are not limited to the coordinates of the four corners of the rectangular boundary contour, the two coordinates of the opposite corners of the rectangular boundary contour, etc. The specific boundary contour coordinates need to be determined according to actual needs and are not limited here. The boundary contour coordinates can be used to guide the target area cropping layer to accurately crop the text data area to ensure that the cropped image contains only text information without redundant background.

[0063] The above-mentioned target area cropping layer is used to describe the network layer that accurately crops the text data area from the label image according to the boundary contour coordinates. The target area cropping layer can use the functions in the open source computer vision library OpenCV to achieve target area cropping. The role of the target area cropping layer can include but is not limited to removing irrelevant background, reducing unnecessary calculations, and ensuring that the text information is complete and clear.

[0064] In an optional embodiment, the designed preprocessing model includes a color space conversion layer, a contour detection correction layer, and a target area cropping layer. During the preprocessing process, first, the color space conversion layer is used to change the color representation of the image to highlight the text features; then, the contour detection correction layer is used to detect and correct the shape of the text area to ensure that the text is facing forward; finally, the target area cropping layer is used to accurately crop the text data area according to the corrected contour to provide high-quality input for text recognition. The above preprocessing process greatly improves the accuracy and efficiency of text recognition, especially in complex scenarios such as power equipment label recognition. Preprocessing can effectively reduce the interference of irrelevant factors, emphasize the required features, and thus improve recognition efficiency.

[0065] Optionally, a contour detection and correction layer is used to detect a boundary contour of a text data region to obtain boundary contour coordinates, including: using a contour detection and correction layer to detect a boundary contour of a text data region; calculating a deflection angle based on the boundary contour, wherein the deflection angle is used to represent an angle of the boundary contour relative to a horizontal plane; correcting the text data region based on the deflection angle to ensure that the text data region and the boundary contour are located on the same horizontal plane; and obtaining boundary contour coordinates based on the corrected text data region.

[0066] The above-mentioned deflection angle is used to describe the rotation angle of the boundary contour of the text data area relative to the horizontal reference line of the image. The role of calculating the deflection angle may include but is not limited to helping understand the actual orientation of the text in the image. For subsequent image correction, this is essential information. Accurate deflection angle measurement can ensure the correctness and readability of the text after correction. The method for calculating the deflection angle may include but is not limited to finding the minimum bounding rectangle or ellipse of the boundary contour through the minimum area bounding rectangle function minAreaRect() or the fitting ellipse function fitEllipse() in OpenCV, and extracting the rotation angle therefrom. The specific method for calculating the deflection angle needs to be determined based on the model design and is not limited here.

[0067] The above correction is used to describe the process of adjusting the position and direction of the image or text area based on the calculated deflection angle so that it is aligned with the horizontal plane. The correction method may include but is not limited to the implementation of the rotation matrix function getRotationMatrix2D() and the affine transformation function warpAffine() in OpenCV. First, use getRotationMatrix2D() to obtain the rotation matrix, and then use warpAffine() to perform an affine transformation on the original image according to the rotation matrix to achieve the correction of the text area. The specific correction method needs to be determined according to the model design and is not limited here.

[0068] In an optional embodiment, the text data area is detected and corrected through a contour detection and correction layer. First, the boundary contour of the text data area is identified through contour detection technology; then, based on the detected boundary contour, the deflection angle of the area relative to the horizontal plane is calculated; then, the text data area is corrected using the calculated deflection angle to ensure that it is adjusted to the same horizontal plane as the boundary contour, thereby improving the alignment and readability of the text, and finally, the boundary contour coordinates are obtained based on the corrected text area to facilitate the subsequent cropping of the text data area based on the boundary contour coordinates. Through the above process, even images in complex environments, such as dense and possibly tilted text on the terminal block label of the secondary panel cabinet, can be efficiently and accurately identified, greatly reducing human errors, reducing potential risks in the operation and maintenance of power equipment, and improving the automation level and operation and maintenance efficiency of the power system.

[0069] Optionally, the image data of the preprocessed text data area is recognized using a label text recognition model to obtain label text data, including: optimizing parameters of the label text recognition model based on a parameter optimization method to find target parameters; applying the target parameters to the label text recognition model to obtain a new label text recognition model; and recognizing the image data of the preprocessed text data area using the new label text recognition model to obtain label text data.

[0070] The above-mentioned parameter optimization is used to describe the process of finding the target parameters by continuously adjusting the parameters of the model. The target parameters found enable the model to have good recognition performance. The types of parameter optimization may include but are not limited to gradient descent method, genetic algorithm and particle swarm optimization algorithm. The specific parameter optimization type needs to be determined according to the model design requirements and is not limited here. The role of parameter optimization may include but is not limited to helping the model find the target parameters and improving the recognition rate of the model.

[0071] The above-mentioned target parameters are used to describe the parameter set found through the parameter optimization process that can make the model have good performance. The types of target parameters may include but are not limited to weights, biases, and learning rates. The specific target parameters need to be determined according to the specific label text recognition model and are not limited here.

[0072] The above-mentioned new label text recognition model is used to describe a model obtained by adjusting parameters through parameter optimization based on the original model.

[0073] In an optional embodiment, the parameters are optimized and the target parameters are obtained by using algorithms such as the gradient descent method and the particle swarm optimization algorithm, and then the obtained target parameters are applied to the label text recognition model to construct a new label text recognition model. In the specific application of power equipment label recognition, the optimized model can more accurately process the pre-processed text data area, and maintain a high recognition accuracy rate even in the case of poor image quality or incorrect text arrangement angle. The above parameter optimization method not only improves the recognition performance, but also enables the system to adapt to a wider range of scenarios and needs, providing solid technical support for the intelligent operation and maintenance of power systems.

[0074] According to another aspect of an embodiment of the present invention, a device for identifying text of electric equipment labels in an electric power system is also provided. The device can execute the method for identifying text of electric equipment labels in an electric power system provided in the above embodiment. The specific implementation method and preferred application scenario are the same as those in the above embodiment and will not be repeated here.

[0075] Figure 2 is a schematic diagram of a device for identifying text on a label of an electric power device in an electric power system according to an embodiment of the present invention. Figure 2 As shown, the device includes: an acquisition module 202, used to acquire label image data of power equipment in the power system, wherein the label image data is used to represent the numbering information and identification information corresponding to the power equipment; a detection module 204, used to detect the label image data using an attention mechanism using a label text detection model to obtain a text data area, wherein the text data area is used to represent the area of ​​the numbering information and the identification information in the label image data; and a recognition module 206, used to recognize the image data of the text data area using a label text recognition model to obtain label text data, wherein the label text data contains the numbering information and the identification information.

[0076] Optionally, the label text detection model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, an attention processing layer and a feature detection layer; the detection module includes: a first feature extraction unit, used to use the first feature extraction layer to perform feature extraction on the label image data to obtain the first label image feature; a second feature extraction unit, used to use the second feature extraction layer to perform feature extraction on the first label image feature to obtain the second label image feature, and the feature scale of the second label image is smaller than the feature scale of the first label image; a feature fusion unit, used to use the feature fusion layer to perform feature fusion on the first label image feature and the second label image feature to obtain a fused feature, and the fused feature is used to represent the feature obtained by splicing label image features of different scales; an attention processing unit, used to use the attention processing layer to perform attention processing on the fused feature to obtain an attention feature; a feature detection unit, used to use the attention feature of the feature detection layer to perform detection to obtain a text data area.

[0077] Optionally, the device also includes: a preprocessing module, which is used to preprocess the text data area using a preprocessing model to obtain a preprocessed text data area; a recognition module, which is used to recognize the image data of the text data area using a label text recognition model to obtain label text data, including: using the label text recognition model to recognize the image data of the preprocessed text data area to obtain label text data.

[0078] Optionally, the preprocessing model includes a color space conversion layer, a contour detection and correction layer, and a target area cropping layer. The preprocessing module includes: a conversion unit, which is used to use the color space conversion layer to convert the input text data area from the original color space to the color brightness space; a detection unit, which is used to use the contour detection and correction layer to detect the boundary contour of the text data area and obtain the boundary contour coordinates, wherein the boundary contour coordinates are used to represent the specific coordinate position information of the boundary contour; a cropping unit, which is used to use the target area cropping layer to crop the text data area based on the boundary contour coordinates to obtain the preprocessed text data area.

[0079] Optionally, the detection unit includes: using a contour detection correction layer to detect the boundary contour of the text data area; calculating a deflection angle based on the boundary contour, wherein the deflection angle is used to represent the angle of the boundary contour relative to a horizontal plane; correcting the text data area based on the deflection angle to ensure that the text data area and the boundary contour are located on the same horizontal plane; and obtaining the boundary contour coordinates based on the corrected text data area.

[0080] Optionally, the recognition module includes: a parameter optimization unit, which is used to optimize the parameters of the label text recognition model based on the parameter optimization method to find the target parameters; an application unit, which is used to apply the target parameters to the label text recognition model to obtain a new label text recognition model; and a recognition unit, which is used to use the new label text recognition model to recognize the image data of the preprocessed text data area to obtain the label text data.

[0081] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0082] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0083] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0084] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0085] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0086] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0087] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0088] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0089] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0090] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for identifying text on power equipment labels in a power system, characterized in that: include: Acquire label image data of an electric power device in an electric power system, wherein the label image data is used to represent number information and identification information corresponding to the electric power device; Using a label text detection model to detect the label image data through an attention mechanism to obtain a text data region, wherein the text data region is used to represent a region of the number information and the identification information in the label image data; The image data of the text data area is recognized by using a label text recognition model to obtain label text data, wherein the label text data includes the number information and the identification information.

2. The method for identifying text on power equipment labels in a power system according to claim 1, characterized in that: The label text detection model includes: a first feature extraction layer, a second feature extraction layer, a feature fusion layer, an attention processing layer and a feature detection layer; the label text detection model is used to detect the label image data through an attention mechanism to obtain a text data area, including: Using the first feature extraction layer to extract features from the label image data, to obtain first label image features; Using the second feature extraction layer to extract features of the first label image, to obtain second label image features, wherein the feature scale of the second label image is smaller than the feature scale of the first label image; Using the feature fusion layer to fuse the first label image feature with the second label image feature to obtain a fused feature, where the fused feature is used to represent a feature obtained by splicing the label image features of different scales; Using the attention processing layer to perform attention processing on the fused features to obtain attention features; The attention feature of the feature detection layer is used to perform detection to obtain the text data area.

3. The method for identifying text on power equipment labels in a power system according to claim 1, characterized in that: The method further comprises: Preprocessing the text data region using a preprocessing model to obtain a preprocessed text data region; Using a label text recognition model to recognize the image data of the text data area to obtain label text data, including: The preprocessed image data of the text data area is recognized using a label text recognition model to obtain label text data.

4. The method for identifying text on power equipment labels in a power system according to claim 3, characterized in that: The preprocessing model includes a color space conversion layer, a contour detection correction layer and a target area clipping layer. The text data area is preprocessed using the preprocessing model to obtain a preprocessed text data area, including: Using the color space conversion layer to convert the input text data area from the original color space to the color brightness space; Detecting the boundary contour of the text data area using the contour detection and correction layer to obtain boundary contour coordinates, wherein the boundary contour coordinates are used to represent specific coordinate position information of the boundary contour; The text data region is clipped using a target region clipping layer based on the boundary contour coordinates to obtain the preprocessed text data region.

5. The method for identifying text on power equipment labels in a power system according to claim 4, characterized in that: Detecting the boundary contour of the text data area using the contour detection correction layer to obtain the boundary contour coordinates includes: Detecting the boundary contour of the text data area using the contour detection correction layer; Calculating a deflection angle based on the boundary contour, wherein the deflection angle is used to represent an angle of the boundary contour relative to a horizontal plane; Correcting the text data region based on the deflection angle to ensure that the text data region and the boundary contour are located on the same horizontal plane; The boundary contour coordinates are acquired based on the corrected text data area.

6. The method for identifying text on power equipment labels in a power system according to claim 3, characterized in that: Using a label text recognition model to recognize the preprocessed image data of the text data area to obtain label text data, including: Optimizing the parameters of the label text recognition model based on a parameter optimization method to find the target parameters; Applying the target parameters to the label text recognition model to obtain a new label text recognition model; The new label text recognition model is used to recognize the preprocessed image data of the text data area to obtain label text data.

7. A device for identifying text on power equipment labels in a power system, characterized in that: include: An acquisition module, used for acquiring label image data of an electric power device in an electric power system, wherein the label image data is used for indicating number information and identification information corresponding to the electric power device; A detection module, used to detect the label image data by using a label text detection model through an attention mechanism to obtain a text data region, wherein the text data region is used to represent a region of the number information and the identification information in the label image data; The recognition module is used to recognize the image data of the text data area by using a label text recognition model to obtain label text data, wherein the label text data includes the number information and the identification information.

8. An electronic device, characterized in that: include: A memory storing an executable program; A processor, configured to run the program, wherein the program executes the method according to any one of claims 1 to 6 when running.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored executable program, wherein when the executable program is executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The invention comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images