PDU terminal and wire label intelligent matching method based on image processing

Abstract

The invention discloses a PDU terminal and wire label intelligent matching method based on image processing, and the method comprises the following steps: S1, sending an original image into a region-of-interest detection network model for detection, and cutting the original image according to a detection result, thereby obtaining a region-of-interest image of the region where the PDU terminal and the wire label are located; s2, the interested image is sent to the terminal label and wire label area detection network model for detection, and according to a detection result, the PDU terminal and wire label images are cut and rotated to be corrected respectively to obtain a terminal label image and a wire label image; s3, sending the terminal label image and the wire rod label image which are rotated and corrected into an OCR identification network model for identification to obtain an identification result image; s4, processing the identification result image through an original image mapping algorithm to obtain a mapping result image; and S5, processing the terminal and the wire rod label in the mapping result image through a terminal and wire rod label matching algorithm to obtain a matching result.

Description

Technical Field

[0001] The present invention relates to an intelligent matching method of PDU terminals and wire labels based on image processing. Background Art

[0002] PDU (Power Distribution Unit) is a product designed to provide power distribution for cabinet-mounted electrical equipment. It has a variety of series specifications with different functions, installation methods and different socket combinations, and can provide suitable rack-mounted power distribution solutions for different power supply environments.

[0003] During the construction of large-scale cabinet groups, the intelligent power distribution system of each cabinet needs to detect whether the PDU terminal labels and wire labels match each other to identify any mis-insertion or mis-use.

[0004] At present, there is a lack of intelligent matching methods for PDU terminal and wire label identification. In industrial scenarios, manual label identification + manual matching is generally used to check whether the terminals and wires are matched correctly. This method is inefficient and also has the problem of high error rate due to human negligence.

[0005] Integrating visual recognition technology into the intelligent power distribution system can automatically identify and match PDU terminals and wire labels, reduce manual intervention, and improve work efficiency. However, visual recognition technology faces the following challenges in practical applications: 1. Different manufacturers may use different label formats and encoding methods, resulting in the system having to handle multiple label standards. The lack of a unified standard increases the complexity of system development and may affect the accuracy of identification.

[0006] 2. Due to the complexity of industrial scenarios, directly using the OCR recognition network to identify PDU terminals and wire labels will usually identify some background elements, which will greatly interfere with the correct identification of PDU terminals and wire labels.

[0007] 3. Due to the diversity of the placement angles of PDU terminals and wire labels, the direct use of OCR recognition network to identify PDU terminals and wire labels usually has a low recognition rate, which in turn affects subsequent intelligent matching. Summary of the invention

[0008] The purpose of the present invention is to provide a method for intelligently matching PDU terminal and wire labels based on image processing to achieve matching detection of PDU terminal labels and wire labels.

[0009] To this end, the present invention provides an intelligent matching method for PDU terminals and wire labels based on image processing, comprising the following steps: S1, obtaining an original image containing PDU terminals and wire labels, and then sending it into an area of ​​interest detection network model for detection, cropping the original image according to the detection result, and obtaining an image of interest in the area where the PDU terminals and wire labels are located; S2, sending the image of interest into a terminal label and wire label area detection network model for detection, cropping and rotating and straightening the PDU terminal and wire label images according to the detection results, to obtain a terminal label image and a wire label image; S3, sending the rotated and straightened terminal label image and wire label image into an OCR recognition network model for recognition to obtain a recognition result image; S4, processing the recognition result image by a mapping original image algorithm to obtain a mapping result image; S5, processing the terminals and wire labels in the mapping result image by a terminal and wire label matching algorithm to obtain a matching result.

[0010] Compared with the prior art, the present invention has the following technical effects: 1. The PDU terminal and wire label images are processed by the region of interest detection network, which can effectively eliminate the interference of non-interest areas (background elements) on the OCR recognition of terminal labels and wire labels; 2. The image of interest is processed by the terminal label and wire label area detection network, which can effectively eliminate the mutual interference of terminal label and wire label OCR recognition, and rotate and straighten the image, greatly improving the recognition rate of terminal label and wire label; 3. The rotated and straightened image is processed by the OCR recognition network, and then processed by the original image mapping algorithm and the terminal and wire label matching algorithm, which can well realize the intelligent matching of PDU terminals and wire labels, thereby improving the accuracy and efficiency of management.

[0011] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. 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: Figure 1 It is a flow chart of the method for intelligently matching PDU terminals and wire labels based on image processing of the present invention; Figures 2 to 10 The images corresponding to the steps in the matching method of the first embodiment of the present invention are shown, wherein: Figure 2It is the image obtained after the original image is processed by the region of interest detection network model; Figure 3 yes Figure 2 The image shown is the image of interest obtained after cropping; Figure 4 yes Figure 3 The image of interest shown is obtained after being processed by the terminal label and wire label area detection network model; Figure 5 yes Figure 4 The image shown is a terminal label image and a wire image obtained by cropping the image; Figure 6 yes Figure 5 The image shown is a rotated and rectified image obtained after the image is rotated and rectified; Figure 7 yes Figure 6 The rotated and straightened image shown is the recognition result image obtained after passing through the OCR recognition model; Figure 8 yes Figure 7 The recognition result image shown is processed by the original image mapping algorithm to obtain a mapping result image; Fig. 9 yes Figure 8 Flowchart showing the mapping result image being processed by the terminal and wire label matching algorithm; Fig.10 yes Figure 8 The matching result diagram of the mapping result image shown; Figures 11 to 18 The images corresponding to the steps in the matching method of the second embodiment of the present invention are shown, wherein: Fig.11 It is the image obtained after the original image is processed by the region of interest detection network model; Fig.12 yes Fig.11 The image shown is the image of interest obtained after cropping; Fig.13 yes Fig.12 The image of interest shown is obtained after being processed by the terminal label and wire label area detection network model; Fig.14 yes Fig.13 The image shown is a terminal label image and a wire image obtained by cropping the image; Fig.15 yes Fig.14 The image shown is a rotated and rectified image obtained after the image is rotated and rectified; Fig.16 yes Fig.15 The rotated and straightened image shown is the recognition result image obtained after passing through the OCR recognition model; Fig.17yes Fig.16 The recognition result image shown is processed by the original image mapping algorithm to obtain a mapping result image; Fig.18 yes Fig.17 The matching result diagram of the mapping result image shown; Figures 19 to 26 The images corresponding to the steps in the matching method of the third embodiment of the present invention are shown, wherein: Fig.19 It is the image obtained after the original image is processed by the region of interest detection network model; Fig. 20 yes Fig.19 The image shown is the image of interest obtained after cropping; Fig.21 yes Fig. 20 The image of interest shown is obtained after being processed by the terminal label and wire label area detection network model; Fig. 22 yes Fig.21 The terminal label image and wire image are obtained by cropping the image shown; Fig.23 yes Fig. 22 The image shown is a rotated and straightened image obtained after being rotated and straightened; Fig.24 yes Fig.23 The rotated and straightened image shown is the recognition result image obtained after passing through the OCR recognition model; Fig.25 yes Fig.24 The recognition result image shown is processed by the original image mapping algorithm to obtain a mapping result image; Fig.26 yes Fig.25 The matching result graph of the mapping result image is shown. DETAILED DESCRIPTION

[0013] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0014] Combined with reference Figure 1 The method for intelligently matching PDU terminals and wire labels based on image processing of the present invention comprises the following steps S1-S5.

[0015] S1. Obtain the original image of the terminal and wire label, and then send it to the region of interest detection network model for detection, and crop the original image according to the detection result to obtain the image of interest; S2. Then send it to the terminal label and wire label region detection network model for detection, and crop and rotate the image of interest according to the detection result to obtain the terminal label image and the wire label image; S3. Then send it to the OCR recognition network for recognition to obtain the recognition result image; S4. Then obtain the mapping result image after processing by the mapping original image algorithm; S5. Then obtain the matching result after processing by the terminal and wire label matching algorithm.

[0016] The network model and algorithm adopted by the present invention are described below.

[0017] 1. Region of Interest Detection Network Model (1) Function: Detect the region of interest for the input terminal and wire label images; (2) Function: Eliminate interference of non-interest areas (background elements) on OCR recognition of terminal labels and wire labels; (3) Input: terminal and wire label images; (4) Output: coordinate values ​​of the area of ​​interest , Area of ​​Interest Category .

[0018] in, Contains the horizontal coordinate of the upper left corner of the region of interest , vertical coordinate , Width of region of interest , height of area of ​​interest , containing the area of ​​interest .

[0019] 2. Terminal label and wire label area detection network model (1) Function: Detect terminal label and wire label areas for the input image of interest; (2) Function: Eliminate mutual interference between terminal label and wire label OCR recognition; (3) Input: image of interest; (4) Output: Terminal label area coordinate values , wire label area coordinate value , terminal and wire label categories .

[0020] in, Contains the horizontal coordinate of the upper left corner of the terminal label area , vertical coordinate , Terminal label area width , Terminal label area height , Contains the horizontal coordinate of the upper left corner of the wire label area , vertical coordinate , Width of wire label area , Wire label area height , Contains positive terminal labels Terminal labels that need to be rotated 90 degrees clockwise , Wire labels that need to be rotated 90 degrees clockwise , Wire labels that need to be rotated 90 degrees counterclockwise .

[0021] 3. OCR recognition network model (1) Function: Perform OCR recognition on the input rotated and straightened image; (2) Function: Detect and identify the terminal labels and wire labels in the input rotated and rectified image, and obtain the corresponding coordinate results and text results for subsequent algorithm processing; (3) Input: rotate and straighten the image; (4) Output: Correct terminal label coordinate recognition result , the correct terminal label text recognition result , the terminal label coordinate recognition result needs to be rotated 90 degrees clockwise , the terminal label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees clockwise , the wire label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees counterclockwise , the wire label text recognition result needs to be rotated 90 degrees counterclockwise .

[0022] in, Contains the coordinates of the upper left corner of the terminal label , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the terminal label that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label that needs to be rotated 90 degrees counterclockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner .

[0023] 4. Mapping original image algorithm (1) Function: Map the coordinate recognition results and text recognition results in the input recognition result image, and map the results back to the terminal and wire label images; (2) Function: Map the results back to the terminal and wire label images for subsequent algorithm processing; (3) Input: coordinate recognition results and text recognition results in the recognition result image; (4) Output: Coordinate mapping result of the correct terminal label on the terminal and wire label images , the text mapping result of the correct terminal label on the terminal and wire label image The coordinate mapping result of the terminal label that needs to be rotated 90 degrees clockwise on the terminal and wire label images The text mapping result of the terminal label that needs to be rotated 90 degrees clockwise on the terminal and wire label images , the coordinate mapping result of the wire label that needs to be rotated 90 degrees clockwise on the terminal and wire label image , the text mapping result of the wire label that needs to be rotated 90 degrees clockwise on the terminal and wire label image , the coordinate mapping result of the wire label that needs to be rotated 90 degrees counterclockwise on the terminal and wire label image , the text mapping result of the wire label on the terminal and wire label image needs to be rotated 90 degrees counterclockwise .

[0024] in, Contains the coordinates of the upper left corner of the correct terminal label coordinate mapping result , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the terminal label coordinate mapping result that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label coordinate mapping result that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label coordinate mapping result that needs to be rotated 90 degrees counterclockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner .

[0025] 5. Terminal and wire label matching algorithm (1) Function: Intelligently match terminal and wire labels based on the coordinate mapping results and text mapping results in the input mapping result image; (2) Function: Realize intelligent matching of PDU terminals and wire labels, thereby improving the accuracy and efficiency of management; (3) Input: coordinate mapping results and text mapping results in the mapping result image; (4) Output: Output the PDU terminal and wire label matching results.

[0026] Figures 2 to 10 The first embodiment of the method for intelligently matching PDU terminals and wire labels according to the present invention is shown below. Figures 2 to 10 Each step of the matching method of this embodiment is described.

[0027] 1. PDU terminal and wire label image acquisition and preprocessing (1) Image acquisition: Use a camera to obtain terminal and wire label image sets, denoted as ;in, represents the i-th terminal and wire label image, and I is the total number of terminal and wire label images; (2) Image preprocessing: Annotate the terminal and wire label regions in the terminal and wire label image set A, and obtain an image set of interest denoted as ;in Indicates the i-th terminal and wire label image The annotated images of interest; The terminal label area and wire label area in the image set of interest B are labeled, so that the terminal label image set is recorded as and the wire label image set is denoted as ,in Represents the i-th image of interest Terminal label image marked, Represents the i-th image of interest Annotated wire label image.

[0028] The terminal label text in the terminal label image set C is annotated to obtain the terminal label text set denoted as ,in Represents the i-th terminal label image The marked terminal label text image; the wire label text in the wire label image set D is marked, so that the wire label text set is recorded as ,in Represents the i-th wire label image Annotated wire label text image.

[0029] 2. Network model training (1) Region of Interest Detection Model Training: Using the image set B of interest and the loss function shown in formula (1) , train the region of interest detection network to obtain the region of interest detection model ; (1) In formula (1), represents the classification loss function, represents the intersection-over-union loss function, represents the distribution focal loss function.

[0030] (2) Terminal label and wire label area detection model training: Using the terminal label image set C, the wire label image set D and the loss function shown in formula (1) , train the terminal label and wire label area detection network to obtain the terminal label and wire label area detection model .

[0031] (3) OCR recognition model training: The terminal label text set E and the wire label text set F are used to train the OCR recognition network to obtain the OCR recognition model PD.

[0032] 3. Reasoning of network models (1) Detection model of region of interest , terminal label and wire label area detection model The three models PD are logically cascaded, and the terminal and wire label images are input to perform forward reasoning of the model.

[0033] (2) Detection model after region of interest After processing, the coordinate values ​​of the area of ​​interest are obtained , area of ​​interest category ,in Contains the horizontal coordinate of the upper left corner of the region of interest , vertical coordinate , Width of region of interest , height of area of ​​interest , Include area of ​​interest ; Draw graphics based on the processed information, as follows Figure 2 shown.

[0034] (3) The coordinate values ​​of the region of interest obtained in step (2) The terminal and wire label images are cropped to obtain images of interest. The images of interest are as follows: Figure 3 As shown, the image of interest is then fed into the terminal label and wire label region detection model In the detection model, the terminal label and wire label area are detected After processing, the coordinate value of the terminal label area is obtained , wire label area coordinate value , terminal and wire label categories ,in Contains the horizontal coordinate of the upper left corner of the terminal label area , vertical coordinate , Terminal label area width , Terminal label area height , Contains the horizontal coordinate of the upper left corner of the wire label area , vertical coordinate , Width of wire label area , Wire label area height , Contains positive terminal labels Terminal labels that need to be rotated 90 degrees clockwise , Wire labels that need to be rotated 90 degrees clockwise , Wire labels that need to be rotated 90 degrees counterclockwise ; Draw graphics based on the processed information, as follows Figure 4 shown.

[0035] (4) The coordinate values ​​of the terminal label area obtained in step (3) , wire label area coordinate value The image of interest is cropped to obtain the terminal label image and the wire label image, as shown in the following figure. Figure 5 shown.

[0036] According to the terminal and wire label categories obtained in step (3) Rotate the terminal label image and wire label image. No rotation required, Rotate 90 degrees clockwise, Rotate 90 degrees clockwise, Rotate 90 degrees counterclockwise to obtain a rotated and straightened image, as shown below: Figure 6 shown.

[0037] (5) The rotated and straightened image is sent to the OCR recognition model PD for processing. After being processed by the OCR recognition model PD, the correct terminal label coordinate recognition result is obtained. , the correct terminal label text recognition result , the terminal label coordinate recognition result needs to be rotated 90 degrees clockwise , the terminal label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees clockwise , the wire label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees counterclockwise , the wire label text recognition result needs to be rotated 90 degrees counterclockwise .

[0038] in, Contains the coordinates of the upper left corner of the terminal label , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the terminal label that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label that needs to be rotated 90 degrees counterclockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner ; Draw a graph based on the processed information to obtain the recognition result image, as shown in the following figure: Figure 7 shown.

[0039] (6) The recognition result image is sent to the original image mapping algorithm for processing. After being processed by the original image mapping algorithm, the coordinate mapping result of the correct terminal label on the terminal and wire label image is obtained. , the text mapping result of the correct terminal label on the terminal and wire label image The coordinate mapping result of the terminal label that needs to be rotated 90 degrees clockwise on the terminal and wire label images The text mapping result of the terminal label that needs to be rotated 90 degrees clockwise on the terminal and wire label images , the coordinate mapping result of the wire label that needs to be rotated 90 degrees clockwise on the terminal and wire label image , the text mapping result of the wire label that needs to be rotated 90 degrees clockwise on the terminal and wire label image , the coordinate mapping result of the wire label that needs to be rotated 90 degrees counterclockwise on the terminal and wire label image , the text mapping result of the wire label on the terminal and wire label image needs to be rotated 90 degrees counterclockwise .

[0040] in, Contains the coordinates of the upper left corner of the correct terminal label coordinate mapping result , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the terminal label coordinate mapping result that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label coordinate mapping result that needs to be rotated 90 degrees clockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner , Contains the coordinates of the upper left corner of the wire label coordinate mapping result that needs to be rotated 90 degrees counterclockwise , the coordinates of the upper right corner , the coordinates of the lower right corner , the coordinates of the lower left corner ; Draw graphics based on the processed information to obtain the mapping result image, as shown in the following figure: Figure 8 shown.

[0041] in, and The mapping relationship between them is as follows: (2) in, and The mapping relationship between them is as follows: (3) in, and The mapping relationship between them is as follows: (4) in, and The mapping relationship between them is as follows: (5) in, , , , and , , , The mapping relationship between them is as follows: (6).

[0042] 7) Send the mapping result image to the terminal and wire label matching algorithm for processing. After the processing of the terminal and wire label matching algorithm, the PDU terminal and wire label matching result is obtained. If the match is successful, the match success mark is returned, and if the match fails, the match false mark is returned. The algorithm idea is to first match the text on the terminal and wire label, and then perform coordinate matching when the text matching is successful. Only when both the text matching and the coordinate matching are successful, it indicates that the matching is successful, otherwise the matching fails. The specific algorithm flow chart is as follows Fig. 9 As shown, the matching results are as follows Fig.10 shown.

[0043] Figures 11 to 18 A second embodiment of the method for intelligently matching PDU terminals and wire labels according to the present invention is shown, which is different from the first embodiment in that the shape of the PDU terminal changes, the text on the PDU terminal label is arranged horizontally, and the text on the wire label is arranged vertically. Figures 19 to 26 A third embodiment of the method for intelligently matching PDU terminals and wire labels according to the present invention is shown, which is different from the first embodiment in that the shape of the PDU terminal is changed, the text on the PDU terminal label is arranged horizontally, and the text on the wire label is arranged vertically.

[0044] In the first and second embodiments, as Fig.10 and Fig.18As shown, the two pictures fail to match. In the third embodiment, Fig.26 As shown in the figure, the image is matched successfully. The reason for failure is that some terminal and wire label texts and coordinates do not match. The reason for success is that the terminal and wire label texts and coordinates are matched successfully.

[0045] The present invention also provides a computer program product, including a computer program, which, when executed, is used to implement the steps of the above-mentioned method for intelligent matching of PDU terminals and wire labels based on image processing.

[0046] The present invention also provides a PDU terminal and wire label intelligent matching detection device based on image processing, which stores a computer program, and when the program is executed, it is used to implement the various steps of the above-mentioned PDU terminal and wire label intelligent matching method based on image processing.

[0047] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for intelligently matching PDU terminals and wire labels based on image processing, characterized in that: include: S1. Obtain the original image containing the PDU terminal and the wire label, and then send it to the region of interest detection network model for detection. According to the detection result, the original image is cropped to obtain the region of interest where the PDU terminal and the wire label are located; S2. Send the image of interest to the terminal label and wire label area detection network model for detection, and crop and rotate the PDU terminal and wire label images respectively according to the detection results to obtain the terminal label image and the wire label image; S3, sending the rotated and rectified terminal label image and wire label image to the OCR recognition network model for recognition, and obtaining a recognition result image; S4, processing the recognition result image by a mapping original image algorithm to obtain a mapping result image; S5. The terminal and wire labels in the mapping result image are processed by a terminal and wire label matching algorithm to obtain a matching result.

2. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The region of interest detection network model is used to detect the region of interest on the input original image and output the coordinate value of the region of interest , Area of ​​Interest Category , wherein the region of interest detection network model is pre-trained.

3. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The terminal label and wire label area detection network model is used to detect the terminal label and wire label area of ​​the input image of interest and output the coordinate value of the terminal label area. , wire label area coordinate value , terminal and wire label categories , wherein the terminal label and wire label area detection network model is pre-trained.

4. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The OCR recognition network model is used to perform OCR recognition on the input rotated and rectified image and output the correct terminal label coordinate recognition result. , the correct terminal label text recognition result , the terminal label coordinate recognition result needs to be rotated 90 degrees clockwise , the terminal label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees clockwise , the wire label text recognition result needs to be rotated 90 degrees clockwise , the wire label coordinate recognition result needs to be rotated 90 degrees counterclockwise , the wire label text recognition result needs to be rotated 90 degrees counterclockwise .

5. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The region of interest detection network model , terminal label and wire label area detection network model The three models, namely, the OCR recognition network model PD, are logically cascaded, the original image is input, and the model forward reasoning is performed to obtain the terminal label image and the wire label image.

6. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The original image mapping algorithm is used to map the coordinate recognition results and text recognition results in the input recognition result image, map the results back to the original image, and output: the coordinate mapping result of the correct terminal label on the original image , the text mapping result of the correct terminal label on the original image , the coordinate mapping result of the terminal label that needs to be rotated 90 degrees clockwise on the original image , the text mapping result of the terminal label on the original image that needs to be rotated 90 degrees clockwise , the coordinate mapping result of the wire label that needs to be rotated 90 degrees clockwise on the original image , the text mapping result of the wire label on the original image needs to be rotated 90 degrees clockwise , the coordinate mapping result of the wire label that needs to be rotated 90 degrees counterclockwise on the original image , the text mapping result of the wire label on the original image needs to be rotated 90 degrees counterclockwise .

7. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 1 is characterized in that: The terminal and wire label matching algorithm is used to perform intelligent matching of terminals and wire labels on the coordinate mapping results and text mapping results in the input mapping result image, and output the PDU terminal and wire label matching results.

8. The method for intelligently matching PDU terminals and wire labels based on image processing according to claim 7 is characterized in that: The terminal and wire label matching algorithm is used to first match the text on the terminal and wire labels, and then perform coordinate matching when the text matching is successful. Only when both the text matching and the coordinate matching are successful, the matching is successful. Otherwise, the matching fails and a matching success mark or a matching failure mark is returned.

9. A computer program product, comprising a computer program, characterized in that When executed, the program is used to implement the steps of the method for intelligent matching of PDU terminals and wire labels based on image processing according to any one of claims 1 to 8.

10. A PDU terminal and wire label intelligent matching detection device based on image processing, characterized in that: It stores a computer program, which, when executed, is used to implement the steps of the method for intelligent matching of PDU terminals and wire labels based on image processing according to any one of claims 1 to 8.

Images