A method for detecting the processing and production quality of a complete switchgear based on image analysis

Abstract

The present application belongs to the technical field of image processing, and particularly relates to a complete switchgear processing production quality detection method based on image analysis, which comprises the following steps: obtaining a to-be-detected image, recognizing a mask image of each cable and a rectangular frame of a wire pressing nose, and recognizing a rectangular frame of each terminal; judging a connection attribution relationship of any cable and any terminal, comprising: determining an overlapping degree and a main direction vector of each rectangular frame, and calculating an affinity degree; constructing an expected trend vector and an actual trend vector, and calculating a trend consistency; fusing a distance between the cable wire pressing nose and the terminal, the affinity degree and the trend consistency to determine a connection confidence; for any cable, obtaining a terminal corresponding to a maximum value of the connection confidence of the cable and all terminals as a connection attribution; if the connection attribution is inconsistent with a standard connection attribution, triggering a warning. The present application improves the accuracy of the connection attribution judgment and ensures the reliability of the production quality evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing. More particularly, the present application relates to a method for detecting the processing and production quality of a complete switchgear based on image analysis. BACKGROUND

[0002] A complete switchgear, such as a high-low voltage switch cabinet, a distribution box, etc., is an indispensable key component in the power system. In the processing and production of a complete switchgear, the correct connection between the internal cable and the terminal strip is the core foundation for ensuring the safe operation of the equipment. Traditional quality detection mainly relies on manual visual point-by-point checking, which has the problems of low efficiency, poor quality consistency, and lack of digital traceability records.

[0003] At present, the automatic detection technology based on image analysis, for example, the nearest neighbor point algorithm based on geometric constraints, locates the center point of the cable pressing nose and the center point of the terminal in the image through the target detection algorithm, calculates the Euclidean distance between the center point of the cable pressing nose and all the terminal center points around it, determines the terminal with the shortest distance as the connection object of the cable, and determines whether the connection relationship between them meets the design drawing, so as to evaluate whether the processing and production quality is qualified.

[0004] However, if the cable and the terminal are simply simplified as a single geometric center point, and the Euclidean distance is used as the judgment criterion for connection attribution, in the scene of high-density arrangement of electrical equipment, due to the variability of cable connection joints, the cable pressing nose may have irregular shapes. The minimum circumscribed rectangle frame recognized by the target detection algorithm will be affected by the irregular shape, causing the coordinates of the positioned center point to be deviated towards the adjacent terminal, and finally resulting in inaccurate judgment of the connection attribution, which affects the evaluation of the processing and production quality of the complete switchgear. SUMMARY

[0005] To solve the technical problem that the shortest Euclidean distance of the cable pressing nose and the geometric center point of the terminal only leads to the connection attribution judgment error, the application provides a complete switch device processing and production quality detection method based on image analysis, which comprises the following steps: obtaining a to-be-detected image, identifying the mask image of each cable and the rectangular frame of the cable pressing nose, and the rectangular frame of each terminal; judging the connection attribution relationship between any cable and any terminal, which comprises the following steps: based on the position distribution of the pixel points in the rectangular frame of the cable pressing nose and the terminal, obtaining the overlapping degree and the main direction vector of each rectangular frame; taking the average of the parallel degree of the overlapping degree and the main direction vector as the affinity; performing skeleton extraction on the mask image of the cable, obtaining the center line path and calculating the centroid of the cable; taking the vector from the centroid of the cable to the centroid of the terminal as the expected trend vector; extracting all the pixel points of the center line path in the rectangular frame of the cable pressing nose and calculating the centroid of the cable pressing nose, and taking the vector from the centroid of the cable pressing nose to the centroid of the terminal as the actual trend vector; determining the trend consistency according to the cosine similarity of the expected trend vector and the actual trend vector; fusing the distance between the cable pressing nose and the terminal, the affinity and the trend consistency to determine the connection confidence of the cable and the terminal; for any cable, obtaining the terminal corresponding to the maximum value of the connection confidence of the cable and all terminals as the connection attribution of the cable; if the connection attribution is inconsistent with the standard connection attribution, triggering a warning.

[0006] The innovation of the application lies in constructing a multi-dimensional connection confidence evaluation model which no longer relies on a single criterion. Firstly, the consistency of the overlapping degree and the main direction vector of the rectangular frame of the cable pressing nose and the terminal is comprehensively evaluated to determine whether the shape is consistent; further, the expected trend and the actual trend of the cable skeleton are compared to determine whether the access mode conforms to the physical wiring rules; finally, the traditional geometric distance is retained as one of the evaluation factors, which is fused with the above two factors to ensure that a connection meets the spatial proximity, shape matching and reasonable wiring principles, effectively eliminates the uncertainty caused by single-dimensional evaluation, and makes the final connection attribution determination more robust and accurate.

[0007] Preferably, the mask image of each cable and the rectangular frame of the cable pressing nose, and the rectangular frame of each terminal are obtained by recognizing the to-be-detected image by using a pre-trained Mask R-CNN model.

[0008] Preferably, the overlapping degree is obtained by extracting the positions of all the pixel points in the rectangular frame of the cable pressing nose and the terminal, counting the number of pixel points with the same position, and taking the ratio of the number to the number of all the pixel points in the rectangular frame of the cable pressing nose as the overlapping degree.

[0009] Preferably, the main direction vector of each rectangular frame is obtained by using a principal component analysis algorithm to analyze the positions of all the pixel points in each rectangular frame to obtain a first principal component direction vector as the main direction vector.

[0010] Preferably, the parallel degree of the major direction vector is equal to the absolute value of the cosine similarity of the major direction vector of the rectangular frame of the crimping nose and the major direction vector of the rectangular frame of the terminal.

[0011] The application uses the absolute value of the cosine similarity between the major direction vectors to evaluate the parallelism of the crimping nose and the terminal in the direction, which can effectively evaluate whether rotation or inclination occurs in the crimping process, and improves the robustness of the direction parallelism discrimination.

[0012] Preferably, the center line path is obtained by applying a skeleton extraction algorithm to the cable mask image.

[0013] Preferably, the center of mass of the cable is equal to the average of the positions of all pixel points on the center line path of the cable; the center of mass of the crimping nose is equal to the average of the positions of all pixel points in the rectangular frame of the crimping nose of the cable; and the center of mass of the rectangular frame is equal to the average of the positions of all pixel points in the rectangular frame.

[0014] Preferably, the trend consistency satisfies the expression: ; in the expression, is the actual and expected trend consistency of the cable pointing to the terminal; . is the expected trend vector of the cable pointing to the terminal; . is the actual trend vector of the cable pointing to the terminal; . is the cosine similarity function; is the natural exponential function.

[0015] The application compares the expected trend vector and the actual trend vector of the cable to evaluate whether the access mode conforms to the physical wiring rules, which can effectively identify the distorted and unnatural connection in the access direction and exclude suspicious connections that do not meet the safety requirements.

[0016] Preferably, the connection confidence of the cable and the terminal satisfies the expression: ; in the expression, is the connection confidence score of the cable and the terminal . is the Euclidean distance between the center of mass of the rectangular frame of the crimping nose of the cable and the center of mass of the rectangular frame of the terminal . is the maximum value of the Euclidean distance between the center of mass of the rectangular frame of the crimping nose of the cable and the center of mass of the rectangular frame of all terminals. For the cable The affinity between the crimping nose and the terminal ; For the cable The actual and expected trend consistency of the cable to the terminal .

[0017] The present application fuses the distance, affinity and trend consistency between the cable crimping nose and the terminal, and constructs a multi-dimensional connection confidence. This multi-dimensional feature fusion method avoids the uncertainty caused by relying on single geometric distance evaluation, and makes the connection attribution judgment more accurate.

[0018] Preferably, the standard connection attribution acquisition method is: according to the label of any cable, extracting the standard connection attribution of the cable from the complete switchgear parameter document; wherein the label of the cable is obtained by extracting the mask image of the cable through an optical character recognition network.

[0019] The present application automatically acquires the cable label through optical character recognition, and extracts the standard connection attribution from the parameter document, realizes the adaptive query and matching process, and significantly improves the quality detection efficiency.

[0020] The present application has the following advantages:

[0021] (1) The present application uses a target detection model to locate the rectangular frame of the cable mask image and the crimping nose and the rectangular frame of the terminal, which provides a fine image basis for subsequent accurate extraction of the skeleton, calculation of the center of mass and analysis of the shape, and guarantees the accuracy of feature calculation;

[0022] (2) By extracting the skeleton of the cable mask image, and introducing the trend consistency of the expected trend and the actual trend of the cable to the terminal, the connection relationship that does not conform to the physical wiring rule can be identified, and the reliability of the detection is further improved;

[0023] (3) By constructing multi-dimensional indicators to evaluate the connection confidence, the defect that the traditional method only relies on the shortest distance in the high-density arrangement scene of electrical equipment to easily lead to misjudgment is overcome, and the robustness of the connection attribution determination is improved. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 is a flowchart schematically showing a complete switchgear processing and production quality detection method based on image analysis in the present application. DETAILED DESCRIPTION

[0025] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0026] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0027] The embodiments of the present application disclose a complete switchgear processing and production quality detection method based on image analysis, referring to Figure 1 , including steps S1 to S5:

[0028] S1, acquiring an image to be detected, identifying the mask image of each cable, the rectangular frame of the wire pressing nose, and the rectangular frame of each terminal.

[0029] It should be noted that, due to the variability and irregularity of the cable connection crimping area form, it is necessary to obtain a variety of information that can describe the component form, size and direction, so that different types of cable connection images can be obtained, and instance-level labeling can be performed, including: marking the pixel-level mask of each cable, the rectangular frame of the terminal and the wire pressing nose area; using the labeled picture as a training set, training using the existing Mask R-CNN model, to ensure that the mask image of each cable, the rectangular frame of the wire pressing nose and the rectangular frame of each terminal can be accurately identified in the subsequent detection process, to provide accurate analysis images for the subsequent process.

[0030] Specifically, an image of the internal terminal row area of the complete switchgear to be detected is acquired, the image is preprocessed, including Gaussian denoising and grayscale processing, and a pre-trained Mask R-CNN model is used to acquire the mask image of each cable and the rectangular frame of the wire pressing nose, as well as the rectangular frame of each terminal; at the same time, the positions of the pixel points in each rectangular frame also need to be obtained, and the specific method is as follows: taking the pixel point at the lower left corner of the image as the origin, taking the direction horizontally to the right at the origin as the positive direction of the axis, and taking the direction vertically upward at the origin as the positive direction of the axis, to realize the construction of the rectangular coordinate system; the position of each pixel point in the image is obtained in the rectangular coordinate system, including the coordinate in the axis direction and the coordinate in the axis direction.

[0031] Considering that there are different types of terminal ports and cables, corresponding labels are provided in the terminal area and cable area during the processing of the switch device, so that the staff can accurately install; therefore, the label of the mask area of each cable is obtained as the label of each cable by using an optical character recognition network (OCR), and the label of the rectangular frame area of each terminal is obtained as the label of each terminal.

[0032] At this point, the mask of each cable, the rectangular frame of the wire pressing nose, and the label in the to-be-detected image are obtained, as well as the rectangular frame and the label of each terminal.

[0033] S2, determining the connection belonging relationship of any cable and any terminal, comprising: based on the position distribution of the pixel points in the rectangular frame of the wire pressing nose and the terminal, determining the overlapping degree and the main direction vector of each rectangular frame, and calculating the affinity.

[0034] It should be noted that after recognizing the rectangular frame of each cable wire pressing nose and terminal, the shape contour of a correctly crimped cable end must be highly consistent with the opening shape of the terminal. By performing principal component analysis on the pixel points in each rectangular frame, the first principal component direction is obtained, that is, the main direction of the rectangular frame area. Under normal circumstances, the terminal row of an electrical device is usually arranged in a horizontal or vertical line, and the main direction of a correctly connected wire pressing nose must be almost parallel to the main direction of the terminal, and there is an overlapping area between the wire pressing nose and the rectangular frame of the connected terminal. Therefore, the affinity between the wire pressing nose and the terminal is evaluated by the changes in the main direction of the rectangular frame area of the wire pressing nose and the terminal and the overlapping degree of the rectangular frame.

[0035] Specifically, the overlapping degree of any cable wire pressing nose and any terminal is obtained in the following manner: the positions of all pixel points in the rectangular frame of the wire pressing nose and the terminal are extracted, the number of pixel points with the same position is counted, and the ratio of the number to the number of all pixel points in the rectangular frame of the wire pressing nose is taken as the overlapping degree.

[0036] The main direction vector of the rectangular frame of any terminal is obtained in the following manner: the positions of all pixel points in the rectangular frame are analyzed by using a principal component analysis algorithm to obtain the first principal component direction vector as the main direction vector. Similarly, the main direction vector of the rectangular frame of any cable wire pressing nose is obtained. It should be noted that the principal component analysis algorithm is a known technology and will not be described here.

[0037] The affinity is determined according to the overlapping degree and the parallel degree of the main direction vector; the affinity satisfies the expression:

[0038]

[0039] In the formula, is the cable the overlap degree between the crimping nose and the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal. the main direction vector of the rectangular frame of the terminal.

[0040] wherein, reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction. reflects the parallel degree between the main direction vectors of the crimping nose and the terminal, the value tends to 1, the included angle between the two main direction vectors is closer to 0 degrees or 180 degrees, which means that the crimping nose and the terminal are more parallel in direction.

[0041] Thus far, the affinity between any crimping nose and any terminal is obtained.

[0042] S3, skeleton extraction is performed on the mask image of the cable to obtain a center line path and calculate the centroid of the cable; a vector from the centroid of the cable to the centroid of the rectangular frame of the terminal is taken as an expected trend vector; all pixel points of the center line path located in the rectangular frame of the crimping nose are extracted and the centroid of the crimping nose is calculated, and a vector from the centroid of the crimping nose to the centroid of the rectangular frame of the terminal is taken as an actual trend vector; the cosine similarity between the expected trend vector and the actual trend vector is determined to determine the trend consistency.

[0043] It should be noted that a normal cable connection not only requires a proper shape, but also that its connection method conforms to physical laws. For example, a cable may be led out from the left wiring channel, loop around once, and finally connect to the terminal on the right. However, from a partial viewpoint in the image, the cable lugs may have a high degree of fit with the terminal, but the connection direction is twisted and unnatural, which is a connection method that does not conform to physical laws. This situation does not meet the requirements of safe production in electrical equipment. Therefore, by comparing and analyzing the actual trend of the cable lugs pointing towards the center of the terminal with the expected trend of the cable, the greater the difference between the two directions, the more suspicious the connection method is.

[0044] Specifically, for any cable mask image, the skeleton extraction algorithm is used to obtain the centerline path of the cable, extract all pixels within the rectangle of the cable nose where the centerline path is located, and use the average value of the positions of all pixels as the centroid of the cable nose; calculate the average value of the positions of all pixels on the centerline path of the cable as the centroid of the cable.

[0045] The vector pointing from the centroid of the cable to the centroid of the rectangle of any terminal is taken as the desired trend vector; the vector pointing from the centroid of the cable lug to the centroid of the rectangle of any terminal is taken as the actual trend vector; it should be noted that the centroid of the rectangle of any terminal is equal to the average position of all pixels within its rectangle; similarly, the centroid of the rectangle of any cable lug is obtained.

[0046] The consistency of the trend is determined by the cosine similarity between the expected trend vector and the actual trend vector; the consistency of the trend satisfies the expression:

[0047]

[0048] In the formula, For cables Pointer terminal The actual trend is consistent with the expected trend; For cables Pointer terminal The expected trend vector; For cables Pointer terminal The actual trend vector; To take the cosine similarity function; It is a natural exponential function.

[0049] in, Reflecting cables Pointer terminal The similarity in direction between the expected trend vector and the actual trend vector. A value closer to 1 indicates that the directions of the two vectors are more similar, meaning the cable... The more consistent the overall direction of the cable with its actual terminal connection direction, the better; conversely, the closer the value is to -1, the worse the cable performance. The more inconsistent the overall direction is with its actual terminal access direction; through Will The value is mapped to the interval [0,1], that is... The closer the value is to -1, Approaching 0.135; The closer the value is to 1, Approaching 1.

[0050] For example, terminals The centroid of the rectangle is Cable The center of mass is Cable The end position is Expected trend vector: Actual trend vector: , Among them, the expected trend was to the right, while the actual trend was upward, indicating a low consistency in the trend, suggesting that the cable... Pointer terminal The more suspicious the access method, the more suspicious it becomes.

[0051] At this point, the consistency between the actual and expected trajectory of any cable pointing to any terminal is obtained.

[0052] S4. Combine the distance, fit, and consistency of the cable crimp lugs and terminals to determine the connection confidence between the cable and the terminals.

[0053] It should be noted that traditional quality inspection methods rely solely on the geometric nearest distance to determine connection attribution. However, the variability of cable connectors within actual electrical equipment, along with the irregular shapes of cable lugs and terminals, can easily lead to offset of the positioning center point, resulting in incorrect connection attribution. Therefore, by integrating the acquired affinity and trend consistency with traditional geometric distance, a comprehensive connection confidence score is constructed. Through this fusion of multi-dimensional features, a connection must meet all of the above conditions to be judged as a high-confidence correct connection. This eliminates the uncertainty brought about by single-dimensional evaluation, making the final connection attribution determination more robust.

[0054] Specifically, the connection confidence level is determined based on the distance between the cable lug and the terminal, the area affinity, and the consistency of the cable routing; the connection confidence level satisfies the expression:

[0055]

[0056] In the formula, For cables With terminals The connection confidence score; For cables The centroid of the rectangular frame of the crimping lug and the terminal The Euclidean distance of the centroid of the rectangle; For cables The maximum Euclidean distance between the centroid of the rectangular frame of the wire nose and the centroids of the rectangular frames of all terminals; For cables crimping lugs and terminals The degree of affinity between them; For cables Pointer terminal The actual trend is consistent with the expected trend.

[0057] in, Reflecting cables With terminals Geometric similarity The smaller the value, the better the cable crimping lugs and terminals The smaller the geometric distance, that is, the smaller the cable in space With terminals The closer; Reflecting cables With terminals The connection confidence score indicates that the higher the value, the better the cable connection confidence score. With terminals Not only are they spatially similar, but they also match closely in form and their access methods conform to physical cabling principles, meaning that the cables... Connected to the terminal The higher the confidence level, the better. It should be noted that in actual complete switchgear systems, because terminals have physical dimensions and are distributed in different locations, It is always greater than zero.

[0058] At this point, the connection confidence level between any cable and any terminal has been obtained.

[0059] S5. For any cable, obtain the terminal corresponding to the maximum connection confidence of its connection with all terminals, and use it as its connection assignment; if the connection assignment is inconsistent with the standard connection assignment, trigger an alarm.

[0060] It should be noted that the wiring within a complete set of switchgear follows the standard wiring diagram corresponding to its equipment model. The correctness of the connection between internal cables and terminal blocks is the core foundation for ensuring the safe operation of the equipment. Therefore, when conducting quality inspection, the standard wiring table corresponding to the switchgear model can be extracted in advance to obtain the standard connection assignment of each cable, and this can be used to assess whether the connection relationship is correct.

[0061] Specifically, for any cable, the connection confidence thereof with all terminals is obtained, and the terminal corresponding to the maximum of the connection confidence is taken as the connection attribution of the cable.

[0062] According to the label of any cable, the standard connection attribution of the cable is extracted from the complete switchgear parameter document. By comparing the connection attribution of the cable with the standard connection attribution, if the connection attribution is consistent with the standard connection attribution, it indicates that the cable is correctly connected with the terminal, which means that the assembly quality of the cable is qualified. If the connection attribution is inconsistent with the standard connection attribution, it means that the assembly quality of the cable is unqualified, which immediately triggers a warning and sends a message that the switchgear processing production quality is unqualified, and notifies relevant personnel to check the wiring.

Claims

1. A method for quality inspection of complete switchgear manufacturing based on image analysis, characterized in that, include: Acquire the image to be inspected, identify the mask image and the rectangular box of the wire lug of each cable, as well as the rectangular box of each terminal; Determine the connection relationship between any cable and any terminal, including: Based on the positional distribution of pixels within the rectangular frames of the pressure nose and terminals, the overlap and the principal direction vector of each rectangular frame are obtained; the average value of the overlap and the parallelism of the principal direction vector is used as the affinity. The skeleton of the cable mask is extracted to obtain the centerline path and calculate the centroid of the cable. The vector pointing from the centroid of the cable to the centroid of the rectangle of the terminal is used as the expected trend vector. All pixels within the rectangle of the wire nose where the centerline path is located are extracted and the centroid of the wire nose is calculated. The vector pointing from the centroid of the wire nose to the centroid of the rectangle of the terminal is used as the actual trend vector. The consistency of the trend is determined based on the cosine similarity between the expected trend vector and the actual trend vector. By integrating the distance, fit, and consistency of the cable lugs and terminals, the connection confidence between the cable and terminals is determined, satisfying the expression: In the formula, For cables With terminals The connection confidence score, For cables The centroid of the rectangular frame of the crimping lug and the terminal The Euclidean distance of the centroid of the rectangle. For cables The maximum Euclidean distance between the centroid of the rectangular frame of the wire lug and the centroids of the rectangular frames of all terminals. For cables crimping lugs and terminals The affinity between them For cables Pointer terminal The actual trend is consistent with the expected trend; For any cable, obtain the terminal corresponding to the maximum connection confidence score of all terminals and use it as its connection assignment; if the connection assignment is inconsistent with the standard connection assignment, trigger an alarm.

2. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The mask images of each cable, the rectangular boxes of the wire lugs, and the rectangular boxes of each terminal are obtained by recognizing the image to be detected using a pre-trained Mask R-CNN model.

3. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The method for obtaining the overlap includes: extracting the positions of all pixels within the rectangular frames of the crimping nose and the terminal, counting the number of pixels with the same position, and taking the ratio of the number of pixels with the same position to the number of pixels within the rectangular frame of the crimping nose as the overlap.

4. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The method for obtaining the principal direction vector of each rectangle is as follows: the position of all pixels in each rectangle is analyzed using the principal component analysis algorithm, and the first principal component direction vector is obtained as the principal direction vector.

5. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The parallelism of the principal direction vector is equal to the absolute value of the cosine similarity between the principal direction vector of the rectangular frame of the pressure nose and the principal direction vector of the rectangular frame of the terminal.

6. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The centerline path is obtained by applying a skeleton extraction algorithm to the cable mask image.

7. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The centroid of the cable is equal to the average position of all pixels along the centerline path of the cable; the centroid of the crimping lug is equal to the average position of all pixels within the rectangular frame of the crimping lug along the centerline path of the cable; the centroid of the rectangular frame of the terminal is equal to the average position of all pixels within its rectangular frame.

8. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The trend consistency satisfies the expression: ； In the formula, For cables Pointer terminal The actual trend is consistent with the expected trend; For cables Pointer terminal The expected trend vector; For cables Pointer terminal The actual trend vector; To take the cosine similarity function; It is a natural exponential function.

9. The method for quality inspection of complete switchgear manufacturing based on image analysis according to claim 1, characterized in that, The standard connection attribution method is as follows: based on the label of any cable, the standard connection attribution of the cable is extracted from the parameter document of the complete set of switchgear; wherein, the cable label is obtained by extracting the mask image of the cable through an optical character recognition network.

Images