A Visual Measurement Method for Transformer Tank Dimensions Based on Augmented Reality Technology

Abstract

The present invention provides a vision measurement method for the size of a transformer box based on augmented reality technology, belonging to the technical field of transformer box size measurement. The method includes obtaining a real-time image of the transformer box through an augmented reality device; using a pre-trained transformer box recognition model to recognize the real-time image of the transformer box to obtain the transformer box category; obtaining the boundary of the transformer box from the recognized transformer box in the form of point cloud; calculating the size of the transformer box according to the boundary of the transformer box; and judging the qualification information of the current transformer box based on the transformer box category and the transformer box size, so as to complete the real-time measurement of the transformer box size. The present invention uses an augmented reality device to accurately identify the transformer box from the environment and obtain size data, and this method can greatly bring convenience to the staff under the condition of ensuring measurement accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer box size measurement, and particularly relates to a vision measurement method for transformer box size based on augmented reality technology. Background Art

[0002] As an important part of the power system, the size of a transformer is often related to its capacity. Therefore, the measurement of the transformer size is of great significance. First, measuring the transformer box size can determine the size reference of the transformer box for subsequent assembly and maintenance of the transformer, ensuring that the sizes and installations of all components are correct. Second, correctly measuring the transformer box size is crucial for the performance evaluation and modeling of the transformer, and provides a reference for optimizing its performance and simulation analysis. In addition, during the transportation and installation of the transformer, it is necessary to measure the size information of the transformer box to facilitate the selection of subsequent transportation tools.

[0003] Currently, the mainstream transformer box measurements include the external dimension accumulation method, the laser projection method, etc. The external dimension accumulation method mainly measures the inner diameter of the transformer oil tank, the thickness of the tank wall, the dimensions of components such as the outer reinforcing iron, etc., and sums them up as the external dimension of the box. The laser projection method mainly uses a laser projector to project a light beam onto various parts of the box, and then measures using the received reflected light to obtain the size information of the box. The mainstream measurement methods are prone to error magnification due to multiple measurements and are easily affected by the environment, resulting in large fluctuations in measurement accuracy. Summary of the Invention

[0004] In view of this, the present invention aims to provide a vision measurement method for transformer box size based on augmented reality technology, which can accurately identify the transformer box from the environment and greatly facilitate the staff under the condition of ensuring measurement accuracy.

[0005] To achieve the above technical effects, the technical solution provided by the present invention is as follows:

[0006] A vision measurement method for transformer box size based on augmented reality technology includes the following steps:

[0007] Obtain a real-time transformer box image through an augmented reality device;

[0008] Use a pre-trained transformer box recognition model to recognize the real-time transformer box image to obtain the transformer box category;

[0009] Obtain the boundary of the transformer box from the recognized transformer box in the form of point cloud;

[0010] Calculate the transformer box size according to the boundary of the transformer box;

[0011] The qualified information of the current transformer box is judged based on the transformer box type and transformer box size, thereby completing the real-time measurement of the transformer box size.

[0012] Furthermore, the augmented reality device collects real-time transformer box images through a binocular camera.

[0013] Furthermore, the boundary of the transformer box is obtained from the identified transformer box by means of a point cloud, specifically including:

[0014] Obtaining a point cloud boundary of the transformer box from the identified transformer box by means of a point cloud method;

[0015] Perform the least squares fitting of the space curve on the point cloud boundary to obtain a rectangular model including the length, width and height of the transformer box;

[0016] The most central vertical line of the rectangular model is taken as the Z axis, and the lowest corner point of the vertical line is taken as the origin. The line segments in the rectangular model that do not have the origin as the endpoint are removed to obtain a simplified transformer box model.

[0017] Convert the endpoint coordinates of the simplified transformer box model from the image coordinate system to the world coordinate system;

[0018] Calculate the transformer box size based on the transformed endpoint coordinates.

[0019] Furthermore, the point cloud boundary of the transformer box is obtained from the identified transformer box by means of a point cloud method, specifically including:

[0020] The pixel point set of the identified transformer box is transformed to the plane xy by coordinate transformation to obtain a plane point set;

[0021] Traverse the plane point set and determine the minimum coordinate value x in the point set min ,y min and the maximum coordinate value x max ,y max ;

[0022] Generate a number of grids on the plane xy according to the set grid size, and make the grid area cover the plane point set area;

[0023] Determine a boundary grid from the grid area, the boundary grid is a real hole, and the eight adjacent grids of the real hole include at least one empty hole, and the real hole and the empty hole are grids containing data points and grids not containing data points respectively;

[0024] The boundary meshes are connected and refined to obtain the point cloud boundary of the transformer box.

[0025] Furthermore, the grid size is set according to the following formula:

[0026]

[0027] Wherein, L is the grid size and n is the number of grids set.

[0028] Furthermore, the number of grids in the x and y directions is determined according to the following formula:

[0029]

[0030]

[0031] Wherein, x_num and y_num are the number of grids in the x and y directions of the plane respectively, and L is the grid size.

[0032] Furthermore, the endpoint coordinates of the simplified transformer box model are converted from the image coordinate system to the world coordinate system, specifically according to the following formula:

[0033]

[0034] Wherein, (u, v) are the pixel coordinates of the pixel point to be converted in the image coordinate system, Z is the depth of the pixel point to be converted, dx and dy are the sizes of the pixel point to be converted in the X-axis and Y-axis directions respectively, (u0, v0) are the origin pixel coordinates of the image coordinate system, f is the focal length of the binocular camera, R and T are the rotation matrix and translation matrix of the binocular camera respectively, (X ω , Y ω , Z ω ) are the coordinates of the pixel point to be converted in three directions in the world coordinate system.

[0035] Furthermore, before the binocular camera captures the real-time transformer box image, it further includes:

[0036] Calibrating the binocular camera to establish a camera imaging geometric model and correct the distortion effect of the lens.

[0037] Furthermore, the transformer box recognition model is trained using the YOLO algorithm.

[0038] Furthermore, the augmented reality device further includes: a portable display device;

[0039] The portable display device is used to display the transformer box category, the transformer box size, and the qualification information of the current transformer box.

[0040] In summary, the present invention provides a visual measurement method for the size of a transformer box based on augmented reality technology, which includes obtaining a real-time image of the transformer box through an augmented reality device; using a pre-trained transformer box recognition model to recognize the real-time image of the transformer box to obtain the type of the transformer box; obtaining the boundary of the transformer box from the recognized transformer box in the form of point cloud; calculating the size of the transformer box according to the boundary of the transformer box; and judging the qualification information of the current transformer box based on the type and size of the transformer box, so as to complete the real-time measurement of the size of the transformer box. The present invention uses an augmented reality device to accurately identify the transformer box from the environment and obtain size data, and this method can greatly bring convenience to the staff under the condition of ensuring the measurement accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings according to these drawings without creative efforts.

[0042] Figure 1 It is a flowchart of a visual measurement method for the size of a transformer box based on augmented reality technology provided by an embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of measuring the size of a transformer box by an augmented reality device provided by an embodiment of the present invention;

[0044] Figure 3 It is a schematic diagram of the formation of a simplified model of a transformer box provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0046] Please refer to Figure 1 , an embodiment of the present invention provides a visual measurement method for the size of a transformer box based on augmented reality technology, including the following steps:

[0047] S11: Obtain a real-time image of the transformer box through an augmented reality device;

[0048] S12: using a pre-trained transformer box recognition model, recognize the real-time transformer box image to obtain the transformer box category;

[0049] S13: obtaining the boundary of the transformer box from the identified transformer box by means of a point cloud;

[0050] S14: Calculate the size of the transformer box according to the boundary of the transformer box;

[0051] S15: judging the qualified information of the current transformer box based on the transformer box type and the transformer box size, thereby completing the real-time measurement of the transformer box size.

[0052] It should be noted that augmented reality technology is a technology that cleverly integrates virtual information with the real world. It widely uses a variety of technical means such as multimedia, three-dimensional modeling, real-time tracking and registration, intelligent interaction, and sensing. It simulates computer-generated virtual information such as text, images, three-dimensional models, music, and videos, and applies them to the real world. The two types of information complement each other, thereby achieving "enhancement" of the real world.

[0053] The visual measurement method proposed in this embodiment applies augmented reality technology to transformer box size measurement, uses augmented reality device to obtain transformer box image, and uses it to determine subsequent box size. This method can complete the transformer box size measurement without setting various components during measurement.

[0054] When the transformer box image is collected and generated, the transformer box recognition model is used to identify the transformer box type in the transformer box image. Conventional transformer box types include European box transformers and American box transformers. Since the two box transformers have different structural designs, the size difference between the two is more obvious. The transformer box recognition model of this embodiment recognizes the transformer box type based on the size difference between the two. The transformer box recognition model can be obtained by machine learning methods using a large number of images of European box transformers and American box transformers.

[0055] In a specific implementation, the transformer box recognition model is configured on a background server, and the augmented reality device itself is configured with a communication device (which may be a 5G communication device) to transmit the image to the server background. The recognition model is used on the background server to identify and obtain the transformer box type and transformer box location (including its boundary box) information, so as to separate it from the external environment for subsequent box size measurement.

[0056] A point cloud refers to a set of points formed by obtaining the spatial coordinates of each sampling point on the surface of an object. By means of the point cloud method, a set of point cloud data of the transformer box can be obtained, so as to determine the distribution area of the transformer box image, and the boundary of the transformer box can be determined according to the size of its distribution area.

[0057] On the basis that the boundary of the transformer box can already be determined in the foregoing steps, calculating the size of the box according to the box boundary can adopt existing well-known technologies, which will not be elaborated here.

[0058] As mentioned above, the size differences between European-style and American-style box transformers are relatively large, and according to their structural designs, the box sizes of the two types of box transformers should meet the data requirements in the corresponding engineering manuals. Therefore, according to the type of box transformer and its size, it can be determined whether this type of box transformer is a qualified product.

[0059] The present invention provides a vision measurement method for the size of a transformer box based on augmented reality technology. This method uses an augmented reality device to accurately identify the transformer box from the environment and obtain size data, which can greatly bring convenience to the staff under the condition of ensuring measurement accuracy.

[0060] In a preferred embodiment of the present invention, the augmented reality device collects real-time images of the transformer box through a binocular camera. That is, the augmented reality device contains two symmetric high-definition micro cameras (binocular cameras). Additionally, an integrated communication chip can be attached to establish a stable network connection with a nearby signal base station, and can upload real-time images to the server backend at high speed through the 5G network. When the signal is poor, it can automatically switch to the 4G network for operation to ensure uninterrupted communication. The schematic diagram of the image for obtaining the size reading of the transformer during the measurement process is as Figure 2 shown. The high-definition micro camera has a wide-angle function. When the movement space of the user is limited, it can be manually adjusted to the wide-angle mode so that the object to be measured (transformer box) is completely mapped in the eyepiece screen. In addition, there is a high-brightness lighting module between the binocular cameras. This module integrates eight high-power small LED white light beads, which can be used by the staff to perform relevant operations under harsh natural lighting conditions such as foggy days and nights. The two micro cameras and the lighting module are integrated together and are centrally powered by a power supply board.

[0061] It can be understood that the augmented reality device collects images through a binocular camera based on the imaging principle of a binocular camera. A binocular camera consists of two cameras and can obtain the depth information of an object in three-dimensional space. The imaging principle of a binocular camera includes binocular disparity, stereo matching, binocular calibration, triangulation, etc.

[0062] Collect the images of the transformer tank based on the binocular imaging principle. In a preferred embodiment of the present invention, before using the binocular camera, it is also necessary to calibrate it to establish the camera imaging geometric model (obtain the internal and external parameters of the camera) and correct the distortion effect of the positive lens, so as to ensure that the captured images can well represent the objects in the real world. Specific methods can include the calibration plate (such as a checkerboard pattern) method or the feature point matching method (such as SIFT), etc. Based on the above principle, the feature point matching method is used to calibrate the camera. First, detect the feature points, such as prominent textures, corners, and edges, etc.; secondly, for each of the said feature points, calculate the corresponding descriptors, which are often vectors; then, perform the matching of the feature points. For the feature points in the test set and the training set, use the feature point matching algorithm to match them, such as using the nearest neighbor matching; finally, through the geometric relationship between the matched feature points, the internal and external parameters of the said camera can be calculated.

[0063] In a preferred embodiment of the present invention, the transformer tank recognition model can be trained using the YOLO algorithm. YOLO (You Only Look Once) is an object detection algorithm based on a deep neural network, used to identify and locate multiple objects in real time in images or videos. The main features of YOLO are high speed and relatively high accuracy, and it can achieve fast object detection in real-time scenarios.

[0064] Adopt the YOLO algorithm on the background server. Its detection speed is very high and it has good accuracy, which is very suitable for scenarios that require object detection. To meet the needs of actual projects, this method can collect multiple groups of training sets to train this YOLO-based object detection model. For each different type of transformer tank, a large amount of data sets need to be obtained. Currently, the mainstream box-type substations in China include American box-type substations and European box-type substations. Among them, American box-type substations are mostly used in low-density residential areas and less important buildings, while European box-type substations are mostly used in more important buildings such as multi-story residences. Through the domestic image library, a large number of rich image data sets can be obtained, including American box-type substations and European box-type substations. Then, use the YOLO algorithm to extract the features of the collected images to obtain the position and its bounding box of the target object. Then, perform the three-dimensional reconstruction of the tank. Based on the principle of multi-view geometry, use multi-view stereo matching to restore the three-dimensional shape of the tank. Finally, based on the restored three-dimensional shape, the YOLO algorithm predicts the category of its bounding box to determine the type attribution of the tank. In addition, the YOLO algorithm can improve the effect of three-dimensional reconstruction based on techniques such as optical flow estimation.

[0065] In a preferred embodiment of the present invention, after the binocular camera is calibrated, stereo correction, parallax calculation, and stereo matching are completed in sequence. Since the above operations are relatively fixed, they will not be described in this article. After that, this embodiment uses a point cloud method to obtain the boundary of the object. Among them, the boundary features of the transformer box can be well extracted through the grid division method. The specific steps are as follows:

[0066] S21: Obtaining a point cloud boundary of the transformer box from the identified transformer box in a point cloud manner.

[0067] As mentioned above, the point cloud data of the transformer box can be obtained through the point cloud method, and the point cloud boundary is the point cloud data located at the boundary determined from these point cloud data.

[0068] S22: Perform the least squares fitting of the space curve on the point cloud boundary to obtain a rectangular parallelepiped model including the length, width and height of the transformer box.

[0069] Least squares fitting is a mathematical approximation and optimization method that uses known data to obtain a straight line or curve so that the sum of the squares of the distances between it and the known data in the coordinate system is minimized. Using Excel's built-in functions, linear data analysis can be easily fitted. Through least squares fitting, a rectangular model containing the length, width and height of the transformer box can be obtained, which can be used to describe the size of the box.

[0070] S23: Taking the most centered vertical line of the rectangular parallelepiped model as the Z axis and the lowest corner point of the vertical line as the origin, eliminating the line segments in the rectangular parallelepiped model that do not have the origin as the endpoint, and obtaining a simplified transformer box model.

[0071] In order to facilitate subsequent size calculations, this step simplifies the rectangular parallelepiped model obtained in the previous step and retains the basic line segments that can describe the actual size of the transformer box.

[0072] S24: Convert the endpoint coordinates of the simplified transformer box model from the image coordinate system to the world coordinate system.

[0073] S25: Calculate the transformer box size based on the converted endpoint coordinates.

[0074] In a preferred embodiment of the present invention, obtaining the point cloud boundary of the transformer box from the identified transformer box by means of a point cloud method specifically includes:

[0075] S31: transforming the identified pixel point set of the transformer box onto the plane xy through coordinate transformation to obtain a plane point set;

[0076] S23: Traverse the plane point set and determine the minimum coordinate value x in the point set min ,ymin and the maximum coordinate value x max , y max ;

[0077] S33: Generate a number of grids on the plane xy according to the set grid size, and make the grid area cover the plane point set area;

[0078] S34: Determine the boundary grids from the grid area. The boundary grids are real holes, and at least one of the 8 adjacent grids of the real hole includes an empty hole. The real hole and the empty hole are grids containing data points and grids not containing data points, respectively;

[0079] S35: Connect the boundary grids and perform a refinement operation to obtain the point cloud boundary of the transformer box body.

[0080] In this embodiment, the grid division method is used to extract the boundary features of the transformer box body. For ease of understanding, a certain 3D point set t = {t, t1, t2,..., t n} in the imaginary space is transformed to the plane xy through a series of coordinate changes to obtain the point set r = {r1, r2,..., r n} in the plane. For each point in the above set, there is t i = (x i , y i , z i ) T , r i = (x i , y i ) T . The uniform grid method is adopted, that is, all points in the plane set r are traversed to obtain x max , x min , y max , y min and the point set is surrounded by a smallest bounding box.

[0081] According to the set grid size and quantity, the data point p i = (x i , y i ) T is placed into the corresponding grid cell (u, v), then the data point p i = (x i , y i ) T and the grid cell (u, v) establish a corresponding relationship, where u ∈ [0, x_num], v ∈ [0, y_num]. After all data are assigned grids, if the grid contains data points, it is called a "real hole", otherwise, the grid is called an "empty hole".

[0082] It is worth noting that if the grid is set to a small size, isolated "holes" will appear due to the uneven distribution of the point cloud in the plane, causing the surrounding grids to be misjudged as boundary grids. If such "holes" are filled, the occurrence of such misjudgments can be reduced.

[0083] The search for boundary meshes is mainly based on the following judgment: if for a "real hole", at least one of its 8 adjacent meshes includes the "empty hole", then the "real hole" can be regarded as a boundary mesh. The set of boundary meshes can be regarded as the "rough boundary" of the object. However, "empty holes" cannot be used as boundary meshes. After obtaining the "rough boundary", it needs to be refined because the "rough boundary" cannot meet the requirements of precise measurement well. The refinement operation can be performed as follows: connect each mesh in the "rough boundary" in sequence to form the initial boundary line, and then undergo a smoothing process to finally obtain the point cloud boundary of the transformer box.

[0084] Input the three-dimensional point cloud boundary and perform the least squares fitting of the space curve to obtain a rectangular model containing the length, width and height of the transformer box. Then, take the most centered vertical line as the Z axis and its lowest corner point as the origin to establish a three-dimensional space coordinate system, and remove the line segments that do not end at the origin to obtain the final simplified transformer box model. The specific process is as follows Figure 3 At this time, record the positions of the endpoints of the model (such as endpoints 1, 2, and 3).

[0085] In a preferred embodiment of the present invention, the grid size is determined according to the following formula:

[0086]

[0087] Where L is the grid size, n is the number of grids set, and is an integer.

[0088] The number of grids in the x and y directions is determined by the following formula:

[0089]

[0090]

[0091] Where x_num and y_num are the number of grids in the x and y directions of the plane, respectively, and L is the grid size.

[0092] In a preferred embodiment of the present invention, the endpoint coordinates of the simplified transformer box model are converted from the image coordinate system to the world coordinate system, specifically according to the following formula:

[0093]

[0094] Where (u, v) are the pixel coordinates of the pixel point to be converted in the image coordinate system, Z is the depth of the pixel point to be converted, dx and dy are the sizes of the pixel point to be converted in the X-axis and Y-axis directions respectively, (u0, v0) are the pixel coordinates of the origin of the image coordinate system, f is the focal length of the binocular camera, and R and T are the rotation matrix and translation matrix of the binocular camera respectively, and (X ω , Y ω , Z ω ) are the coordinates of the pixel point to be converted in three directions in the world coordinate system.

[0095] Taking Figure 3 endpoint 1 as an example, assuming its pixel coordinates in the image coordinate system are (u, v), then its coordinates in the world coordinate system after conversion according to the above formula are P(X ω , Y ω , Z ω ). For Figure 3 the 12-segment shown, its length L 12 can be obtained by the following formula for the length between two points in space:

[0096]

[0097] Since point 1 is selected as the origin of the coordinate system, there is Similarly, it is stipulated that the length of the transformer tank is max{L 12 , L 13}, the width of the transformer tank is min{L 12 , L 13}, and the height of the transformer tank is L 14 . At this time, the size of the transformer tank can be determined, that is, {length * width * height} = {max{L 12 , L 13} * min{L 12 , L 13} * L 14}.

[0098] In a preferred embodiment of the present invention, the augmented reality device further includes: a portable display device;

[0099] The portable display device is used to display the transformer tank category, the transformer tank size, and the qualification information of the current transformer tank. In a specific embodiment, the portable display device uses wearable AR glasses. When the transformer tank size data is acquired and transmitted to the background for the type judgment of the tank. The background has loaded the data in the transformer tank engineering manual and judges whether the tank belongs to qualified products. Finally, the judged tank type, the size data of the tank, and the information on whether it belongs to qualified products are transmitted to the wearable AR glasses through waveguide display technology.

[0100] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A transformer box size visual measurement method based on augmented reality technology, characterized in that: The steps include: Obtain real-time transformer box images through augmented reality devices; Using a pre-trained transformer box recognition model, the real-time transformer box image is recognized to obtain the transformer box category; Obtaining the boundary of the transformer box from the identified transformer box by means of a point cloud; Calculating the size of the transformer box according to the boundary of the transformer box; Based on the transformer box type and the transformer box size, the qualified information of the current transformer box is judged, thereby completing the real-time measurement of the transformer box size; the augmented reality device collects the real-time transformer box image through the binocular camera; The boundary of the transformer box is obtained from the identified transformer box by point cloud method, including: Obtaining a point cloud boundary of the transformer box from the identified transformer box by using the point cloud method; Performing least square fitting of a spatial curve on the boundary of the point cloud to obtain a rectangular parallelepiped model including the length, width and height of the transformer box; The most central vertical line of the rectangular parallelepiped model is taken as the Z axis, and the lowest corner point of the vertical line is taken as the origin, and the line segments in the rectangular parallelepiped model that do not have the origin as the endpoint are removed to obtain a simplified transformer box model; Converting the endpoint coordinates of the simplified transformer box model from the image coordinate system to the world coordinate system; Calculate the transformer box size based on the transformed endpoint coordinates.

2. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 1 is characterized in that: Obtaining the point cloud boundary of the transformer box from the identified transformer box by the point cloud method specifically includes: The pixel point set of the identified transformer box is transformed to the plane xy by coordinate transformation to obtain a plane point set; Traverse the plane point set and determine the minimum coordinate value x in the point set min ,y min and the maximum coordinate value x max ,y max ; Generate a number of grids on the plane xy according to a set grid size, and make the grid area cover the plane point set area; Determine a boundary grid from the grid area, wherein the boundary grid is a real hole, and eight adjacent grids of the real hole include at least one empty hole, and the real hole and the empty hole are grids containing data points and grids not containing data points, respectively; The boundary meshes are connected and refined to obtain the point cloud boundary of the transformer box.

3. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 2 is characterized in that: The grid size is determined according to the following formula: Where L is the grid size and n is the number of grids set.

4. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 2 is characterized in that: The number of grids in the x and y directions is determined by the following formula: Where x_num and y_num are the number of grids in the x and y directions of the plane, respectively, and L is the grid size.

5. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 1 is characterized in that: The endpoint coordinates of the simplified transformer box model are converted from the image coordinate system to the world coordinate system, specifically according to the following formula: Where, (u, v) is the pixel coordinate of the pixel to be converted in the image coordinate system, Z is the depth of the pixel to be converted, dx, dy are the dimensions of the pixel to be converted in the X-axis and Y-axis directions, (u0, v0) is the pixel coordinate of the origin of the image coordinate system, f is the focal length of the binocular camera, R, T are the rotation matrix and translation matrix of the binocular camera, (X ω ,Y ω ,Z ω ) are the coordinates of the pixel to be converted in three directions in the world coordinate system.

6. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 1 is characterized in that: Before the binocular camera collects the real-time transformer box image, it also includes: The binocular camera is calibrated to establish the camera imaging geometry model and correct the lens distortion effect.

7. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 1 is characterized in that: The transformer box recognition model is trained using the YOLO algorithm.

8. The method for visually measuring transformer box dimensions based on augmented reality technology according to claim 1 is characterized in that: The augmented reality device further includes: a portable display device; The portable display device is used to display the transformer box type, the transformer box size and the qualification information of the current transformer box.

Images