Power grid operation and maintenance method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a power grid operation and maintenance method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps that a first image and a second image of a target area are acquired based on a visual identification system, the visual identification system comprises a first camera and a second camera which are fixedly arranged on the same horizontal datum line at a preset distance, the parameters of the first camera and the second camera are the same, and the optical axes of the first camera and the second camera are parallel to each other; constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image; and inputting the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and executing a power grid operation and maintenance task based on the trigger area. By adopting the method, the operation and maintenance efficiency of the power grid can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of smart grid technology, and in particular to a grid operation and maintenance method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] RPA (Robotic Process Automation) is a new type of automation technology, namely robotic process automation. In the field of power grid operation and maintenance technology, RPA is usually used to perform high-frequency and repetitive tasks.

[0003] In the existing technology, RPA can monitor the triggering operations of users in the user interaction interface and learn from them to achieve automated operation and maintenance of the power grid.

[0004] However, due to the poor accuracy of RPA in identifying the trigger operation in the trigger area corresponding to the user interaction interface in the prior art, the efficiency of power grid operation and maintenance is low. Summary of the invention

[0005] Based on this, it is necessary to provide a power grid operation and maintenance method, device, computer equipment, computer-readable storage medium and computer program product that can improve the efficiency of power grid operation and maintenance in response to the above technical problems.

[0006] In a first aspect, the present application provides a power grid operation and maintenance method, comprising:

[0007] Acquiring a first image and a second image of a target area based on a visual recognition system, the visual recognition system comprising a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, the first camera and the second camera having the same parameters and optical axes parallel to each other;

[0008] Constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0009] The target image is input into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and a power grid operation and maintenance task is performed based on the trigger area.

[0010] In one of the embodiments, the correction matrix is ​​constructed based on the relative transformation matrix of the first camera and the second camera, including: obtaining the relative position relationship between the first camera and the second camera, and determining a rotation transformation matrix and a translation transformation matrix according to the relative position relationship; determining the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determining the correction matrix according to the relative transformation matrix.

[0011] In one of the embodiments, the correction matrix is ​​determined according to the relative transformation matrix, including: obtaining the extrinsic matrix of the first camera and the second camera, and constructing a calibration model of the visual recognition system; determining the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic matrix and the calibration model; and determining the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0012] In one of the embodiments, the training method of the trigger area prediction model includes: training the initial trigger area prediction model based on training data to obtain a candidate trigger area prediction model, and determining the prediction error of the candidate trigger area prediction model; when the prediction error is less than a preset error, determining the candidate trigger area prediction model as the trigger area prediction model; when the prediction error is greater than the preset error, optimizing the candidate trigger area prediction model based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0013] In one of the embodiments, determining the prediction error of the candidate trigger area prediction model includes: acquiring a first image and a second image of a training area based on a visual recognition system; and correcting the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image; inputting the training image into the candidate trigger area prediction model to obtain a predicted trigger area output by the candidate trigger area prediction model; determining a trigger distribution line based on a real trigger area, and determining a projected trigger area based on the trigger distribution line and the predicted trigger area; and determining the prediction error of the candidate trigger area prediction model based on the projected trigger area and the predicted trigger area.

[0014] In one of the embodiments, the method of determining a trigger distribution line based on a real trigger area includes: determining a candidate trigger distribution line based on the real trigger area; optimizing the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0015] In a second aspect, the present application also provides a power grid operation and maintenance device, including:

[0016] an acquisition module, configured to acquire a first image and a second image of a target area based on a visual recognition system, wherein the visual recognition system comprises a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, wherein the first camera and the second camera have the same parameters and their optical axes are parallel to each other;

[0017] A first execution module is used to construct a correction matrix based on the relative transformation matrix of the first camera and the second camera, and perform correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0018] The second execution module is used to input the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and execute power grid operation and maintenance tasks based on the trigger area.

[0019] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in any embodiment of the first aspect when executing the computer program.

[0020] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any one of the embodiments of the first aspect above.

[0021] In a fifth aspect, the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements the method described in any embodiment of the first aspect above.

[0022] The above-mentioned power grid operation and maintenance method, device, computer equipment, computer-readable storage medium and computer program product first obtain the first image and the second image of the target area based on the visual recognition system, the visual recognition system includes a first camera and a second camera fixedly set at a preset distance on the same horizontal reference line, the parameters of the first camera and the second camera are the same, and the optical axes are parallel to each other; then a correction matrix is ​​constructed based on the relative transformation matrix of the first camera and the second camera, and the first image and the second image are corrected based on the correction matrix to obtain the target image; then, the target image is input into a pre-trained trigger area prediction model to obtain the trigger area in the target area output by the trigger area prediction model, and the power grid operation and maintenance task is performed based on the trigger area. The power grid operation and maintenance method provided by the present application performs correction processing on the first image and the second image of the target area obtained by the visual recognition system based on the correction matrix, and then inputs the corrected image into the trigger area prediction model to obtain the trigger area in the target area, which can effectively improve the accuracy of identifying the trigger area, thereby effectively improving the efficiency of power grid operation and maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A schematic diagram of a flow chart of a power grid operation and maintenance method in one embodiment;

[0025] Figure 2 A schematic flow chart of a method for constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera in one embodiment;

[0026] Figure 3 is a flow chart of a method for determining the correction matrix according to the relative transformation matrix in one embodiment;

[0027] Figure 4 A schematic diagram of a flow chart of a method for training a trigger region prediction model in one embodiment;

[0028] Figure 5 A schematic flow chart of a method for determining a prediction error of the candidate trigger region prediction model in one embodiment;

[0029] Figure 6 A schematic flow chart of a method for determining a trigger distribution line based on a real trigger area in one embodiment;

[0030] Figure 7 A schematic diagram of a flow chart of a power grid operation and maintenance method in another embodiment;

[0031] Figure 8 is a structural block diagram of a power grid operation and maintenance device in one embodiment;

[0032] Fig. 9 is an internal structure diagram of a computer device in one embodiment;

[0033] Fig.10 is an internal structure diagram of a computer device in another embodiment;

[0034] Fig.11 A schematic diagram of a trigger area prediction model framework in an embodiment;

[0035] Fig.12 A schematic diagram of a visual recognition system for describing an epipolar geometry model in one embodiment;

[0036] Fig.13 Schematic diagram of affine projection in one embodiment. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0038] RPA (Robotic Process Automation) is a new type of automation technology, namely robotic process automation. In the field of power grid operation and maintenance technology, RPA is usually used to perform high-frequency and repetitive tasks.

[0039] In the existing technology, RPA can monitor the triggering operations of users in the user interaction interface and learn from them to achieve automated operation and maintenance of the power grid.

[0040] However, due to the poor accuracy of RPA in identifying the trigger operation in the trigger area corresponding to the user interaction interface in the prior art, the efficiency of power grid operation and maintenance is low.

[0041] In view of this, the present application provides a power grid operation and maintenance method, which can improve the accuracy of RPA identification of trigger operations in the trigger area corresponding to the user interaction interface, thereby effectively improving the efficiency of power grid operation and maintenance.

[0042] The power grid operation and maintenance method provided in the present application may be executed by a computer device, which may be a server or a terminal.

[0043] In an exemplary embodiment, Figure 1 As shown, a power grid operation and maintenance method is provided, the method comprising the following steps:

[0044] Step 101: Acquire a first image and a second image of a target area based on a visual recognition system.

[0045] The visual recognition system comprises a first camera and a second camera which are fixedly arranged at a preset distance on the same horizontal reference line. The parameters of the first camera and the second camera are the same, and the optical axes are parallel to each other.

[0046] Optionally, the preset distance can be preset by a technician according to actual needs. The first camera and the second camera of the visual recognition system are used to simulate the process of the user's binocular perception of the target area.

[0047] This parameter refers to the camera parameter, that is, the parameter that affects the camera's imaging quality and camera characteristics. The parameter can be focal length, aperture, shutter speed, ISO sensitivity, field of view, white balance, and exposure compensation.

[0048] In some exemplary embodiments, the visual recognition system may be constructed based on binocular stereoscopic vision principle. The first camera and the second camera of the visual recognition system may respectively capture images of the target area at different viewing angles.

[0049] Step 102: construct a correction matrix based on the relative transformation matrix of the first camera and the second camera, and perform correction processing on the first image and the second image based on the correction matrix to obtain a target image.

[0050] The relative transformation matrix is ​​used to describe the spatial relationship between the two cameras, including position (translation) and direction (rotation). The relative transformation matrix is ​​usually composed of a rotation matrix and a translation vector.

[0051] In one possible implementation, corresponding feature points may be first found in the first image and the second image captured by the first camera and the second camera, and then a basic matrix or a homography matrix may be determined based on the features, and then rotation and translation information may be extracted from the basic matrix or the homography matrix to construct the relative transformation matrix.

[0052] Furthermore, the intrinsic parameter matrix of the first camera and the second camera is obtained, and the correction matrix is ​​constructed according to the intrinsic parameter matrix and the relative transformation matrix. The first image and the second image are corrected based on the correction matrix to obtain the target image.

[0053] In another possible implementation manner, the extrinsic parameter matrices of the first camera and the second camera may be first obtained, and the relative transformation matrix of the first camera and the second camera may be determined according to the extrinsic parameter matrices.

[0054] Furthermore, the intrinsic parameter matrix of the first camera and the second camera is obtained, and the correction matrix is ​​constructed according to the intrinsic parameter matrix and the relative transformation matrix. The first image and the second image are corrected based on the correction matrix to obtain the target image.

[0055] Step 103: input the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and perform power grid operation and maintenance tasks based on the trigger area.

[0056] Optionally, the trigger area prediction model may be a neural network model pre-trained by a technician.

[0057] In some exemplary embodiments, the neural network model may be a model based on the AlexNet network. The trigger region prediction model may be composed of five convolutional layers and three fully connected layers, and its basic framework may be as follows: Fig.11As shown in the figure. Specifically, except for the first convolution layer with a convolution kernel size of 5×5, the convolution kernels of the remaining convolution layers all use a 3×3 V convolution kernel with a convolution step of 2 pixels. All convolution layers are followed by a ReLu layer to improve the nonlinear expression ability of the model; the first, second, and fifth convolution layers are followed by a pooling layer, which uses a maximum pooling operation with a step size of 2. The last fully connected layer is the output layer. The trigger area prediction model can output the 5-dimensional parameters of the trigger area, which are: the center position, width, height, and rotation angle of the trigger area.

[0058] The above-mentioned power grid operation and maintenance method first obtains the first image and the second image of the target area based on the visual recognition system, and the visual recognition system includes a first camera and a second camera fixedly set at a preset distance on the same horizontal reference line, and the parameters of the first camera and the second camera are the same, and the optical axes are parallel to each other; then, a correction matrix is ​​constructed based on the relative transformation matrix of the first camera and the second camera, and the first image and the second image are corrected based on the correction matrix to obtain the target image; then, the target image is input into a pre-trained trigger area prediction model to obtain the trigger area in the target area output by the trigger area prediction model, and the power grid operation and maintenance task is performed based on the trigger area. The power grid operation and maintenance method provided in the present application performs correction processing on the first image and the second image of the target area obtained by the visual recognition system based on the correction matrix, and then inputs the corrected image into the trigger area prediction model to obtain the trigger area in the target area, which can effectively improve the accuracy of identifying the trigger area, thereby effectively improving the efficiency of power grid operation and maintenance.

[0059] In an optional embodiment of the present application, as described above, corresponding feature points can be first found in the first image and the second image taken by the first camera and the second camera, and then a basic matrix or homography matrix is ​​determined based on the features, and then rotation information and translation information are extracted from the basic matrix or homography matrix to construct the relative transformation matrix.

[0060] For example, in an ideal state, the optical axes of the first camera and the second camera are parallel to each other. However, in actual applications, during the installation or production of the two cameras of the visual recognition system, it is difficult to ensure that the optical axes of the first camera and the second camera are perpendicular to the baseline and completely parallel to each other at the same time. The optical axes of the first camera and the second camera will intersect somewhere in space. Therefore, the epipolar geometry model can be used to describe the imaging process of the visual recognition system. Specifically, Fig.12 shown.

[0061] In some exemplary embodiments, the epipolar geometry model can describe the spatial geometric relationship between the projection positions of the same spatial target point on the left and right imaging planes of the first camera and the second camera. Fig.13As shown in the figure, O1 and O2 are the optical centers of the first camera and the second camera respectively. Together with the midpoint P in space, they determine the epipolar plane. The epipolar plane intersects with the left and right imaging planes of the first camera and the second camera at the epipolar lines l1 and l2 respectively, where l1 is regarded as the projection of the spatial point P onto the second camera on the left imaging plane, and l2 is the projection of the first camera on the right imaging plane. The intersection of the epipolar line and the baseline is called the pole, which is represented by e1 and e2. e1 is the projection point of the optical center O2 of the second camera on the first camera plane. Similarly, e2 is the projection point of the optical center O2 of the first camera on the right imaging plane. The intersection points of the epipolar line with the spatial point P and the line connecting the optical centers of the first camera and the second camera are the projection points P1 and P2 of the spatial point P on the two imaging planes. In a binocular vision system, the projection of any P in space on the two imaging planes must satisfy the above-mentioned epipolar geometry relationship, that is, the matching point corresponding to the projection point in the left imaging plane must be located on the epipolar line on the right imaging plane. Therefore, the relative position between the two cameras is solved through the epipolar geometry constraint relationship, and then the point pairs that satisfy the constraint relationship are found on the two imaging planes, which are the feature points mentioned above.

[0062] Furthermore, in the epipolar geometry model, the projection point on the right camera imaging plane is transformed to the left imaging plane through the rotation process R and the translation process T, that is, Among them, the translation vector represents the translation process of the optical center O1 to O2 in the polar plane. Since the three coplanar vectors satisfy the constraint that the mixed product is 0, the three vectors on the polar plane satisfy the following relationship: The outer product between vectors is transformed into a linear operation form of matrix-vector product by introducing an antisymmetric matrix, and then By combining them, we can construct the relative posture relationship between the two projection points under the left and right perspectives: ,in, represents the antisymmetric matrix and defines the essential matrix Describe the relative position between the first camera and the second camera. Substitute the intrinsic parameter matrix K of the first camera and the second camera into the above formula, continue to convert the physical coordinates under the camera coordinates to pixel coordinates, and use p1 and p2 to represent the pixel coordinates of the projection point on the left and right images respectively, then: , where K1 and K2 are the intrinsic parameter matrices of the first camera and the second camera respectively. The essential matrix E selects multiple sets of correspondences through the camera calibration process for linear solution, so that the feature points are found through the transformation relationship between the two projected images.

[0063] In an exemplary embodiment, Figure 2 As shown, the correction matrix is ​​constructed based on the relative transformation matrix of the first camera and the second camera, including the following steps:

[0064] Step 201: Obtain a relative position relationship between the first camera and the second camera, and determine a rotation transformation matrix and a translation transformation matrix according to the relative position relationship.

[0065] Step 202: determine the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determine the correction matrix according to the relative transformation matrix.

[0066] In a possible implementation, the relative position relationship between the first camera and the second camera may be determined based on the epipolar geometry model and the method described above, and a rotation transformation matrix and a translation transformation matrix may be determined according to the relative position relationship.

[0067] In an exemplary embodiment, Figure 3 As shown, determining the correction matrix according to the relative transformation matrix includes the following steps:

[0068] Step 301: Obtain the external parameter matrices of the first camera and the second camera, and construct a calibration model of the visual recognition system;

[0069] Step 302: determining a relative position relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic parameter matrix and the calibration model;

[0070] Step 303: determine the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0071] For example, the essence of correction is to make the imaging plane parallel to the fixed baseline of the camera. By performing projection transformation on the left and right images, the re-projected image points satisfy the property of parallel alignment of the epipolar lines in the corrected imaging plane. Therefore, the poles will be mapped to infinity respectively, so that the two images can achieve ideal imaging effects.

[0072] In some exemplary embodiments, the image points on the new imaging plane and the original imaging plane satisfy the following equation: , and are the corrected image points, and It describes the mapping relationship between the left and right image planes of the first camera and the second camera respectively. The corrected image points in the above formula are brought into the relative pose relationship formula to obtain the constraints satisfied by the original corresponding matching image points and the homography transformation matrix: According to the parallel alignment principle, the poles of the corrected imaging plane will be mapped to infinity, and then the solution of the essential matrix will be derived. Then, the constraint conditions such as epipolar coincidence and image loss minimization are used to complete the solution process of the homography matrix.

[0073] Furthermore, the present application uses the Bouguet epipolar correction method encapsulated in OpenCV to perform stereo correction on the first image and the second image. The Bouguet method uses the relative transformation matrix between the first camera and the second camera to construct the correction matrix. First, the transformation process corresponding to the first camera and the second camera is solved. The first camera on the left side of the binocular vision system is used as a reference. The rotation transformation matrix R and the translation transformation matrix T are used to represent the rotation and translation movement of the second camera on the right side of the binocular vision system relative to the first camera on the left side. Therefore, the three-dimensional space target point Image point on the right camera imaging plane and Satisfies the transformation relationship in the following formula: . Target point During the transformation process of projecting onto the imaging planes of the first camera and the second camera, the coordinate systems of the first camera and the second camera also undergo rotation and translation transformation processes respectively. and Respectively represent the external parameter matrices of the first camera and the second camera during the projection of point P, then point P and image point and They satisfy the following transformation relationship respectively: , the parameter matrices of the first camera and the second camera are obtained by the camera calibration process respectively. The calibration process of the visual recognition system is the same as that of the monocular camera, except that the minimum reprojection error of the spatial point projection in the left and right views needs to be considered simultaneously in the parameter optimization process. Therefore, the parameter calibration process of the visual recognition system can be expressed as the mathematical model in the following formula:

[0074] , the obtained external parameter matrices of the first camera and the second camera are brought into the above two equations to solve the relative pose relationship between the two camera coordinate systems: .

[0075] Furthermore, the correction matrix of the imaging planes of the first camera and the second camera is constructed. During the correction process, the left and right imaging planes are first rotated by half of the relative matrix R. The purpose of this is to move the imaging planes of the first camera and the second camera to a position parallel to each other, but the epipolar lines of the two planes have not been aligned at this time: Next, the relative rotation matrix R of the first camera and the second camera is decomposed into and And create the transformation matrix : , in order to achieve parallel alignment of the epipolar lines in the two image planes, the baselines between the two imaging planes and the optical centers of the connecting cameras should also remain parallel, so the poles should be at infinity, so the transformation matrix The first vector in The direction should be consistent with the translation vector between the optical centers of the two cameras: , where T is a mutually orthogonal translation vector, Along the image direction and with vertical, so satisfy: , and and Orthogonal, then: In this way, the correction matrices of the first camera and the second camera are obtained respectively and : .

[0076] In an exemplary embodiment, Figure 4 As shown, the training method of the trigger area prediction model includes the following steps:

[0077] Step 401: training the initial trigger region prediction model based on the training data to obtain a candidate trigger region prediction model, and determining a prediction error of the candidate trigger region prediction model;

[0078] Step 402: if the prediction error is less than a preset error, determine the candidate trigger region prediction model as the trigger region prediction model;

[0079] Step 403: When the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0080] In some exemplary embodiments, in order to obtain training data, the positive and negative samples of the trigger area in the data set can be rotated to the horizontal direction to form a regular sample without direction, and scaled to 48×48×3, and the remaining area is padded with 0. The convolutional neural network model is used to classify the trigger distribution line. In order to increase the number of training images and improve the robustness of the model, the training image is randomly translated in the X and Y axis directions, and the maximum translation distance does not exceed 3 pixels. The initial learning rate of the initial trigger area prediction model is set to 0.001. After every 7 cycles (a total of about 40 cycles in the training process), the learning rate is reduced by 90%, and the weight is updated using the RMSprop method.

[0081] Furthermore, after the initial trigger area prediction model is trained based on the training data, the prediction error of the candidate trigger area prediction model is determined. If the prediction error is less than the preset error, the candidate trigger area prediction model is determined as the trigger area prediction model; if the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0082] In an exemplary embodiment, Figure 5 As shown, determining the prediction error of the candidate trigger area prediction model includes the following steps:

[0083] Step 501: acquiring a first image and a second image of a training area based on a visual recognition system; and performing correction processing on the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image;

[0084] Step 502: input the training image into the candidate trigger region prediction model to obtain a predicted trigger region output by the candidate trigger region prediction model;

[0085] Step 503: determining a trigger distribution line based on the actual trigger area, and determining a projected trigger area according to the trigger distribution line and the predicted trigger area;

[0086] Step 504: Determine a prediction error of the candidate trigger area prediction model according to the projected trigger area and the predicted trigger area.

[0087] In some exemplary embodiments, a first image and a second image of a training area are first acquired based on a visual recognition system, and the first image and the second image are corrected based on correction matrices of the first camera and the second camera to obtain a training image. The training image is then input into the candidate trigger area prediction model to obtain a predicted trigger area output by the candidate trigger area prediction model.

[0088] Furthermore, a trigger distribution line is determined based on the actual trigger area, and a projected trigger area is determined based on the trigger distribution line and the predicted trigger area. Specifically, the geometric center of the predicted trigger area is projected to each trigger distribution line, and the Euclidean distance between it and each projection point is calculated. The trigger distribution line corresponding to the minimum distance is the reference line. However, the point-line distance can only reflect the distance between the center point of the predicted trigger area and the line where the trigger distribution line is located, rather than the line segment where the trigger distribution line is located. Therefore, it cannot fully reflect the actual correspondence between the predicted trigger area and the trigger distribution line. Therefore, the present application calculates the Euclidean distance between the center point of the predicted trigger area and the two end points of each trigger distribution line, and uses the average as the final point-line distance.

[0089] Furthermore, taking the selected trigger distribution line as a reference, the predicted trigger area is affine transformed to obtain the projected trigger area. The affine transformation is defined as: ,in, is the position coordinate of each point in the predicted trigger area, is the position coordinate of the corresponding point in the projected trigger area. To determine the affine transformation matrix, it is necessary to estimate the rotation angle of the predicted trigger area. and translation vector The values ​​are as follows:

[0090] Rotation angle estimation: Since the slope k of the trigger distribution line is known, the angle between the projected trigger area and the horizontal direction can be expressed as , then the rotation angle of the affine transformation is is the angle between the predicted trigger area and the horizontal direction. Translation vector estimation: Taking the geometric center of the predicted trigger area and the projected trigger area as the reference, calculate the translation vector between the two, expressed as , the center of the projected trigger area is located on the trigger distribution line and is 100 meters away from the center of the predicted trigger area. Specifically, first calculate Projection point on the trigger distribution line If the projection point is on the trigger distribution line, the current projection point can be used as the center of the projection trigger area; otherwise, the distance between the two end points of the trigger distribution line needs to be calculated. The endpoint corresponding to the minimum distance is taken as the center of the projection trigger area, such as Fig.13 shown.

[0091] In an optional embodiment of the present application, in order to measure the deviation between the predicted trigger area and the actual trigger area, two evaluation methods, point indicators and rectangle indicators, can be used. Since point indicators are generally used to evaluate trigger points, and rectangle indicators comprehensively consider the differences in regional direction, spatial position and size, they are suitable for the evaluation of predicted trigger areas. As mentioned above, since the target area only contains a few triggerable areas, directly comparing the predicted trigger area with it cannot fully reflect the triggerable probability of the area. Therefore, a trigger distribution straight line is introduced to characterize the continuous distribution of the actual trigger area, and an affine transformation is performed on the predicted trigger area based on it. By calculating the rectangle indicators of the predicted trigger area and the projection area, a more reasonable predicted trigger area evaluation can be achieved. The evaluation indicators of the predicted trigger area are as follows: ,in, They represent the predicted trigger area and the projected trigger area respectively. is the corresponding direction.

[0092] Further, ,in, Respectively represent the position coordinates of the two end points of the trigger distribution line. , .

[0093] In an exemplary embodiment, Figure 6 As shown, the method for determining the trigger distribution straight line based on the real trigger area includes the following steps:

[0094] Step 601, determining a candidate trigger distribution line based on the real trigger area;

[0095] Step 602: Optimize the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0096] In some exemplary embodiments, most of the current trigger area detection methods use rectangular indicators to evaluate the quality of the predicted area. The existing evaluation indicators cannot fully reflect the actual trigger probability of the predicted area. In order to overcome the shortcomings of the above evaluation method, the predicted area is affine transformed based on the trigger distribution line to obtain the projection area, and the predicted area is evaluated more reasonably by calculating the ratio and angle difference between the predicted area and the projection area. The closer the distance between the predicted area and the trigger distribution line and the smaller the angle difference with the projection area, the greater the probability of its triggerability.

[0097] The present application proposes a method for automatically generating trigger distribution lines from discrete real trigger areas. The most direct method to generate trigger distribution lines is to connect the geometric centers of each real trigger area respectively, but this will result in a large number of invalid connections. Therefore, it is considered to connect the geometric centers of the trigger areas that overlap each other, but invalid lines still exist at this time. The invalid lines in the trigger area are mainly reflected in two aspects: one is redundant lines, which are mainly manifested in that there are multiple trigger distribution lines between the same pair of starting points, especially when multiple areas overlap each other. This leads to multiple similar projection areas in the same area during the projection of the predicted area, which has limited improvement on the accuracy of the algorithm, but reduces the operating efficiency. The second is discontinuous lines, which are mainly manifested in that the trigger distribution line is not completely located on the object, but intersects with the background part. This leads to a large overlap between the projection area and the object during the projection of the predicted area, resulting in detection failure. Therefore, the candidate trigger distribution line can be optimized based on the redundant line optimization algorithm and the discontinuous line optimization algorithm to obtain the trigger distribution line.

[0098] In an exemplary embodiment, Figure 7 As shown, another power grid operation and maintenance method is provided, the method comprising the following steps:

[0099] Step 701: acquiring a first image and a second image of a target area based on a visual recognition system, wherein the visual recognition system includes a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, wherein the first camera and the second camera have the same parameters and their optical axes are parallel to each other;

[0100] Step 702: Obtain the relative position relationship between the first camera and the second camera, and determine a rotation transformation matrix and a translation transformation matrix according to the relative position relationship; determine the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix;

[0101] Step 703: Obtain the external parameter matrices of the first camera and the second camera, and construct a calibration model of the visual recognition system; determine the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the external parameter matrix and the calibration model; determine the correction matrix according to the rotation transformation matrix and the relative posture relationship;

[0102] Step 704: correct the first image and the second image based on the correction matrix to obtain a target image; input the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and perform power grid operation and maintenance tasks based on the trigger area.

[0103] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0104] Based on the same inventive concept, the embodiment of the present application also provides a power grid operation and maintenance device for implementing the power grid operation and maintenance method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more power grid operation and maintenance device embodiments provided below can refer to the limitations of the power grid operation and maintenance method above, and will not be repeated here.

[0105] In an exemplary embodiment, Figure 8As shown, a power grid operation and maintenance device is provided, including: an acquisition module 801, a first execution module 802 and a second execution module 803, wherein:

[0106] An acquisition module 801 is used to acquire a first image and a second image of a target area based on a visual recognition system, wherein the visual recognition system includes a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, wherein the first camera and the second camera have the same parameters and their optical axes are parallel to each other;

[0107] A first execution module 802 is used to construct a correction matrix based on the relative transformation matrix of the first camera and the second camera, and perform correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0108] The second execution module 803 is used to input the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and execute a power grid operation and maintenance task based on the trigger area.

[0109] In one embodiment, the first execution module 802 is specifically used to obtain the relative position relationship between the first camera and the second camera, and determine the rotation transformation matrix and the translation transformation matrix according to the relative position relationship; determine the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determine the correction matrix according to the relative transformation matrix.

[0110] In one embodiment, the first execution module 802 is specifically used to obtain the external parameter matrices of the first camera and the second camera, and construct a calibration model of the visual recognition system; determine the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the external parameter matrix and the calibration model; determine the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0111] In one embodiment, the second execution module 803 is also used to train the initial trigger area prediction model based on the training data to obtain a candidate trigger area prediction model, and determine the prediction error of the candidate trigger area prediction model; when the prediction error is less than a preset error, the candidate trigger area prediction model is determined as the trigger area prediction model; when the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0112] In one embodiment, the second execution module 803 is specifically used to acquire the first image and the second image of the training area based on the visual recognition system; and perform correction processing on the first image and the second image based on the correction matrix of the first camera and the second camera to obtain the training image; input the training image into the candidate trigger area prediction model to obtain the predicted trigger area output by the candidate trigger area prediction model; determine the trigger distribution line based on the real trigger area, and determine the projected trigger area according to the trigger distribution line and the predicted trigger area; determine the prediction error of the candidate trigger area prediction model according to the projected trigger area and the predicted trigger area.

[0113] In one embodiment, the second execution module 803 is specifically used to determine a candidate trigger distribution line based on the real trigger area; optimize the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0114] Each module in the above-mentioned power grid operation and maintenance device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0115] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a power grid operation and maintenance method is implemented.

[0116] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.10As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a power grid operation and maintenance method is implemented.

[0117] Those skilled in the art will understand that Fig. 9 and Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0118] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0119] Acquiring a first image and a second image of a target area based on a visual recognition system, the visual recognition system comprising a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, the first camera and the second camera having the same parameters and optical axes parallel to each other;

[0120] Constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0121] The target image is input into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and a power grid operation and maintenance task is performed based on the trigger area.

[0122] In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the relative position relationship between the first camera and the second camera, and determining a rotation transformation matrix and a translation transformation matrix according to the relative position relationship; determining the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determining the correction matrix according to the relative transformation matrix.

[0123] In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the extrinsic matrix of the first camera and the second camera, and constructing a calibration model of the visual recognition system; determining the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic matrix and the calibration model; determining the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0124] In one embodiment, the processor also implements the following steps when executing the computer program: training the initial trigger area prediction model based on the training data to obtain a candidate trigger area prediction model, and determining the prediction error of the candidate trigger area prediction model; when the prediction error is less than a preset error, determining the candidate trigger area prediction model as the trigger area prediction model; when the prediction error is greater than the preset error, optimizing the candidate trigger area prediction model based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0125] In one embodiment, the processor also implements the following steps when executing the computer program: acquiring a first image and a second image of a training area based on a visual recognition system; and correcting the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image; inputting the training image into the candidate trigger area prediction model to obtain a predicted trigger area output by the candidate trigger area prediction model; determining a trigger distribution line based on the actual trigger area, and determining a projected trigger area based on the trigger distribution line and the predicted trigger area; and determining a prediction error of the candidate trigger area prediction model based on the projected trigger area and the predicted trigger area.

[0126] In one embodiment, when the processor executes the computer program, the following steps are also implemented: determining a candidate trigger distribution line based on the real trigger area; optimizing the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0127] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0128] Acquiring a first image and a second image of a target area based on a visual recognition system, the visual recognition system comprising a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, the first camera and the second camera having the same parameters and optical axes parallel to each other;

[0129] Constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0130] The target image is input into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and a power grid operation and maintenance task is performed based on the trigger area.

[0131] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: obtaining the relative position relationship between the first camera and the second camera, and determining a rotation transformation matrix and a translation transformation matrix based on the relative position relationship; determining the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determining the correction matrix based on the relative transformation matrix.

[0132] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: obtaining the extrinsic matrix of the first camera and the second camera, and constructing a calibration model of the visual recognition system; determining the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic matrix and the calibration model; and determining the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0133] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the initial trigger area prediction model is trained based on the training data to obtain a candidate trigger area prediction model, and the prediction error of the candidate trigger area prediction model is determined; when the prediction error is less than a preset error, the candidate trigger area prediction model is determined as the trigger area prediction model; when the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0134] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: acquiring a first image and a second image of a training area based on a visual recognition system; and correcting the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image; inputting the training image into the candidate trigger area prediction model to obtain a predicted trigger area output by the candidate trigger area prediction model; determining a trigger distribution line based on the actual trigger area, and determining a projected trigger area based on the trigger distribution line and the predicted trigger area; and determining a prediction error of the candidate trigger area prediction model based on the projected trigger area and the predicted trigger area.

[0135] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: determining a candidate trigger distribution line based on the real trigger area; optimizing the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0136] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0137] Acquiring a first image and a second image of a target area based on a visual recognition system, the visual recognition system comprising a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, the first camera and the second camera having the same parameters and optical axes parallel to each other;

[0138] Constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image;

[0139] The target image is input into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and a power grid operation and maintenance task is performed based on the trigger area.

[0140] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: obtaining the relative position relationship between the first camera and the second camera, and determining a rotation transformation matrix and a translation transformation matrix based on the relative position relationship; determining the rotation transformation matrix and the translation transformation matrix as the relative transformation matrix, and determining the correction matrix based on the relative transformation matrix.

[0141] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: obtaining the extrinsic matrix of the first camera and the second camera, and constructing a calibration model of the visual recognition system; determining the relative posture relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic matrix and the calibration model; and determining the correction matrix according to the rotation transformation matrix and the relative posture relationship.

[0142] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: the initial trigger area prediction model is trained based on the training data to obtain a candidate trigger area prediction model, and the prediction error of the candidate trigger area prediction model is determined; when the prediction error is less than a preset error, the candidate trigger area prediction model is determined as the trigger area prediction model; when the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

[0143] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: acquiring a first image and a second image of a training area based on a visual recognition system; and correcting the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image; inputting the training image into the candidate trigger area prediction model to obtain a predicted trigger area output by the candidate trigger area prediction model; determining a trigger distribution line based on the actual trigger area, and determining a projected trigger area based on the trigger distribution line and the predicted trigger area; and determining a prediction error of the candidate trigger area prediction model based on the projected trigger area and the predicted trigger area.

[0144] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: determining a candidate trigger distribution line based on the real trigger area; optimizing the candidate trigger distribution line based on a redundant line optimization algorithm and a discontinuous line optimization algorithm to obtain the trigger distribution line.

[0145] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0146] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0147] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A power grid operation and maintenance method, characterized in that: The method comprises: Acquiring a first image and a second image of a target area based on a visual recognition system, wherein the visual recognition system includes a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, wherein the first camera and the second camera have the same parameters and their optical axes are parallel to each other; constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera, and performing correction processing on the first image and the second image based on the correction matrix to obtain a target image; The target image is input into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and a power grid operation and maintenance task is performed based on the trigger area.

2. The method according to claim 1, characterized in that The constructing a correction matrix based on the relative transformation matrix of the first camera and the second camera includes: Acquire a relative position relationship between the first camera and the second camera, and determine a rotation transformation matrix and a translation transformation matrix according to the relative position relationship; The rotation transformation matrix and the translation transformation matrix are determined as the relative transformation matrix, and the correction matrix is ​​determined according to the relative transformation matrix.

3. The method according to claim 2, characterized in that The determining the correction matrix according to the relative transformation matrix comprises: Obtaining the external parameter matrices of the first camera and the second camera, and constructing a calibration model of the visual recognition system; Determine a relative position relationship between the first camera and the second camera according to the relative transformation matrix, the extrinsic parameter matrix and the calibration model; The correction matrix is ​​determined according to the rotation transformation matrix and the relative posture relationship.

4. The method according to claim 1, characterized in that: The training method of the trigger area prediction model includes: Training the initial trigger region prediction model based on the training data to obtain a candidate trigger region prediction model, and determining a prediction error of the candidate trigger region prediction model; In a case where the prediction error is less than a preset error, determining the candidate trigger area prediction model as the trigger area prediction model; When the prediction error is greater than the preset error, the candidate trigger area prediction model is optimized based on the prediction error until the prediction error of the candidate trigger area prediction model is less than the preset error.

5. The method according to claim 4, characterized in that The determining the prediction error of the candidate trigger area prediction model includes: Acquire a first image and a second image of a training area based on a visual recognition system; and perform correction processing on the first image and the second image based on correction matrices of the first camera and the second camera to obtain a training image; Inputting the training image into the candidate trigger region prediction model to obtain a predicted trigger region output by the candidate trigger region prediction model; Determine a trigger distribution line based on the real trigger area, and determine a projected trigger area according to the trigger distribution line and the predicted trigger area; A prediction error of the candidate trigger area prediction model is determined according to the projected trigger area and the predicted trigger area.

6. The method according to claim 5, characterized in that The determining of the trigger distribution straight line based on the real trigger area includes: Determine a candidate trigger distribution line based on the real trigger area; The candidate trigger distribution straight line is optimized based on a redundant straight line optimization algorithm and a discontinuous straight line optimization algorithm to obtain the trigger distribution straight line.

7. A power grid operation and maintenance device, characterized in that: The device comprises: an acquisition module, configured to acquire a first image and a second image of a target area based on a visual recognition system, wherein the visual recognition system comprises a first camera and a second camera fixedly arranged at a preset distance on the same horizontal reference line, wherein the first camera and the second camera have the same parameters and their optical axes are parallel to each other; A first execution module, configured to construct a correction matrix based on a relative transformation matrix of the first camera and the second camera, and perform correction processing on the first image and the second image based on the correction matrix to obtain a target image; The second execution module is used to input the target image into a pre-trained trigger area prediction model to obtain a trigger area in the target area output by the trigger area prediction model, and execute power grid operation and maintenance tasks based on the trigger area.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images