Substation three-dimensional modeling method based on self-calibration technology

Abstract

The invention discloses a transformer substation three-dimensional modeling method based on a self-calibration technology, and relates to the technical field of transformer substation three-dimensional modeling, and the transformer substation three-dimensional modeling method based on the self-calibration technology comprises the following steps: S1, determining the spatial position of point cloud data; s2, coarse registration: performing coarse registration on the point cloud of the power transformation equipment without the specific mark and the point cloud of the power transformation equipment with the specific mark according to the relative position relation of the equipment; s3, data cleaning: performing cleaning and denoising processing on the point cloud to ensure data quality; s4, precise registration: fitting point cloud line and surface information for the point cloud of the power transformation equipment without the specific mark, and carrying out precise registration; according to the transformer substation three-dimensional modeling method based on the self-calibration technology, the calculation complexity can be effectively managed, efficient descriptors are used in the coarse registration stage to reduce the data volume, and therefore the overall registration speed is increased, and robustness can be improved.

Description

Technical Field

[0001] The invention belongs to the technical field of substation three-dimensional modeling, and in particular relates to a substation three-dimensional modeling method based on a self-calibration technology. Background Art

[0002] In recent years, with the continuous development of smart devices and digital concepts, applications such as virtual reality technology, smart cities, and digital factories are actively promoting the operation and progress of society. Among them, accurate three-dimensional digital models provide realistic expressions for real scenes, equipment and facilities, and surrounding environments. Traditional modeling technology uses professional software and manual markings to create the surface shape of objects. The massive amount of data leads to low efficiency and high labor intensity. With the development of science and technology, digital photogrammetry technology has been widely used in scene reconstruction, and point cloud registration plays an irreplaceable role in three-dimensional reconstruction. It accurately transforms point cloud data in different postures to the same coordinate system, and then merges these point cloud data into a complete and continuous three-dimensional model, realizing real-time three-dimensional automated modeling of physical targets of electrical equipment, lines, and operators, laying the foundation for intelligent management and control of personnel at the operation site;

[0003] At present, the more classic and widely used point cloud registration method is the iterative closest point algorithm (ICP). For coarse registration, CHEN et al. evaluated 9 popular local feature descriptors using 8 popular datasets obtained by different techniques. These 9 descriptors are Spin Image (SI), Shape Context (3-DSC), Unique Shape Context (USC), Direction Histogram Signature (SHOT), Local Surface Patches (LSP), Point Feature Histogram (PFH), Fast Point Feature Histogram (FPFH), Triple-Spin Image (TriSI) and Rotational Projection Statistics (RoPS). In general, RoPS is the best performing descriptor in terms of descriptiveness, USC and TriSI are the most robust descriptors, and in terms of efficiency, SI has very good performance on different numbers of points. In addition, the initial registration is performed based on the intrinsic shape signature (ISS) feature points and the sampling consistency registration (SAC-IA) algorithm, and then the iterative nearest neighbor (ICP) algorithm is used for precise registration. These point cloud registration methods are highly descriptive, but not very robust, prone to errors in corresponding point pairs, and have complex calculations and low registration efficiency. Summary of the invention

[0004] The purpose of the present invention is to provide a substation three-dimensional modeling method based on self-calibration technology with a simple structure and reasonable design in order to solve the above problems.

[0005] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0006] A three-dimensional modeling method for a substation based on a self-calibration technology comprises the following steps:

[0007] S1. Determine the spatial position of the point cloud data, including:

[0008] S101, collecting point cloud data of the transformer in the substation by scanning;

[0009] S102, for the point cloud of the substation equipment without specific marks, construct a point cloud coordinate system for each scanned point cloud according to the arrangement information of the point cloud scene equipment and the ground information;

[0010] S103, for the point cloud of the substation equipment with specific marks, directly establish its relative rotation and translation relationship with the world coordinate system, and obtain the precise position and posture of the substation equipment in the world coordinate system;

[0011] S2, coarse registration: the point cloud of the substation equipment without specific marks is roughly registered with the point cloud of the substation equipment with specific marks according to the relative position relationship of the equipment, and the preliminary spatial relationship between the point cloud of the equipment without mark and the point cloud of the equipment with mark is established;

[0012] S3, data cleaning: clean and denoise the point cloud to ensure data quality;

[0013] S4. Precise registration: Based on the point cloud of the substation equipment without specific marks, the point cloud lines and surface information are fitted, and precise registration is performed to obtain detailed geometric feature information and complete the construction of the three-dimensional model of the substation equipment.

[0014] As a further optimization scheme of the present invention, the point cloud of the substation equipment without specific marks is assumed to be the point cloud P to be aligned, and the point cloud set of the substation equipment with specific marks is assumed to be the reference point cloud set {Q1,...,Qi}, i∈[1,...,Q], Q is the number of reference point cloud sets, and the point clouds are roughly aligned according to the relative position relationship of the equipment.

[0015] As a further optimization solution of the present invention, in step S2, the specific process of coarse registration is as follows:

[0016] Step 1: Set the number of initial optimal matching reference point clouds to M = 0;

[0017] Step 2: Select point cloud Qi from the reference point cloud set {Q1,...,Qi};

[0018] Step 3: Determine whether the point cloud P to be registered and the reference point cloud Qi have the same type of equipment; if so, proceed to step 4; if not, proceed to step 10;

[0019] Step 4: Randomly select a pair of the same type of devices Ai and Bi from the registration point cloud P and the reference point cloud Qi, Ai∈P,Bi∈Qi;

[0020] Step 5: Find the device pairs Ai-Aj, Bi-Bj with the same distance and direction as Ai and Bi in the point cloud to be registered P and the reference point cloud Qi, Aj∈P, Bj∈Qi;

[0021] Step 6: Determine whether there is a device pair with the same distance and direction. If so, proceed to step 7; if not, return to step 4;

[0022] Step 7: For each pair of devices, calculate the number of devices that can match the remaining devices of the point cloud P to be registered with the reference point cloud Qi devices, where the maximum number of devices that can match for each pair of devices is N;

[0023] Step 8: Determine whether N is greater than M. If yes, proceed to step 9; if no, proceed to step 10;

[0024] Step 9: Update M=N, and update the optimal matching reference point cloud to Qm=Qi;

[0025] Step 10: Delete Qi from the reference point cloud set;

[0026] Step 11: determine whether the number of reference point clouds in the reference point cloud set is 0, if yes, proceed to step 12; if no, return to step 2;

[0027] Step 12: Determine the best matching reference point cloud as Qm, calculate the relative rotation and translation relationship between the point cloud to be registered P and the best matching reference point cloud as Qm, and complete the rough registration.

[0028] As a further optimization solution of the present invention, in step S4, a RANSAC (Random Sample Consensus) algorithm is used to fit the straight line and plane shape information.

[0029] As a further optimization scheme of the present invention, the specific steps include S401, using RANSAC to match the lines and surfaces of two point clouds; S402, performing precise alignment by minimizing the distance between the matched lines and surfaces.

[0030] As a further optimization solution of the present invention, in step S3, specifically including:

[0031] S301, removing noise points: identifying and removing obvious noise points by statistical analysis methods;

[0032] S302, removing isolated points: using neighborhood analysis of the point cloud to detect and remove isolated points that are not sufficiently associated with surrounding points;

[0033] S303, noise reduction processing: using a smoothing algorithm to reduce random noise in the point cloud;

[0034] S304, data sampling: Use voxel grid filters or random sampling methods to reduce the number of points in the point cloud to reduce the amount of calculation while retaining the main features of the point cloud.

[0035] As a further optimization scheme of the present invention, the minimization loss function of the precise registration is as follows:

[0036]

[0037] Among them, let the equations of the two matching straight lines be A i1 x+B i1 y+C i1 =0 and A i2 x+B i2 y+C i2 =0, i∈[1,..,K], K is the number of matching lines; let the equations of the two matching surfaces be A j1 x+B j1 y+C j1 z+D j1 =0 and A j2 x+B j2 y+C j2 z+D j2 =0, j∈[1,..,N], N is the number of matching faces.

[0038] The present invention has at least the following beneficial effects: the present invention can generate a high-precision three-dimensional model through effective point cloud registration, and fuse point cloud data from different perspectives or time points into a complete three-dimensional model. This data fusion not only improves the integrity of the model, but also ensures the continuity and consistency of information. Through the staged registration strategy (coarse registration and fine registration), the computational complexity can be effectively managed. The coarse registration stage uses efficient descriptors to reduce the amount of data, thereby improving the overall registration speed and can improve robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flowchart of the steps of the substation three-dimensional modeling method of the present invention;

[0040] Figure 2 It is a flow chart of the rough registration steps of the present invention. DETAILED DESCRIPTION

[0041] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0042] like Figure 1 , Figure 2 As shown, a substation 3D modeling method based on self-calibration technology includes the following steps:

[0043] S1. Determine the spatial position of the point cloud data, including:

[0044] S101, firstly collect point cloud data of the transformer of the substation by scanning;

[0045] S102, for the point cloud of the substation equipment without specific marks, construct a point cloud coordinate system for each scanned point cloud according to the arrangement information of the point cloud scene equipment and the ground information;

[0046] S103, for the point cloud of the substation equipment with specific marks, directly establish its relative rotation and translation relationship with the world coordinate system, and obtain the precise position and posture of the substation equipment in the world coordinate system;

[0047] Specifically, it includes importing the collected point cloud data into the data processing tool, performing preliminary processing such as downsampling and denoising on the point cloud data to improve the data quality;

[0048] For point clouds of substation equipment without specific markings, a coordinate system is established using the layout of the equipment and ground features to align and process the point cloud data;

[0049] Point clouds of substation equipment with specific marks, find these known mark points in the point cloud data, and record their actual world coordinates. Using the positional relationship of the mark points, calculate the rotation and translation matrices from the point cloud coordinate system to the world coordinate system, where the rotation matrix: defines how to rotate the point cloud coordinate system to match the direction of the world coordinate system; the translation vector: defines how the origin of the point cloud coordinate system moves to the corresponding position of the world coordinate system; apply the calculated rotation and translation matrices to the point cloud data to convert it to the world coordinate system. This allows different point cloud data to be compared and analyzed in the same reference frame.

[0050] S2, coarse registration: the point cloud of the substation equipment without specific marks is roughly registered with the point cloud of the substation equipment with specific marks according to the relative position relationship of the equipment, and the preliminary spatial relationship between the point cloud of the equipment without mark and the point cloud of the equipment with mark is established;

[0051] Specifically include:

[0052] Assume that the point cloud of the substation equipment without specific marks is the point cloud to be registered P, and the point cloud set of the substation equipment with specific marks is the reference point cloud set {Q1,...,Qi}, i∈[1,...,Q], Q is the number of reference point cloud sets, and perform rough registration of point clouds according to the relative position relationship of the equipment. The specific process of rough registration is as follows:

[0053] Step 1: Set the number of initial optimal matching reference point clouds to M = 0;

[0054] Step 2: Select point cloud Qi from the reference point cloud set {Q1,...,Qi};

[0055] Step 3: Determine whether the point cloud to be registered P and the reference point cloud Qi have the same type of equipment; if so, proceed to step 4; if not, proceed to step 10;

[0056] Make sure that the devices in the point cloud to be registered P and the reference point cloud Qi are of the same type to ensure the rationality of the registration. If the device types in the two point clouds are consistent, their geometric features and sizes may be similar, so it is possible to find an effective match. Only when the device types are consistent, proceed to the subsequent registration steps. This ensures that the device structures in the registration process are compatible, thereby improving the accuracy and effect of the registration.

[0057] Step 4: Randomly select a pair of the same type of devices Ai and Bi from the registration point cloud P and the reference point cloud Qi, Ai∈P,Bi∈Qi;

[0058] Step 5: Find the device pairs Ai-Aj, Bi-Bj with the same distance and direction as Ai and Bi in the point cloud to be registered P and the reference point cloud Qi, Aj∈P, Bj∈Qi;

[0059] Step 6: Determine whether there is a device pair with the same distance and direction. If so, proceed to step 7; if not, return to step 4;

[0060] Among them, the direction between Ai and Bi can be obtained by calculating the angle between the line connecting Ai and Bi and the coordinate axis, or by calculating the direction vector from Ai to Bi and normalizing it. The distance between Ai and Bi is di = |Ai-Bi|. Similarly, the direction between Aj and Bj can be obtained by calculating the angle between the line connecting Aj and Bj and the coordinate axis, or by calculating the direction vector. The distance between Aj and Bj is dj = |Aj-Bj|. If the distance and direction of (Aj, Bj) are consistent with (Ai, Bi), it is considered that a matching device pair has been found, otherwise it is not.

[0061] Step 7: For each pair of devices, calculate the number of devices that can match the remaining devices of the point cloud P to be registered with the reference point cloud Qi devices, where the maximum number of devices that can match for each pair of devices is N;

[0062] Step 8: Determine whether N is greater than M. If yes, proceed to step 9; if no, proceed to step 10;

[0063] For example: the point cloud P to be registered contains 4 pairs of devices (A1, B1), (A2, B2), (A3, B3), and (A4, B4);

[0064] The reference point cloud Q_i contains 4 pairs of devices (C1, D1), (C2, D2), (C3, D3), and (C_4, D_4); Assume that the device pair (A1, B1) in the point cloud P to be registered is selected. After removing (A1, B1) from the point cloud P to be registered, the remaining device pairs are (A2, B2), (A3, B3), and (A4, B4);

[0065] Number of matches N: Calculate the number of matches between the remaining device pairs {(A2, B2), (A3, B3), and (A4, B4)} and the device pairs in the reference point cloud Qi. Assume that in this process, 2 pairs of device pairs are found, such as (A2, B2) and (C3, D3), (A4, B4) and (C1, D1), so N = 2. Assume that the set threshold M is 3, N is less than M, which means that after removing (A1, B1), the remaining number of matches is not enough to meet our threshold condition. In other words, this pair of device pairs may have a greater impact on the registration quality, and removing them may lead to poor registration results.

[0066] Step 9: Update M=N, and update the optimal matching reference point cloud to Qm=Qi;

[0067] Step 10: Delete Qi from the reference point cloud set;

[0068] Step 11: determine whether the number of reference point clouds in the reference point cloud set is 0, if yes, proceed to step 12; if no, return to step 2;

[0069] Step 12: Determine the best matching reference point cloud as Qm, calculate the relative rotation and translation relationship between the point cloud to be registered P and the best matching reference point cloud as Qm, and complete the rough registration.

[0070] S3, data cleaning: clean and denoise the point cloud to ensure data quality;

[0071] Specifically include:

[0072] S301, removing noise points: identifying and removing obvious noise points by statistical analysis methods;

[0073] S302, removing isolated points: using neighborhood analysis of the point cloud to detect and remove isolated points that are not sufficiently associated with surrounding points;

[0074] S303, noise reduction processing: using a smoothing algorithm to reduce random noise in the point cloud;

[0075] S304, data sampling: Use voxel grid filters or random sampling methods to reduce the number of points in the point cloud to reduce the amount of calculation while retaining the main features of the point cloud.

[0076] S4, precise registration: Based on the point cloud of the substation equipment without specific marks, fit the point cloud line and surface information for precise registration; use the RANSAC (Random Sample Consensus) algorithm to fit the straight line and plane shape information. The specific steps include S401, using RANSAC to match the line and surface of two point clouds; S402, precise registration by minimizing the distance between the matching lines and surfaces; obtain detailed geometric feature information and complete the construction of the three-dimensional model of the substation equipment;

[0077] The minimization loss function for fine registration is as follows:

[0078]

[0079] Among them, let the equations of the two matching straight lines be A i1 x+B i1 y+C i1 =0 and A i2 x+B i2 y+C i2 =0, i∈[1,..,K], K is the number of matching lines; let the equations of the two matching surfaces be A j1 x+B j1 y+C j1 z+D j1 =0 and A j2 x+B j2 y+C j2 z+D j2 =0, j∈[1,..,N], N is the number of matching faces;

[0080] The position and posture of the point cloud data are adjusted through an optimization algorithm to minimize the value of a loss function. The loss function usually measures the difference between the registered point cloud and the target point cloud, such as the distance between points or the overlap error. By minimizing this loss function, the most accurate point cloud registration result can be obtained to improve the accuracy of the substation 3D model.

[0081] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A three-dimensional modeling method for a substation based on self-calibration technology, characterized in that: The following steps are involved: S1. Determine the spatial position of the point cloud data, including: S101, collecting point cloud data of the transformer in the substation by scanning; S102, for the point cloud of the substation equipment without specific marks, construct a point cloud coordinate system for each scanned point cloud according to the arrangement information of the point cloud scene equipment and the ground information; S103, for the point cloud of the substation equipment with specific marks, directly establish its relative rotation and translation relationship with the world coordinate system, and obtain the precise position and posture of the substation equipment in the world coordinate system; S2, coarse registration: the point cloud of the substation equipment without specific marks is roughly registered with the point cloud of the substation equipment with specific marks according to the relative position relationship of the equipment, and the preliminary spatial relationship between the point cloud of the equipment without mark and the point cloud of the equipment with mark is established; S3, data cleaning: cleaning and denoising the point cloud; S4. Precise registration: Based on the point cloud of the substation equipment without specific marks, the point cloud lines and surface information are fitted, and precise registration is performed to obtain detailed geometric feature information and complete the construction of the three-dimensional model of the substation equipment.

2. A substation three-dimensional modeling method based on self-calibration technology according to claim 1, characterized in that: Assume that the point cloud of the substation equipment without specific marks is the point cloud to be registered P, and the point cloud set of the substation equipment with specific marks is the reference point cloud set {Q1,...,Qi}, i∈[1,...,Q], Q is the number of reference point cloud sets, and perform rough point cloud registration according to the relative position relationship of the equipment.

3. The three-dimensional modeling method of a substation based on self-calibration technology according to claim 2 is characterized in that: In step S2, the specific process of coarse registration is as follows: Step 1: Set the number of initial optimal matching reference point clouds to M = 0; Step 2: Select point cloud Qi from the reference point cloud set {Q1,...,Qi}; Step 3: Determine whether the point cloud P to be registered and the reference point cloud Qi have the same type of equipment; if so, proceed to step 4; if not, proceed to step 10; Step 4: Randomly select a pair of the same type of devices Ai and Bi from the registration point cloud P and the reference point cloud Qi, Ai∈P,Bi∈Qi; Step 5: Find the device pairs Ai-Aj, Bi-Bj with the same distance and direction as Ai and Bi in the point cloud to be registered P and the reference point cloud Qi, Aj∈P, Bj∈Qi; Step 6: Determine whether there is a device pair with the same distance and direction. If so, proceed to step 7; if not, return to step 4; Step 7: For each pair of devices, calculate the number of devices that can match the remaining devices of the point cloud P to be registered with the reference point cloud Qi devices, where the maximum number of devices that can match for each pair of devices is N; Step 8: Determine whether N is greater than M. If yes, proceed to step 9; if no, proceed to step 10; Step 9: Update M=N, and update the optimal matching reference point cloud to Qm=Qi; Step 10: Delete Qi from the reference point cloud set; Step 11: determine whether the number of reference point clouds in the reference point cloud set is 0, if yes, proceed to step 12; if no, return to step 2; Step 12: Determine the best matching reference point cloud as Qm, calculate the relative rotation and translation relationship between the point cloud to be registered P and the best matching reference point cloud as Qm, and complete the rough registration.

4. The three-dimensional modeling method of a substation based on self-calibration technology according to claim 3 is characterized in that: In step S4, the RANSAC algorithm is used to fit the straight line and plane shape information.

5. The three-dimensional modeling method of a substation based on self-calibration technology according to claim 4 is characterized in that: The specific steps include S401, using RANSAC to match the line and surface of two point clouds; S402, performing precise alignment by minimizing the distance between the matched lines and surfaces.

6. A three-dimensional modeling method for a substation based on self-calibration technology according to claim 5, characterized in that: In step S3, it specifically includes: S301, removing noise points: identifying and removing obvious noise points by statistical analysis methods; S302, removing isolated points: using neighborhood analysis of the point cloud to detect and remove isolated points that are not sufficiently associated with surrounding points; S303, noise reduction processing: using a smoothing algorithm to reduce random noise in the point cloud; S304, data sampling: Use voxel grid filters or random sampling methods to reduce the number of points in the point cloud to reduce the amount of calculation while retaining the main features of the point cloud.

7. A three-dimensional modeling method for a substation based on self-calibration technology according to claim 6, characterized in that: The minimization loss function for fine registration is as follows: Among them, let the equations of the two matching straight lines be A i1 x+B i1 y+C i1 =0 and A i2 x+B i2 y+C i2 =0, i∈[1,..,K], K is the number of matching lines; let the equations of the two matching surfaces be A j1 x+B j1 y+C j1 z+D j1 =0 and A j2 x+B j2 j+C j2 z+D j2 =0, j∈[1,..,N], N is the number of matching faces.

Images