Method, system and equipment for positioning object in transformer substation and medium
Data is collected through drones and cart-mounted lidars, and lightweight three-dimensional models are automatically generated, solving the problem of difficulty in updating three-dimensional models in the substation and improving positioning accuracy and the availability of operation safety control.
Patent Information
- Application Number
- CN202411079875.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-15
AI Technical Summary
In the prior art, three-dimensional models in the substation require manual measurement creation, resulting in measurement errors, long cycles and high costs, difficult to update in time, affecting the reliability of operation safety control.
Data is collected by drones and cart-mounted lidar, and lightweight three-dimensional models are automatically optimized through point cloud processing technology, linear features are extracted, clustered and fitted road boundaries, and object positioning is combined with point cloud segmentation network to realize the substation Accurate positioning of objects.
It improves the position accuracy and update frequency of the substation model, enhances the availability of operation safety control, reduces the cost and time of manual operation, and improves the ability to adapt to changes in equipment in the substation.
Smart Images

Figure CN118962691A_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to the technical field of point cloud data processing of physical objects in a substation, and in particular to a method, system, equipment and medium for positioning objects in a substation. [Background technology]
[0002] With the gradual completion of the Beidou ground-based augmentation platform, the placement of Beidou positioning tags on substation workers and vehicles can achieve centimeter-level positioning accuracy, and has broad application prospects in power grid operation and inspection services. Real-time positioning needs to be combined with an accurate proportional three-dimensional substation model to achieve intelligent operation safety management and control. However, at present, manual on-site measurements are required to create three-dimensional models. On the one hand, due to measurement errors, the position of the equipment in the station in the model does not match the actual situation. On the other hand, manual modeling has a long cycle and high cost. After infrastructure construction or equipment replacement in the substation, the three-dimensional model is difficult to update in time, making it difficult to actually apply operation safety management that relies on accurate location information. As the front end of the three-dimensional safety management and control system for power operations, the accuracy and lightweight of the three-dimensional model are very important, and are the basis for the large-scale application and promotion of the three-dimensional safety management and control system for operations.
[0003] Therefore, it is necessary to study a method, system, device and medium for locating objects in a substation to address the shortcomings of the existing technology and to solve or alleviate one or more of the above problems. [Summary of the invention]
[0004] In view of this, the present invention provides a method, system, equipment and medium for positioning objects in a substation, abandoning the original manual patch splicing modeling method. After the lightweight three-dimensional point cloud modeling uses drones and carts with multiple laser scanning devices to collect on-site data, it automatically optimizes and generates an integrated point cloud model of key components based on the degree of attention to the power equipment. It can quickly generate complete model data for the operating area and is more adaptable to scenarios such as adding new equipment, temporary construction, and frequent equipment changes in the substation.
[0005] In one aspect, the present invention provides a method for locating an object in a substation, the method comprising:
[0006] S1, perform lidar scanning in the traversable area of the substation to obtain point cloud data of the substation;
[0007] S2, extracting a feature point set containing straight line features from the point cloud data;
[0008] S3, obtain a set of characteristic straight lines through a set of characteristic points containing straight line features,
[0009] S4, clustering and screening the set of characteristic straight lines to obtain valid road boundary straight line segments;
[0010] S5, performing road boundary fitting through effective road boundary straight line segments to obtain the road boundary of the accessible area of the substation;
[0011] S6, performing point cloud semantic segmentation on the point cloud data of the substation through a point cloud segmentation network and extracting feature points of physical objects in the substation to obtain a feature point set of each object in the substation;
[0012] S7, integrating the road boundary of the traversable area of the substation and the feature point set of each object in the substation, completes the positioning of the objects in the substation.
[0013] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein S2 specifically includes:
[0014] Each scan line of the laser radar is used as a horizontal index, and three consecutive points on the same scan line that meet the preset constraints are used as a set of horizontal candidate feature points;
[0015] Taking the radial angle change between adjacent laser radar scanning lines as a constraint condition, a radial candidate feature point set is obtained;
[0016] According to the above aspects and any possible implementation, an implementation is further provided, wherein the constraints preset in the process of acquiring the horizontal candidate feature point set in S2 are specifically:
[0017] Replace three consecutive points with three points separated by a distance of d, project the point cloud onto the XOY plane, and use the angle relationship of the three points to establish constraint conditions for feature point extraction;
[0018] The three points on the same scan line l are p l,i-d , P l,i , p l,i+d , the coordinates are (x l,i-d ,y l,i-d ), (x l,i ,y l,i ), (x l,i+d ,y l,i+d ) Every two adjacent points can determine a vector, defining α i,j The positional relationship of the three points is described as a vector angle as follows:
[0019]
[0020] Under flat conditions such as road surfaces, α i,j The value is small, close to 0°, at the road boundary, α i,j Close to 90°, select threshold α th =30°, d=5, keep [p l,i-d , p l,i+d ] interval, the coefficient αl,k The largest point p l,k As the feature point, id ≤ k ≤ i + d, by traversing each scan line, a set of horizontal candidate feature points is constructed.
[0021] The constraint conditions for the radial angle change in S2 are specifically:
[0022] Preset p l-1,i , p l,i are two points located on adjacent scan lines with the same azimuth, and their coordinates are (x l-1,i ,y l-1,i , z l-1,i ), (x l,i ,y l,i , z l,i ), let the radial angle between the two points be
[0023] Take β th =20°, when p l,i When located at the road boundary, the radial angle between adjacent scan lines compared to a flat road surface Changes occur if It is believed that p l,i is a qualified feature point. For each set of radial data, only the first point that meets the constraint conditions is extracted from the inside to the outside as the feature point. Each set of radial data is traversed to construct a radial candidate feature point set.
[0024] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein S3 specifically includes:
[0025] S31: The horizontal candidate feature point set is based on each scanning line, and any two adjacent feature points are used as the first and last endpoints to define a discrete curve. A preset constraint method is executed for each discrete curve to retain the horizontal feature straight line set that meets the constraint conditions.
[0026] S32: The radial candidate feature point set is first discretized, and then, taking each radial line as a unit, any two adjacent feature points are used as the first and last endpoints to define a discrete curve, and a preset constraint method is executed for each discrete curve to retain a radial feature straight line set that meets the constraint conditions;
[0027] S33: Fusing and summarizing the transverse feature straight line set and the radial feature straight line set to obtain a set of feature straight lines.
[0028] As described above and any possible implementation method, an implementation method is further provided, wherein the clustering and screening method in S4 classifies straight line segments through K-Means clustering, horizontal analysis and vertical analysis, and RANSAC-based screening, and selectively retains feature straight line segments as valid road boundary straight line segments.
[0029] According to the aspects described above and any possible implementation method, an implementation method is further provided, in which the road boundary fitting process in S5 fits the road boundary through cubic B-spline interpolation, and optimizes the interpolated and fitted road boundary through inter-cluster boundary smoothing processing to obtain the road boundary of the accessible area of the substation.
[0030] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the semantic segmentation process in S6 specifically includes:
[0031] S61: The segmentation process of the point cloud data is restricted by using a smooth threshold as a segmentation condition, and an improved Euclidean clustering segmentation is performed;
[0032] S62: Perform missing optimization, feature position encoding, network learning and normalization on the results after clustering segmentation.
[0033] According to the above aspects and any possible implementation manner, a system for locating an object in a substation is further provided, and the system for locating an object includes the following modules:
[0034] The point cloud data acquisition module is used to perform laser radar scanning in the traversable area of the substation to obtain the point cloud data of the substation;
[0035] A feature point set extraction module is used to extract feature point sets containing straight line features from point cloud data;
[0036] The feature line set acquisition module acquires the feature line set through the feature point set containing the line features.
[0037] An effective road boundary straight line segment acquisition module is used to cluster and screen a set of characteristic straight lines to obtain effective road boundary straight line segments;
[0038] A road boundary fitting module is used to fit the road boundary through valid road boundary straight line segments to obtain the road boundary of the accessible area of the substation;
[0039] The object feature point acquisition module is used to perform point cloud semantic segmentation on the point cloud data of the substation through the point cloud segmentation network and extract feature points of the physical objects in the substation to obtain the feature point set of each object in the substation;
[0040] The object positioning module is used to integrate the road boundaries of the traversable area of the substation and the feature point set of each object in the substation to complete the positioning of objects in the substation.
[0041] According to the aspects described above and any possible implementation methods, an electronic device is further provided, including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program, and when the computer program is executed by the processor, the steps of the method for locating objects in a substation are implemented.
[0042] According to the aspects described above and any possible implementation, a computer-readable storage medium is further provided, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the method for locating objects in a substation are implemented.
[0043] Compared with the prior art, the present invention can achieve the following technical effects:
[0044] The present invention uses a laser radar mounted on an unmanned aerial vehicle or a ground vehicle to survey and scan the substation and the transmission lines. The software can process the point cloud data and automatically generate a lightweight point cloud model suitable for operational safety management and control, thereby greatly improving the position accuracy and update frequency of the substation model and improving the availability of operational safety management and control.
[0045] Of course, any product implementing the present invention does not necessarily need to achieve all of the above-mentioned technical effects at the same time.
Brief Description of the Drawings
[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1 is a flow chart of an object positioning method provided by an embodiment of the present invention;
[0048] Figure 2 is a point cloud straight line feature description diagram provided by an embodiment of the present invention;
[0049] Figure 3 is a schematic diagram of feature point extraction provided by an embodiment of the present invention;
[0050] Figure 4 is a supplementary map of missed feature points provided by an embodiment of the present invention;
[0051] Figure 5is a flow chart of a characteristic line extraction algorithm provided by an embodiment of the present invention;
[0052] Figure 6 is a characteristic straight line extraction result diagram provided by an embodiment of the present invention;
[0053] Figure 7 is a straight line segment clustering flow chart provided by an embodiment of the present invention;
[0054] Figure 8 is a schematic diagram of inter-cluster boundary smoothing processing provided by an embodiment of the present invention;
[0055] Fig. 9 It is a flow chart of smoothness-based Euclidean clustering provided by one embodiment of the present invention;
[0056] Fig.10 It is a Kdtree data structure diagram provided by an embodiment of the present invention;
[0057] Fig.11 It is an EdgeConv graph convolution graph provided by an embodiment of the present invention;
[0058] Fig.12 is a TransformerNet network structure diagram provided by an embodiment of the present invention;
[0059] Fig.13 It is a connection attention and splicing module provided by an embodiment of the present invention;
[0060] Fig.14 It is a schematic diagram of an electronic device provided by an embodiment of the present invention. [Specific implementation method]
[0061] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0062] It should be clear that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0063] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.
[0064] Embodiment 1:
[0065] like Figure 1As shown, the present invention provides a method for locating an object in a substation, and the object locating method comprises:
[0066] S1, using drones, unmanned vehicles, or a combination of drones and unmanned vehicles, to perform lidar scanning of the substation in the traversable area to obtain point cloud data of the substation;
[0067] S2, extracting a feature point set containing straight line features from the point cloud data;
[0068] Each scan line of the laser radar is used as a horizontal index, and three consecutive points on the same scan line that meet the preset constraints are used as a set of horizontal candidate feature points;
[0069] Taking the radial angle change between adjacent laser radar scanning lines as a constraint condition, a radial candidate feature point set is obtained;
[0070] S3, obtaining a set of feature lines through a lateral candidate feature point set and a radial candidate feature point set,
[0071] S4, clustering and screening the set of characteristic straight lines to obtain valid road boundary straight line segments;
[0072] S5, performing road boundary fitting through effective road boundary straight line segments to obtain the road boundary of the accessible area of the substation;
[0073] S6, a point cloud segmentation network based on point cloud transformer performs point cloud semantic segmentation on the point cloud data of the substation and extracts feature points of physical objects in the substation to obtain a set of feature points of each object in the substation;
[0074] S7, integrating the road boundary of the traversable area of the substation and the feature point set of each object in the substation, completes the positioning of the objects in the substation.
[0075] In a specific embodiment, S2 first defines and extracts the spatial straight line features of the road boundary based on a complex outdoor environment. The straight line features in the point cloud data can be regarded as approximate straight lines, but not absolute straight lines. In complex point cloud data, it is difficult to accurately extract all straight line features. The reasons are mainly manifested in the following three situations:
[0076] 1) The laser radar point cloud data consists of multiple circular scanning lines. When the scanning line radius of the radar is large, several consecutive points on the same scanning circle can be regarded as approximate straight line segments, such as Figure 2 (a) The straight line segment marked 1.
[0077] 2) When the relative undulation of the ground is more obvious, the laser radar scanning line will be affected by the terrain changes, forming an approximate pseudo-straight line feature, such as Figure 2(b) The straight line segment with 2 on the left.
[0078] 3) There is a certain error in the laser radar distance measurement, which will also be affected by the roughness of the object. Points on the same straight line segment will not be absolutely aligned, such as Figure 2 (b) The straight line segment with 3 on the right.
[0079] The present invention designs a simple, fast and comprehensive feature line extraction algorithm. In order to improve the algorithm's computational efficiency and facilitate the algorithm to process point cloud data, the algorithm constructs the collected three-dimensional point cloud data into a manner that can be horizontally indexed by each scanning line according to the scanning characteristics of the laser radar. It can be seen from the distribution of the point clouds of roads and road boundaries that, in general, when the three-dimensional laser radar scans a flat road surface, the small curve formed by three consecutive points on the same scanning line is relatively smooth, and the angle formed by these three points is close to a flat angle; when the radar scans the road boundary, the angle formed by these three points will decrease sharply compared to the flat road surface. In order to overcome the roughness of the ground, the present invention replaces three consecutive points with three points spaced d apart from each other. In order to improve the operating efficiency of the algorithm, without affecting the experimental results, the point cloud is projected onto the XOY plane, and the constraint conditions are established using the angle relationship of the three points to extract feature points.
[0080] like Figure 3 As shown, assume that the three points on the same scan line l are p l,i-d , p l,i , p l,i+d , the coordinates are (x l,i-d ,y l,i-d ), (x l,i ,y l,i ), (x l,i+d ,y l,i+d ) Every two adjacent points can determine a vector. The present invention defines α i,j The positional relationship of the three points is described as a vector angle as follows:
[0081]
[0082] Under flat conditions such as road surfaces, α i,j The value is small, close to 0°, at the road boundary, α i,j Close to 90°, considering the complexity of the environment and the robustness of the algorithm, the present invention selects a relatively loose threshold α th =30°, d=5, so as to extract as many feature points as possible. In order to ensure that the feature points are representative enough, the algorithm only retains the feature points in [p l,i-d , p l,i+d ] interval, the coefficient α l,k The largest point p l,kAs a feature point, id≤k≤i+d. By traversing each scan line, a candidate feature point set F1 is constructed.
[0083] In special cases, such as Figure 4 As shown, when the radar scan line coincides with the road boundary, the points on the same scan line are relatively smooth, which will result in failure to meet the above constraints and cause feature point misses. In view of this situation, the present invention uses the radial angle change between points on adjacent scan lines to supplement feature points.
[0084] Assume that p l-1,i , p l,i are two points located on adjacent scan lines with the same azimuth, and their coordinates are (x l-1,i ,y l-1,i, z l-1,i ), (x l,i ,y l,i , z l,i ). Let the radial angle between the two points be Take β th =20°, when p l,i When located at the road boundary, the radial angle between adjacent scan lines compared to a flat road surface There will be relatively obvious changes if I think p l,i is a feature point that meets the conditions. In order to improve the efficiency of the algorithm and eliminate the influence of complex environment on feature point extraction, for each set of radial data, only the first point that meets the constraint conditions is extracted from the inside to the outside as a feature point. The present invention constructs a candidate feature point set F2 by traversing each set of radial data.
[0085] The S3 specifically includes:
[0086] After feature point extraction, two representative candidate feature point sets F1 and F2 (ie, a horizontal candidate feature point set and a vertical candidate feature point set) are constructed.
[0087] S31: The horizontal candidate feature point set is based on each scanning line, and any two adjacent feature points are used as the first and last endpoints to define a discrete curve. A preset constraint method is executed for each discrete curve to retain the horizontal feature straight line set that meets the constraint conditions.
[0088] Specifically, taking each scan line as a unit, any two adjacent feature points are used as the first and last endpoints to define a discrete curve. The specific algorithm for each discrete curve is as follows:
[0089] Input: Any discrete curve S c .
[0090] Output: whether to store Sc as a characteristic straight line.
[0091] Step 1: Calculate the discrete curve S c The number of points included N c and S c The length D c :
[0092]
[0093] Step 2: If S c At the same time, the constraints N c >N,D c > D, go to step 3, otherwise refuse to store S c
[0094] Step 3: Select S c The starting point (x s ,y s ) and the end point (x e ,y e ) as a reference point, establish a straight line model Ax+By+C=0
[0095] Step 4: Calculate S separately c Each point (x i ,y i ) to the line Ax+By+C=0 i
[0096] Step 5: If S c Each point (x i ,y i ) to the line Ax+By+C=0 i All meet d i <d, then S is stored in the form of a characteristic straight line c , otherwise refuse to store S c .
[0097] The algorithm converts the discrete curve S c The length D c It is defined as the distance from the starting point to the end point. N represents the minimum number of points contained in the characteristic line, D represents the minimum length of the characteristic line, and d i Indicates the maximum error allowed for each point on the characteristic line. In the algorithm experiment, N = 10, D = 0.2m, d i =0.05m. The corresponding algorithm flow chart is as follows Figure 5 shown.
[0098] S32: The radial candidate feature point set is first discretized, and then, taking each radial line as a unit, any two adjacent feature points are used as the first and last endpoints to define a discrete curve, and a preset constraint method is executed for each discrete curve to retain a radial feature straight line set that meets the constraint conditions;
[0099] Specifically, since the feature points in the point set F2 are relatively densely distributed, the present invention firstly discretizes F2 using the value of N, and then uses the same processing method as F1 to extract the feature straight lines.
[0100] S33: Fusing and summarizing the transverse characteristic straight line set and the radial characteristic straight line set to obtain a characteristic straight line set. Specifically, the processing results of the integrated point sets F1 and F2 are combined to obtain a characteristic straight line set.
[0101] The clustering and screening method in S4 achieves the classification of straight line segments through K-Means clustering, horizontal analysis and vertical analysis, and RANSAC-based screening, and selectively retains characteristic straight line segments as valid road boundary straight line segments.
[0102] In a specific embodiment, through the steps S1-S3 in S4, a set of characteristic straight line segments can be obtained, and they are relatively independent of each other. In order to exclude the approximate straight line segments formed by interference points (for example, scanning arcs formed by ground undulations, random false detections caused by environmental factors, etc.), the present invention is based on the similarity of data, and realizes the classification of straight line segments through K-Means clustering, horizontal analysis and vertical analysis, and RANSAC-based screening, and selectively retains stable and valuable characteristic straight line segments.
[0103] K-Means clustering is a commonly used iterative clustering analysis method that can classify similar data with a known number of categories. An important assumption of using K-Means is that the similarity between data can be measured by Euclidean distance directly or through conversion. The specific steps are as follows:
[0104] Input: initial clustering samples and number of clusters K.
[0105] Output: final clustering results and the cluster center of each group.
[0106] Step 1: Cluster initialization. Select K data objects as the initial cluster centers. The selection of K cluster centers will have a decisive influence on the clustering results, so the clustering effect can be improved by designing the initial cluster centers.
[0107] Step 2: Set the iteration termination condition. Usually the maximum number of cycles or the change error of the cluster center is selected as the clustering termination condition.
[0108] Step 3: Update the class to which the sample object belongs. Calculate the distance between each data object and the cluster center, and assign the data object to the class with the closest distance according to the distance optimization criterion.
[0109] Step 4: Update the cluster center. Each time a sample object is assigned, the cluster center will calculate the average vector based on the existing objects in the cluster and use it as the cluster center for the next iteration.
[0110] Step 5: Repeat steps 3-4 and stop when the iteration termination condition in step 2 is met.
[0111] The K-Means algorithm has the advantages of simplicity and high efficiency. It is the most widely used clustering algorithm and meets the requirements of the present invention for clustering algorithms. Considering the operation efficiency of the algorithm, the present invention uses the K-Means algorithm to perform a total of two clustering analyses on the characteristic straight line segments according to different clustering criteria. The basic process of the algorithm is as follows: Figure 6 shown.
[0112] After two K-Means clusterings, the straight line segments are simply classified according to similar properties. Obviously, simply clustering cannot achieve an ideal data classification. The present invention further screens effective straight line segments by analyzing the spatial positions of the straight line segments horizontally and vertically.
[0113] Specifically:
[0114] (a) Horizontal analysis
[0115] Transverse analysis refers to the integration, reconstruction, and screening of straight line segments based on the radar's transverse scan lines. In the process of extracting characteristic straight lines, a road boundary may be divided into several relatively short and fragmented lines due to factors such as the roughness of the road boundary in a complex environment and the obstruction of gravel and soil. On the other hand, due to the scanning characteristics of the lidar, a scan line with a large radius will form a scanning arc that approximates a straight line segment on the far road surface. The purpose of transverse analysis is to reduce the number of straight line segments by integrating and reconstructing the short and fragmented straight line segments on the same scan line, while screening out some useless straight line segments that do not belong to the road boundary.
[0116] The horizontal analysis process uses the cluster of clustering results as the basic unit, and merges short and fragmented straight line segments with similar distances and located on the same scan line in each cluster. The merged straight line segments have the two farthest points as endpoints, and the direction is consistent with the laser radar scanning direction. Then, according to the characteristic straight line extraction algorithm set by S1-S3, the reconstructed straight line is judged again, and the straight line segments that meet the standards are retained.
[0117] (b) Longitudinal analysis
[0118] The longitudinal analysis also uses the cluster of C2 clustering results as the basic unit. The main purpose is to roughly screen all the straight line segments in the cluster and eliminate the false detection straight line segments with obvious errors. After two K-Means clustering, we simply classified the characteristic straight line segments with similarity. For cluster i, the longitudinal analysis needs to re-count the average inclination information in each cluster based on the clustering result C2, which is recorded as Any straight line segment in cluster i The weight of the current straight line segment needs to be calculated by the following formula Where n is the number of straight line segments in a cluster and k is an integer.
[0119]
[0120] if Weight of line segment Less than the given threshold W th If the straight line segment does not meet the road boundary evaluation criteria through longitudinal analysis, it is discarded. Otherwise, it is retained.
[0121] In the road boundary fitting process in S5, the road boundary is fitted by cubic B-spline interpolation, and the interpolated and fitted road boundary is optimized by inter-cluster boundary smoothing to obtain the road boundary of the passable area of the substation.
[0122] In a specific embodiment, cubic B-spline interpolation fits the road boundary. Since the road shape is usually expressed as a straight line or a standard curve, it is necessary to use a road boundary model to represent the road boundary information. Road model knowledge can transform road boundary detection into a problem of solving a road model. Usually, the expression method of the road model is based on linear, parabolic and hyperbolic road models. However, this type of model has poor adaptability, and a slight change in the shape of the road boundary will cause a drastic change in the model parameters. Under complex outdoor environmental conditions, these changes in parameters may affect the overall stability of the model. The changes in the cubic B-spline model parameters between adjacent frames are relatively stable, and have local manipulability, can be composed into any shape, and can describe a wider range of road boundary conditions than other models. Based on the above-mentioned excellent characteristics of the model, the present invention uses a cubic uniform B-spline curve model to fit the road boundary.
[0123] Assumption C i (i=0, 1, 2, 3, ..., n) are the n+1 control points of the B-spline curve, and the expression of the B-spline curve can be obtained as follows:
[0124]
[0125] In the above formula, k is the order of the B-spline curve, N i,k(t) is the k-order B-spline basis function, which can be defined by the recursive formula as follows:
[0126]
[0127] U=[u0,u1,u2,u3,...,u n+k+1 ] is the node vector of the k-order B-spline function. If the nodes are evenly distributed along the parameter axis, the B-spline curve is uniform; if the nodes along the parameter axis are unevenly distributed, the B-spline curve is uneven. When constructing a curve, the time complexity of using a non-uniform B-spline function is greater than that of a uniform B-spline function, and there is no significant difference in the effects of the two. And as the order of the spline curve increases, the number of control points required increases, which is not conducive to local control. Therefore, in the present invention, a cubic uniform B-spline curve model is used to fit the road boundary.
[0128] The cubic B-spline basis function is as follows:
[0129]
[0130] The present invention performs interpolation fitting of road boundaries based on the above conclusions. Boundary fitting is also performed in clusters, and the starting point, middle point, and end point of each straight line in the cluster are extracted in turn as feature points. If type = 1, the 3k points obtained are sorted from small to large according to the y coordinate, otherwise they are sorted from small to large according to the x coordinate, and the result is expressed as. The purpose of this is to obtain a smoother boundary curve. The above derivation and inverse control points are performed on the arranged road boundary point sequence as interpolation points, and finally 3k-1 cubic B-spline fitting curves formed by interpolation of 3k points in the cluster are obtained through the obtained control points, which are the road boundaries.
[0131] Specifically, the inter-cluster boundary smoothing process is as follows: the fitting method is performed in clusters, which results in discontinuous fitting curves between clusters. In a straight road environment, most of the road boundary straight line segments on each side are clustered into one category, and there is no need for special processing of the curves between clusters. However, when passing through a curved road environment, there are cases where the road boundary straight line segments on the same side are divided into different categories. In response to this problem, the present invention further adjusts the road boundary curves formed between different clusters. In S1-S3, the optimal straight line model for RANSAC screening is obtained for each cluster. For two clusters located on the same road boundary in a curved environment, if the two RANSAC straight line models have an intersection within a close range of the endpoints of the curve part, it is necessary to smooth the fitting curves of the two clusters. Regarding boundary smoothing, the end point of the previous curve can be and the starting point of the next curve Arrange them in order as a known interpolation sequence, that is, add a new sequence at the cluster discontinuity to form a cubic B-spline fitting curve, such as Figure 8 The red curve is shown in Figure 2. After inter-cluster smoothing, the algorithm's adaptability to environments such as curves is improved, and the robustness of the detection system is enhanced.
[0132] The semantic segmentation process in S6 specifically includes:
[0133] S61: The segmentation process of the point cloud data is restricted by using a smooth threshold as a segmentation condition, and an improved Euclidean clustering segmentation is performed;
[0134] S62: Perform missing optimization, feature position encoding, network learning and normalization on the results after clustering segmentation.
[0135] In a specific embodiment, the improved Euclidean clustering segmentation process in S61 is specifically as follows:
[0136] First, the planes in the scene are segmented by the R ANSAC algorithm to remove a large number of irrelevant environmental point clouds, but it is still impossible to determine how many objects to be tested exist in the scene, and to which cluster each point in the space belongs. The essence of Euclidean clustering is to classify points with similar distances into one category. Assuming that there are n points in the point cloud C, the Euclidean distance is defined as the degree of closeness between two points, and the distance between neighboring points is used as the standard to achieve point cloud clustering segmentation. The specific segmentation process is: for the preprocessed point cloud data set P, determine a query point P i , set the distance threshold r, where the calculation formula of the threshold r is:
[0137]
[0138] Where: p i With p j are two points in the point cloud; p ik With p jk is any point in the neighborhood of two points.
[0139] Find the n nearest neighbor points P through KD-Tree j (j=1, 2, 3, ..., n); Calculate the Euclidean distance d from the n nearest neighbor points to the query point according to the above formula j ; Compare distance d j With the distance threshold r, the points less than r are classified into class M. The segmentation is completed when the number of points in class M no longer increases.
[0140] The traditional Euclidean clustering method has good real-time segmentation performance, but it is prone to under-segmentation and over-segmentation. To address this problem, the present invention improves the traditional Euclidean clustering algorithm and uses a smooth threshold as a segmentation condition to limit the segmentation process, thereby preventing incorrect clustering while retaining good real-time performance. The smoothness parameter is used to describe the degree of change of the object surface. For the query point Pi The smoothness parameter of is defined as:
[0141]
[0142] Where: X i P i The position in the camera coordinate system; S is the spherical neighborhood with R as the radius; N is the total number of points in the dataset within the neighborhood.
[0143] When the query point is located on a smooth surface, the neighborhood points are symmetrically distributed, so the smoothness c is small; however, when the query point is located on the edge, there is only one direction in the area, and the smoothness c is large. Therefore, by judging the size of the smoothness, it can be determined whether the query point is an edge point. The algorithm flow chart is as follows Fig. 9 shown.
[0144] For the acquired point cloud information, semantic segmentation of point cloud can be performed and common entities in substations such as transformers can be abstracted into data models. With the advent of the information age, research in the field of artificial intelligence has become a general trend. How to recognize objects in complex scenes is a very important research topic in deep learning. Point cloud semantic segmentation is an important means of analyzing and understanding three-dimensional scenes, and has broad application prospects and research value in many fields. The rapid development of three-dimensional point cloud data acquisition technology is very beneficial to point cloud segmentation.
[0145] In recent years, with the rapid development of deep learning and three-dimensional point cloud data acquisition technology, the use of deep learning to directly perform semantic segmentation on point cloud data has received more attention. Methods based on deep learning can extract deep semantic information of point clouds and mine important features between data. However, many deep neural networks currently do not make better use of local information between point clouds, cannot process larger-scale point cloud data, and cannot deal with the problem of partial missing point cloud data, resulting in low segmentation accuracy. In response to the above problems, focusing on the extraction of local features of three-dimensional point clouds, there is a multi-level point cloud semantic segmentation network based on an effective local feature extraction module, which is optimized and improved for the problem of missing point clouds. (1) Construction of local feature extraction module. For large-scale unordered point cloud data, the present invention establishes a Kd-tree data structure organization for it. The Kdtree data structure is as follows: Fig.10 as shown.
[0146] The KNN algorithm is used to find the nearest K neighboring points in the Euclidean space of each point, and redundant information is added to each point by encoding the local area. The features of the neighboring points output above are aggregated together to obtain the aggregated features of each point, thereby improving the receptive field of each point. (2) Construction of semantic segmentation network. The present invention uses a general autoencoder network structure and stacks multiple local feature extraction modules to significantly increase the receptive field of each point. By analyzing the characteristics of the FPS, IDIS and RS sampling methods, the RS sampling method is used to downsample large-scale point cloud data, extract point cloud feature information in a hierarchical manner and splice it through a jump cascade method, integrate multi-resolution feature information, make full use of the features of shallow and deep networks, enrich the features learned by the overall network, and improve the accuracy of predicting semantic categories. (3) Point cloud completion and sampling method adjustment. Considering the impact of partial missing point cloud data on the overall point cloud semantic segmentation, a method for completing point cloud data is proposed. First, the boundary of an object is extracted, all internal hole boundaries are found, and points are evenly scattered on the internal holes to complete the holes. At the same time, various clustering methods are analyzed, and an improved sampling method based on K-means clustering is proposed. By using K-means clustering point cloud different feature information multiple times, the point cloud data sampled by the network is made more uniform.
[0147] Feature position encoding based on feature distance Specifically: There are many feature extraction methods suitable for point clouds in the current point cloud learning field. Some methods only process point cloud coordinates, while other methods can learn the features of point cloud normals, colors, and other attributes. The above methods extract the intrinsic features of point clouds as much as possible. In addition, since the introduction of PointNet++, sampling and grouping have become effective means for point cloud feature extraction and reducing network calculations. Point cloud sampling obtains key points. After grouping, the point cloud is divided into multiple local areas. The extraction of local area features is not involved in previous methods. However, the features of local areas still cannot well represent the associated features inside the point cloud. The connection between points is abstracted into the global features and is not extracted independently. Edge convolution (EdgeConv) is introduced into TransformerNet. The PCT network learns features through sampling and grouping methods similar to PointNet++, and encodes the association between features with reference to the spatial position of the point cloud. Compared with PCT, edge convolution can abstract the correlation of point clouds in high-dimensional feature space. The connection between points can be directly abstracted as edges in the graph, which can better abstract the connection between points. Therefore, the feature distance encoding (Feature Distance Encoding) based on the present invention uses edge convolution to abstract point cloud features.
[0148] like Fig.11As shown in the figure, the larger blue dots are key point features, the smaller blue dots are neighboring point features, and the purple dotted lines connecting the dots represent associated features. Edge convolution uses the KNN algorithm to query the features of key points and neighboring points, and constructs the local area into a graph. The points are the nodes of the graph, and the associations between the points are the edges of the graph. Finally, the edges of the graph are convolved. Assume that the input point cloud is P = {p1, p2, p3, ..., p n}∈R d , d is the feature dimension of the current point cloud (the d defined here is the d in the R superscript, which is different from the d in the above mutual interval d), the local area is represented as a directed graph G = (V, E), the feature of each point is the vertex, and the edge E of the directed graph is composed of the relationship between points. More precisely, for each point p in P i Query the k points p closest to the feature i,j , define the directed edge as E = {e i,j}, edge feature e i,j The definition is shown in formula 1. Where h θ A layer of perceptron is used to extract local area features, and finally the local features are aggregated to the features of the current point through the maximum pooling function, as shown in equations (2) and (3).
[0149] e i,j =h θ (p i , p j -p i ) (1)
[0150] e′ i,j =ReLU(θ m ·(p j -p i )+φ m ·p i ) (2)
[0151] p′ i =max e′ i,j (3)
[0152] Transformer-based point cloud learning network: TransformerNet network structure is as follows: Fig.12 As shown, the network is divided into a classification network and a segmentation network. Fig.12The left side is the encoder, the middle is the segmentation network decoder, and the right side is the classification network encoder. The TransformerNet encoder consists of input embedding (InputEmbedding), feature position encoding (Feature Position Encoding), connection attention (Connect Attention), concatenation and regularization (Concat&Norm) and LBR. The connection attention consists of four layers of attention (AT) layers. L1 Norm is applied to the AT layer. Therefore, the AT layer first scales the features, then normalizes the first dimension of the features through Softmax, and finally normalizes the second dimension of the features through L1 Norm. The calculation process is shown in equations (4), (5), (6), and (7).
[0153]
[0154]
[0155] Four AT layers are connected in series to form a connection attention module, followed by a Concat&Norm module to concatenate and normalize the feature input and output, such as Fig.13 shown.
[0156] For a given point cloud feature input The output of each self-attention layer in the connection attention is the input of the next layer. The connection attention calculation process is shown in the formula.
[0157] F1=AT 1 (F in ) (8)
[0158] F i =AT 2 (F i--1 ), i = 2, 3, 4 (9)
[0159] F out =Norm(concat(F in ,F1,F2,F3,F4)) (10)
[0160] Normalization is one of the means to avoid network overfitting. For the input point cloud data F = f i,j,k ∈R b×n×d, where b is the batchsize, n is the number of points, and d is the feature dimension. LayerNorm normalizes the features of each point, calculates the mean and variance of the features of each point, and normalizes the features of the point. In the network, the self-attention mechanism calculates and extracts the internal correlation of the features of each point, while LayerNorm normalizes the internal features of this feature. The two complement each other. Feature position encoding calculates the connection between points, while BatchNorm normalizes the features of all points in a batch. Therefore, BatchNorm is applied in feature position encoding to avoid problems such as extreme values or vanishing gradients of the feature values of the associated features.
[0161] Finally, by integrating the road boundaries of the substation's traversable area and the feature point set of each object in the substation, the objects in the substation can be positioned.
[0162] Embodiment 2:
[0163] The present invention provides a substation object positioning system, the object positioning system comprising the following modules:
[0164] A point cloud data acquisition module is used to use a drone, an unmanned vehicle, or a combination of a drone and an unmanned vehicle to perform a laser radar scan of the substation in a traversable area to obtain point cloud data of the substation;
[0165] A feature point set extraction module is used to extract feature point sets containing straight line features from point cloud data;
[0166] The feature line set acquisition module is used to acquire a set of feature lines through a lateral candidate feature point set and a radial candidate feature point set.
[0167] An effective road boundary straight line segment acquisition module is used to cluster and screen a set of characteristic straight lines to obtain effective road boundary straight line segments;
[0168] A road boundary fitting module is used to fit the road boundary through valid road boundary straight line segments to obtain the road boundary of the accessible area of the substation;
[0169] The object feature point acquisition module is used to perform point cloud semantic segmentation on the point cloud data of the substation based on the point cloud segmentation network of the point cloud transformer and extract feature points of the physical objects in the substation to obtain the feature point set of each object in the substation;
[0170] The object positioning module is used to integrate the road boundaries of the traversable area of the substation and the feature point set of each object in the substation to complete the positioning of objects in the substation.
[0171] Embodiment 3:
[0172] See also Fig.14 As shown, the present invention also provides an electronic device 100; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.
[0173] The memory 101 can be used to store the computer program 103, and the processor 102 implements the steps of the method for locating objects in a substation described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.
[0174] The memory 101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.
[0175] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.
[0176] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a method for locating an object in a substation, and the processor 102 can execute the plurality of instructions to implement:
[0177] S1, using drones, unmanned vehicles, or a combination of drones and unmanned vehicles, to perform lidar scanning of the substation in the traversable area to obtain point cloud data of the substation;
[0178] S2, extracting a feature point set containing straight line features from the point cloud data;
[0179] S3, obtaining a set of feature lines through a lateral candidate feature point set and a radial candidate feature point set,
[0180] S4, clustering and screening the set of characteristic straight lines to obtain valid road boundary straight line segments;
[0181] S5, performing road boundary fitting through effective road boundary straight line segments to obtain the road boundary of the accessible area of the substation;
[0182] S6, a point cloud segmentation network based on point cloud transformer performs point cloud semantic segmentation on the point cloud data of the substation and extracts feature points of physical objects in the substation to obtain a set of feature points of each object in the substation;
[0183] S7, integrating the road boundary of the traversable area of the substation and the feature point set of each object in the substation, completes the positioning of the objects in the substation.
[0184] Embodiment 4:
[0185] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, the steps of each of the above-mentioned method embodiments can be implemented.
[0186] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory and read-only memory (ROM).
[0187] The upgraded application of the lightweight substation three-dimensional model of the present invention eliminates the tedious operations such as manual splicing and patching of the substation three-dimensional model, and combines positioning technology to realize remote perception and precise control of the attributes and behaviors of personnel on the work site, and realizes full perception and control management of permissions, behaviors, information, and materials, thereby improving the level of safety and quality control. Artificial intelligence digital auxiliary means are used to improve the rigor and adaptability of security layout plans, avoiding possible human errors and the resulting safety hazards. Researching and developing automated and intelligent means to achieve lightweight three-dimensional modeling of substations and transmission lines not only has important scientific and technological value for the scientific and technological progress of the power industry, but also has very important significance for improving the reliability and scientific management level of power system operation, improving the comprehensive intelligent management level of power grids, and bringing huge economic and social benefits.
[0188] The above is a detailed introduction to a method, system, device and medium for locating objects in a substation provided by the embodiments of the present application. The description of the above embodiments is only used to help understand the method and core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
[0189] For example, certain words are used in the specification and claims to refer to specific components. Those skilled in the art should understand that hardware manufacturers may use different nouns to refer to the same component. This specification and claims do not use differences in names as a way to distinguish components, but use differences in the functions of components as the criteria for distinction. As mentioned throughout the specification and claims, "including" and "comprising" are open-ended terms, so they should be interpreted as "including / including but not limited to". "Approximately" means that within an acceptable error range, those skilled in the art can solve the technical problem within a certain error range and basically achieve the technical effect. The subsequent description of the specification is a preferred embodiment of the present application, but the description is for the purpose of illustrating the general principles of the present application, and is not used to limit the scope of the present application. The scope of protection of the present application shall be determined by the definition of the attached claims.
[0190] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a product or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a product or system. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the product or system including the elements.
[0191] It should be understood that the term "and / or" used in the present invention is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects before and after are in an "or" relationship.
[0192] The above description shows and describes several preferred embodiments of the present application, but as mentioned above, it should be understood that the present application is not limited to the form disclosed in the present invention, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the application concept described in the present invention through the above teachings or the technology or knowledge in the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present application, and should be within the scope of protection of the claims attached to the present application.
Claims
1. A method for locating objects in a substation, characterized in that: The object positioning method comprises: S1, perform lidar scanning in the traversable area of the substation to obtain point cloud data of the substation; S2, extracting a feature point set containing straight line features from the point cloud data; S3, obtain a set of characteristic straight lines through a set of characteristic points containing straight line features, S4, clustering and screening the set of characteristic straight lines to obtain valid road boundary straight line segments; S5, performing road boundary fitting through effective road boundary straight line segments to obtain the road boundary of the accessible area of the substation; S6, performing point cloud semantic segmentation on the point cloud data of the substation through a point cloud segmentation network and extracting feature points of physical objects in the substation to obtain a feature point set of each object in the substation; S7, integrating the road boundary of the traversable area of the substation and the feature point set of each object in the substation, completes the positioning of the objects in the substation.
2. The method for locating objects in a substation according to claim 1, characterized in that: The S2 specifically includes: Each scan line of the laser radar is used as a horizontal index, and three consecutive points on the same scan line that meet the preset constraints are used as a set of horizontal candidate feature points; The radial angle change between adjacent laser radar scanning lines is used as a constraint to obtain a set of radial candidate feature points.
3. The method for locating objects in a substation according to claim 2, characterized in that: The constraints preset in the process of acquiring the horizontal candidate feature point set in S2 are specifically: Replace three consecutive points with three points separated by a distance of d, project the point cloud onto the XOY plane, and use the angle relationship of the three points to establish constraint conditions for feature point extraction; The three points on the same scan line l are p l,i-d ,p l,i ,p l,i+d , the coordinates are (x l,i-d ,y l,i-d ), (x l,i ,y l,i ), (x l,i+d ,y l,i+d ) Every two adjacent points can determine a vector, defining α i,j The positional relationship of the three points is described as a vector angle as follows: Under flat conditions such as road surfaces, α i,j The value is very small, close to 0°, at the road boundary, α i,j Close to 90°, select threshold α th =30°, d=5, keep [p l,i-d ,p l,i+d ] interval, the coefficient α l,k The largest point p l,k As feature points, id≤k≤i+d, a set of horizontal candidate feature points is constructed by traversing each scan line; The constraint conditions for the radial angle change in S2 are specifically: Preset p l-1,i ,p l,i are two points located on adjacent scan lines with the same azimuth, and their coordinates are (x l-1,i ,y l-1,i ,z l-1,i ), (x l,i ,y l,i ,z l,i ), let the radial angle between the two points be Take β th =20°, when p l,i When located at the road boundary, the radial angle between adjacent scan lines compared to a flat road surface Changes occur if It is believed that p l,i is a qualified feature point. For each set of radial data, only the first point that meets the constraint conditions is extracted from the inside to the outside as the feature point. Each set of radial data is traversed to construct a radial candidate feature point set.
4. The method for locating objects in a substation according to claim 1, characterized in that: The S3 specifically includes: S31: The horizontal candidate feature point set is based on each scanning line, and any two adjacent feature points are used as the first and last endpoints to define a discrete curve. A preset constraint method is executed for each discrete curve to retain the horizontal feature straight line set that meets the constraint conditions. S32: The radial candidate feature point set is first discretized, and then, taking each radial line as a unit, any two adjacent feature points are used as the first and last endpoints to define a discrete curve, and a preset constraint method is executed for each discrete curve to retain a radial feature straight line set that meets the constraint conditions; S33: Fusing and summarizing the transverse feature straight line set and the radial feature straight line set to obtain a set of feature straight lines.
5. The method for locating objects in a substation according to claim 1, characterized in that: The clustering and screening method in S4 achieves the classification of straight line segments through K-Means clustering, horizontal analysis and vertical analysis, and RANSAC-based screening, and selectively retains characteristic straight line segments as valid road boundary straight line segments.
6. The method for locating objects in a substation according to claim 1, characterized in that: In the road boundary fitting process in S5, the road boundary is fitted by cubic B-spline interpolation, and the interpolated and fitted road boundary is optimized by inter-cluster boundary smoothing to obtain the road boundary of the passable area of the substation.
7. The method for locating objects in a substation according to claim 1, characterized in that: The semantic segmentation process in S6 specifically includes: S61: The segmentation process of the point cloud data is restricted by using a smooth threshold as a segmentation condition, and an improved Euclidean clustering segmentation is performed; S62: Perform missing optimization, feature position encoding, network learning and normalization on the results after clustering segmentation.
8. A substation object positioning system, characterized in that: The object positioning system includes the following modules: The point cloud data acquisition module is used to perform laser radar scanning in the traversable area of the substation to obtain the point cloud data of the substation; A feature point set extraction module is used to extract feature point sets containing straight line features from point cloud data; The feature line set acquisition module acquires the feature line set through the feature point set containing the line features. An effective road boundary straight line segment acquisition module is used to cluster and screen a set of characteristic straight lines to obtain effective road boundary straight line segments; A road boundary fitting module is used to fit the road boundary through valid road boundary straight line segments to obtain the road boundary of the accessible area of the substation; The object feature point acquisition module is used to perform point cloud semantic segmentation on the point cloud data of the substation through the point cloud segmentation network and extract feature points of the physical objects in the substation to obtain the feature point set of each object in the substation; The object positioning module is used to integrate the road boundaries of the traversable area of the substation and the feature point set of each object in the substation to complete the positioning of objects in the substation.
9. An electronic device, characterized in that: It comprises a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program, and when the computer program is executed by the processor, the steps of a method for locating objects in a substation are implemented as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a method for locating an object in a substation as claimed in any one of claims 1 to 7 are implemented.
