Method for automatically connecting power transmission tower points in remote sensing image based on hidden markov model

Abstract

This application relates to an automatic connection method for transmission tower points in remote sensing imagery based on a Hidden Markov Model (HMM). The method includes steps such as preprocessing the raw data of the transmission tower point set, feature extraction, HMM modeling, path decoding and generation, post-processing of engineering constraints, path optimization, and interactive verification. Outliers are filtered using a density clustering algorithm, observed feature vectors are calculated, the connection pattern of the transmission tower points is modeled using an HMM, the optimal state sequence is solved using the Viterbi algorithm to generate preliminary connection paths, post-processing of engineering constraints is performed, and finally, path optimization and manual interactive verification are conducted. This method integrates HMM probabilistic modeling and spatial features to improve connection accuracy, meet engineering specifications, and continuously optimize model parameters through manual interactive correction and incremental learning mechanisms, thereby improving connection efficiency and quality and promoting the development of power line planning technology.

Description

Technical Field

[0001] This invention belongs to the field of power line planning technology and relates to an automatic connection method for transmission tower points in remote sensing images based on hidden Markov models. Background Technology

[0002] In the early stages of power transmission line planning, the connection of transmission towers on maps was mainly done manually. This method was not only inefficient but also susceptible to human factors, leading to significant differences in the accuracy and rationality of the connection results.

[0003] With technological advancements, several automatic connection methods based on simple algorithms have emerged, such as the minimum spanning tree method and the nearest neighbor method. However, these methods present numerous problems in practical applications. While the minimum spanning tree method can generate a tree-like network connecting all transmission towers, it fails to adequately consider the complexity of the terrain and engineering constraints of transmission lines, easily generating connection paths that do not meet actual construction requirements. The nearest neighbor method only considers the local proximity relationships between transmission towers, ignoring the overall route of the transmission line and terrain factors, potentially leading to frequent turns and detours in the connection path, increasing construction difficulty and cost.

[0004] The design and construction of power transmission lines must comply with a series of engineering specifications and constraints, such as the span, turning angle, and slope between transmission towers. However, most existing wiring methods fail to fully consider these engineering constraints, resulting in the need for extensive subsequent adjustments to the generated wiring paths to meet actual engineering requirements, thus increasing the time and cost of design and construction.

[0005] Remote sensing imagery data contains rich geographic information and spatial distribution characteristics of transmission towers. However, existing methods for connecting transmission towers often fail to fully utilize this information, merely using remote sensing imagery as a background layer for simple visualization, without delving into the useful data to guide the connection work. Summary of the Invention

[0006] To address the problems existing in the traditional methods mentioned above, this invention proposes an automatic connection method for transmission tower points in remote sensing images based on hidden Markov models. This method can fully utilize the geographical information and spatial features in remote sensing image data, reduce manual intervention, improve the accuracy and efficiency of connection, and reduce the probability of errors.

[0007] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions: On the one hand, an automatic connection method for transmission tower points in remote sensing images based on hidden Markov models is provided, the method including the following steps: Obtain the transmission tower point set, which includes information on multiple transmission tower points, including the longitude, latitude, and altitude of the tower's location.

[0008] A spatial index is constructed on the original data of the transmission tower point set, and outliers are filtered out using a density clustering algorithm to obtain the preprocessed transmission tower point set.

[0009] Based on the preprocessed set of transmission tower points, the observation feature vector of each transmission tower point is extracted.

[0010] Based on the observed feature vectors, power grid topology, and geographic information and spatial features in remote sensing image data, the parameters of the Hidden Markov Model are constructed.

[0011] Based on the spatial distribution of the transmission tower point set and the power grid topology, the starting point of the transmission line is determined. Starting from the starting point, the next transmission tower point is determined sequentially according to spatial proximity, generating an ordered observation sequence.

[0012] Based on the ordered observation sequence and the parameters of the hidden Markov model, the optimal state sequence is solved to generate the preliminary connection path.

[0013] Based on the initial connection path and engineering specifications, compliance adjustments were made to obtain the adjusted path.

[0014] The adjusted path is simplified and smoothed, and the smoothed path result and geographic information data are loaded into the GIS platform. The path is then registered and displayed by manual dragging and dropping, and the parameters of the Hidden Markov Model are updated to obtain the final optimized path.

[0015] One of the above technical solutions has the following advantages and beneficial effects: The aforementioned method for automatically connecting transmission tower points in remote sensing imagery based on Hidden Markov Models (HMMs) includes steps such as preprocessing the raw data of the transmission tower point set, feature extraction, HMM modeling, path decoding and generation, post-processing of engineering constraints, path optimization, and interactive verification. Outliers are filtered using density clustering algorithms, observed feature vectors are calculated, connection patterns of transmission tower points are modeled using HMMs, the Viterbi algorithm is used to solve for the optimal state sequence to generate preliminary connection paths, post-processing of engineering constraints is performed, and finally, path optimization and manual interactive verification are conducted. This method integrates HMM probabilistic modeling and spatial features to improve connection accuracy, meet engineering specifications, and continuously optimize model parameters through manual interactive correction and incremental learning mechanisms, thereby improving connection efficiency and quality and promoting the development of power line planning technology. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating an automatic connection method for transmission tower points in remote sensing images based on a hidden Markov model, as shown in one embodiment. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0020] It should be noted that, in this document, the reference to "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The presentation of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will understand that the embodiments described herein can be combined with other embodiments. The term "and / or" as used herein refers to any combination of one or more of the associated listed items, and all possible combinations, including such combinations.

[0021] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] In one embodiment, such as Figure 1 As shown, an automatic connection method for transmission tower points in remote sensing images based on a hidden Markov model is provided, which may include the following processing steps 100 to 114: Step 100: Obtain the transmission tower point set, which includes information on multiple transmission tower points, including the longitude, latitude, and altitude of the transmission tower's location.

[0023] Specifically, the transmission tower point set is a point set P = { containing longitude, latitude, and altitude.} }

[0024] Step 102: Construct a spatial index for the original data of the transmission tower point set, and use a density clustering algorithm to filter out outliers to obtain the preprocessed transmission tower point set.

[0025] Specifically, a spatial index is constructed, and a density clustering algorithm is used to filter out outliers and remove those with neighborhood radii. r The number of interior points is less than k The transmission tower points are obtained by normalizing the geographic coordinate data, calculating the Euclidean distance between adjacent towers and the terrain complexity parameters, and obtaining the preprocessed transmission tower point set.

[0026] Step 104: Extract the observation feature vector of each transmission tower point based on the preprocessed transmission tower point set.

[0027] Specifically, the observed feature vector of each transmission tower point is calculated based on the preprocessed transmission tower point set. ,in It is the direction angle. The nearest neighbor distance. For local density, For elevation gradient, For consistency of direction.

[0028] Step 106: Construct Hidden Markov Model parameters based on the observed feature vectors, power grid topology, and geographic information and spatial features in remote sensing image data.

[0029] Specifically, the parameters of the Hidden Markov Model (HMM) are defined as follows:

[0030] Where S is the state space, used to represent the possible connection modes of the transmission tower points, including but not limited to: straight extension, left-side obstacle avoidance, right-side obstacle avoidance, clockwise turning connection, and counterclockwise turning connection; V is the observation space, representing the transmission tower coordinates and feature vectors. ; Let be the state transition probability, and let the state transition probability satisfy: ,in To calculate the cost of terrain barriers based on a digital elevation model (DEM); For the probability of launch, The feature distribution is fitted using a Gaussian mixture model; initial probability. The starting point state distribution is determined based on the power grid topology.

[0031] Step 108: Based on the spatial distribution of the transmission tower point set and the power grid topology, determine the starting point of the transmission line. Starting from the starting point, determine the next transmission tower point in sequence according to spatial proximity to generate an ordered observation sequence.

[0032] Specifically, based on the spatial distribution of transmission tower points and the starting point of the power grid topology, an ordered observation sequence is generated based on spatial proximity. .

[0033] Step 110: Based on the ordered observation sequence and the parameters of the hidden Markov model, solve for the optimal state sequence and generate the preliminary connection path.

[0034] Specifically, the Viterbi algorithm with distance constraints is used to solve for the optimal state sequence. The state transition distance is limited to a distance of less than 2×Euclidean distance between adjacent transmission tower points. Within the range; output the initial connection path. 。

[0035] Step 112: Based on the initial connection path and engineering specifications, make compliance adjustments to obtain the adjusted path.

[0036] Specifically, the cost function for route planning is constructed as follows:

[0037] use Algorithm pair Make local adjustments to ensure that the span and turning angle comply with GB 50545-2010 "Design Specification for 110kV~750kV Transmission Lines with Space".

[0038] Step 114: Simplify and smooth the adjusted path, load the smoothed path results and geographic information data into the GIS platform, register and display them by manual dragging, and update the parameters of the Hidden Markov Model to obtain the final optimized path.

[0039] Specifically, based on the adjusted path, geometric simplification and manual interactive optimization are performed. Specifically: the path is simplified using the Douglas-Peucker algorithm and smoothed using B-spline curves; the optimized path results are overlaid and displayed on the GIS platform, supporting manual drag-and-drop adjustments of specific sections; the manually corrected and confirmed path data is added to the HMM training set for incremental updates of model parameters, especially the state transition probability matrix A and the emission probability matrix B, enhancing the model's adaptive capability.

[0040] The aforementioned method for automatically connecting transmission tower points in remote sensing imagery based on Hidden Markov Models (HMMs) includes steps such as preprocessing the raw data of the transmission tower point set, feature extraction, HMM modeling, path decoding and generation, post-processing of engineering constraints, path optimization, and interactive verification. Outliers are filtered using density clustering algorithms, observed feature vectors are calculated, connection patterns of transmission tower points are modeled using HMMs, the Viterbi algorithm is used to solve for the optimal state sequence to generate preliminary connection paths, post-processing of engineering constraints is performed, and finally, path optimization and manual interactive verification are conducted. This method integrates HMM probabilistic modeling and spatial features to improve connection accuracy, meet engineering specifications, and continuously optimize model parameters through manual interactive correction and incremental learning mechanisms, thereby improving connection efficiency and quality and promoting the development of power line planning technology.

[0041] In one embodiment, step 102 includes: using an R-tree based on a hierarchical index as a spatial index to organize and query spatial data; determining the neighborhood radius r and the number of points threshold k based on the density, geographical range, and business requirements of the actual transmission tower distribution; traversing each transmission tower point p in the transmission tower point set P, with transmission tower point p as the center and neighborhood radius r as the radius, querying all transmission tower points within the neighborhood in the R-tree spatial index, and counting the number of points within the neighborhood; for each transmission tower point p, if the number of points within its neighborhood is less than the number of points threshold k, then marking the point as an outlier; after all points have been traversed, removing all marked outliers from the transmission tower point set P to obtain the preprocessed transmission tower point set.

[0042] Specifically, an R-tree based on a hierarchical index is chosen as the spatial index to organize and query spatial data. The neighborhood radius *r* and the point count threshold *k* are determined based on the density of the actual transmission tower distribution, geographical range, and business requirements for density clustering calculations. For example, in mountainous areas where transmission towers are relatively sparsely distributed, the neighborhood radius *r* can be appropriately increased and the point count threshold *k* decreased; while in densely distributed areas such as plains, the neighborhood radius *r* can be decreased and the point count threshold *k* increased. Each point *p* in the transmission tower point set *P* is traversed. Using point *p* as the center and a neighborhood radius *r* as the radius, all transmission tower points within that neighborhood are queried in the R-tree spatial index, and the number of points within the neighborhood is counted. For each point *p*, if the number of points within its neighborhood is less than *k*, the point is marked as an outlier. After all points have been traversed, all marked outliers are removed from the transmission tower point set *P*, resulting in the transmission tower point set *P'* after density clustering filtering.

[0043] In one embodiment, the observed feature vector includes: orientation angle, nearest neighbor distance, local density, gradient, and orientation consistency; step 104 includes: calculating the orientation angle between each transmission tower point (excluding the last transmission tower point) and its next adjacent point in the preprocessed transmission tower point set; the range of the orientation angle is... For the last transmission tower point in the preprocessed transmission tower point set, take the direction angle from the previous transmission tower point to the current point; accelerate the search using KD-Tree, query the nearest neighbor point of each transmission tower point in the preprocessed transmission tower point set and calculate the Euclidean distance to obtain the nearest neighbor distance; use the number of neighboring points within the neighborhood radius r as the local density; use the elevation change rate per unit distance between neighboring points as the elevation gradient; use the cosine of the angle between the current direction and the previous direction as the direction consistency.

[0044] Specifically, the geographic coordinate data is normalized, and the Euclidean distance between adjacent towers and the terrain complexity parameters are calculated; this includes the following steps: Step 1041: Determine that the normalization range of the three dimensions of longitude, latitude, and altitude is in the interval [0,1]. Perform normalization calculations for the longitude, latitude, and altitude of each point in the transmission tower point set P'. The formula for longitude normalization is: Calculate the normalized longitude value for each point, where, Indicates normalized longitude. Indicates longitude. Indicates the maximum longitude. This represents the minimum longitude. Similarly, normalized latitude and normalized altitude values ​​are calculated.

[0045] Step 1042: For each point in the transmission tower point set P', determine its next adjacent transmission tower point according to the transmission line route from the starting point to the ending point. Based on the normalized three-dimensional coordinates (longitude, latitude, and altitude) of the two adjacent transmission tower points, calculate the distance between them using the Euclidean distance formula. Combined with surrounding terrain information (such as terrain slope and aspect information) and the crossing situation of the transmission line between towers, establish a calculation model for terrain complexity parameters.

[0046] Step S1043: Calculate the observed feature vector for each transmission tower point. ,in It is the direction angle. The nearest neighbor distance. For local density, For elevation gradient, For consistency of direction.

[0047] Step S1044: Specifically, for the transmission tower point set P' Each point in ( i=1,2,…,n When i is not the last point, calculate the relationship between that point and its next neighboring point. The direction angle between them , Calculated direction angle The range is within [−π, π], when i = n When the current point is reached, the direction angle from the previous point to the current point is taken (in reverse). Nearest neighbor distance. The search is accelerated using a KD-Tree, which queries the nearest neighbor of each point and calculates the Euclidean distance. Local density. This represents the number of adjacent points within a radius r. Elevation gradient This represents the rate of change of altitude per unit distance between adjacent points. ,in This represents the absolute value of the elevation difference between adjacent points. This represents the horizontal distance (two-dimensional Euclidean distance) between adjacent points and the current point. Directional consistency. Let cosine be the angle between the current direction and the direction of the previous line segment. ,in The vector of the preceding line segment. This is the vector of the current line segment.

[0048] In one embodiment, step 106 includes: determining the state space based on the actual connection status and possible connection modes of the transmission tower points; determining the observation space based on the coordinates of the transmission tower points and the observed feature vectors; initializing all elements of the state transition probability matrix to preset small non-zero values, or assigning a preset relatively large initial probability to a common state transition path based on prior knowledge; each element in the state transition probability matrix... Indicates from state i Transition to state j The state transition probabilities satisfy:

[0049] in, α , β , c These are weights used to control the changes in orientation angle, the deviation between the nearest neighbor distance and the average distance, and the impact of terrain obstacle costs on the state transition probability; − | is a state i and state j corresponding direction angle and The absolute value of the interpolation, For transmission tower points i nearest neighbor distance, The average distance, For the calculation of the transmission tower point using digital elevation model To the transmission tower point The terrain obstacle cost is estimated by analyzing factors such as terrain undulation and obstacle distribution between two points and using a path analysis algorithm.

[0050] Based on the observed feature vectors, a Gaussian mixture model is used to fit the feature distribution to obtain the transmission probability; based on the topology of the transmission network, the distribution of the starting state is determined, and the initial probability is determined based on the distribution of the starting state.

[0051] Specifically, the process of Hidden Markov Model (HMM) modeling includes: Step 1061: The state space S represents the possible connection modes of the transmission tower point. Based on the actual connection situation and possible connection modes of the transmission tower point, all states contained in the state space S are explicitly listed. S is defined as S = {straight line extension, left obstacle avoidance, right obstacle avoidance, clockwise turning connection, counterclockwise turning connection}. Each state represents a specific connection mode of the transmission tower point in the transmission line.

[0052] Step 1062: The observation space V represents the coordinates and eigenvectors of the transmission tower. The contents of the observation space V are determined, including the coordinates of the transmission towers and the previously calculated observation feature vectors. This information is integrated into an observation vector, which is used to represent the observable features of each transmission tower point in the HMM.

[0053] Step 1063: Initialize the state transition probability matrix A = [ ], its dimension is N×N ( N (where S is the number of states in the state space S), and each element in the matrix... Indicates from state i Transition to state j The probability of state transition can be initially determined by setting all elements to a small non-zero value, or by assigning relatively large initial probabilities to certain common state transition paths based on prior knowledge. satisfy:

[0054] Among them, parameters a、b、c Weights (parameters) are used to control the changes in orientation angle, the deviation between the nearest neighbor distance and the average distance, and the impact of terrain obstacle costs on the state transition probability. (trained using historical route data); for each state transition i→j Calculate the corresponding and Calculate their difference | − | and nearest neighbor distance and average distance The difference | |;Calculate using digital elevation model That is, from the transmission tower point arrive The terrain obstacle cost is estimated by analyzing factors such as terrain undulation and obstacle distribution between two points and using a path analysis algorithm.

[0055] Step 1064, Launch Probability The feature distribution is fitted using a Gaussian mixture model; for each state... j Collect the observed feature vectors of all transmission tower points corresponding to this state. This forms the observation dataset for that state, for each state. j The observed dataset is used to train the Gaussian Mixture Model (GMM) algorithm, determining its parameters, including mixing coefficients, mean vector, and covariance matrix. After training, for a given observed vector v, its state is calculated based on the probability density function of the GMM. j under the emission probability That is, the observation vector belongs to the state. j The probability is given by the formula:

[0056] Step 1065: Based on the topology of the transmission network, analyze and determine the distribution of the starting state and the initial probability. The starting point state distribution is determined based on the power grid topology.

[0057] In one embodiment, the Hidden Markov Model parameters include: state space, observation space, state transition probability matrix, emission probability, and initial probability; step 110 includes: based on the initial probability and the first observation vector of the ordered observation sequence... At t=1, the observation vector of state j in the state space is emitted. The product of the probability of state j and the initial probability of state j is used as the initial state probability of state j; and the path tracing pointer of state j is recorded. (Indicates the initial state); when For each state in the state space, under the state transition distance constraint, the preceding state i that satisfies the distance constraint is selected; the transition distance from state i is calculated. i to state j State transition probability And combine the state probability of the previous time step. And the observation vector emitted from the current state j launch probability The candidate probability values ​​are obtained as follows:

[0058] in, N Let S be the size of the state space, and record the state such that... The largest previous state i The index.

[0059] exist When finding the state index with the highest probability. Determine the joint probability of the optimal state sequence corresponding to the entire observation sequence O; from Begin by tracing the path pointers sequentially. ψ t ( j Backtracking yields the optimal state sequence. ,in , Until back to The initial state at time; based on the optimal state sequence Q* and the ordered observation sequence O, each state is... qᵢ * Corresponding transmission tower point pᵢ Connect them sequentially to form a preliminary connection path. .

[0060] Specifically, step 110 includes: Step 1100: Based on the spatial distribution of the transmission tower point set and the power grid topology, determine the starting point of the transmission line. Starting from the starting point, determine the next transmission tower point in sequence according to spatial proximity, generating an ordered observation sequence. , Step 1101: Based on the initial probability vector π and the first observation vector of the observation sequence... At t=1, calculate each state initial state probability ,in The observation vector is emitted from state j. The probability; record the corresponding path tracing pointer. (Indicates the initial state). For t =2 to n For each state Under the condition of state transition distance constraint, that is, the Euclidean distance between adjacent transmission tower points is less than 2× Within the range, select the previous state that satisfies the distance constraint. Calculate from state i to state j transition probability And combine the state probability of the previous time step. And the observation vector emitted from the current state j probability Obtain candidate probability values ,inN Let S be the size of the state space. Records such that... The largest previous state i The index. In t = n When finding the state index with the highest probability. ,at this time This is the joint probability of the optimal state sequence corresponding to the entire observation sequence O. From Begin by tracing the path pointers sequentially. By backtracking, the optimal state sequence is obtained. ,in , Until back to t The initial state when = 1.

[0061] Step 1102: Based on the optimal state sequence Q* And an ordered observation sequence O, for each state qᵢ* Corresponding transmission tower points pᵢ Connect them sequentially to form a preliminary connection path. .

[0062] In one embodiment, step 112 includes: determining the weighting coefficients of the total route length, gradient penalty cost, and turning angle penalty cost in the cost function of the route planning based on the focus and priority of the engineering design; obtaining the total route length based on the Euclidean distance between adjacent transmission towers according to the preliminary connection path; calculating the gradient penalty cost of each segment according to the gradient of the actual terrain and certain penalty rules based on the gradient of each segment in the preliminary connection path, and summing the gradient penalty costs of all segments to obtain the gradient penalty cost; checking the angle of each turn in the preliminary connection path, and calculating the turning angle penalty cost based on the degree of deviation from the allowable range for turning angles that do not meet the requirements of the transmission line design specifications; weighting and summing the total route length, gradient penalty cost, turning angle penalty cost, and the weighting coefficients of the total route length, gradient penalty cost, and turning angle penalty cost to obtain the cost function of the route planning; and using the cost function of the route planning and... The algorithm makes local adjustments to the initial connection path to ensure that the span and turning angle meet the preset standard requirements, thus obtaining the adjusted path.

[0063] Specifically, the post-processing of engineering constraints includes: determining the weight coefficients of each part of the cost function based on the focus and priority of the engineering design. l、m、n Based on the preliminary connection path The total route length is obtained by summing the Euclidean distances between adjacent transmission towers. The slope of each segment in the preliminary connection path is analyzed, and the slope penalty cost for each segment is calculated according to a certain penalty rule based on the actual terrain slope value. Then, the slope penalty costs of all segments are summed and multiplied by a weighting coefficient. β The total gradient penalty cost is obtained; the angles at each bend in the initial connection path are checked, and for bend angles that do not meet the requirements of the transmission line design specifications, the bend angle penalty cost is calculated based on the degree of deviation from the allowable range. The cost function is then summarized. Next, the initial connection path is replaced with the locally optimal path obtained through the A* algorithm. The corresponding local segment is used to obtain the updated preliminary connection path. .

[0064] In one embodiment, step 114 includes: simplifying the adjusted path using the Douglas-Peucker algorithm and smoothing it using B-spline curves; overlaying and displaying the optimized path results in a GIS platform, supporting manual dragging and adjustment of specific segments; adding the manually corrected and confirmed path data to the HMM training set for incrementally updating model parameters, especially the state transition probability matrix A and the emission probability matrix B, to enhance the model's adaptive capability.

[0065] Specifically, the path optimization and interaction process includes: simplification of the path based on the Douglas-Peucker algorithm, combined with B-spline curve smoothing; loading the smoothed path data and related terrain, feature, and other geographic information data into the GIS platform, ensuring that the coordinate systems of all data are consistent and correctly registered for display; using existing path editing tools, allowing users to adjust specific points or segments on the path through mouse dragging and other operations; recording all modifications made by the user to the path during manual interactive adjustments, including the coordinates of path points before and after modification, timestamps of adjustments, etc.; organizing the manually corrected path data to form a new dataset that can be used for HMM model training; and updating the HMM model parameters, including incrementally updating the state transition probability matrix A and the emission probability matrix B.

[0066] In one embodiment, an obstacle avoidance penalty term is introduced for A. * Algorithm optimization; obstacle avoidance penalty is used to avoid rivers, mountains and no-construction areas.

[0067] In one embodiment, the state space includes, but is not limited to, straight-line extension, left-side obstacle avoidance, right-side obstacle avoidance, clockwise turning connection, and counterclockwise turning connection.

[0068] This method integrates Hidden Markov Model (HMM) probabilistic modeling with spatial features (direction, density, terrain) to improve connection accuracy; density clustering adaptively removes outliers to reduce noise interference; a cost function and A* algorithm enforce compliance with national design standards (span, slope, turning angle); obstacle avoidance penalties avoid sensitive areas (rivers, mountains); manual interactive correction and incremental learning mechanisms continuously optimize model parameters; path simplification and smoothing improve constructability; spatial indexing accelerates neighbor point search; constrained Viterbi algorithm reduces invalid state transitions and improves connection efficiency; the transmission tower connection pattern is abstracted into an HMM state transition problem, combining geographical features and engineering constraints to achieve high-precision automatic connection, and interactive incremental learning improves system adaptability, solving the problem of how to automatically connect transmission tower points to lines in remote sensing imagery.

[0069] It should be understood that, although the above process Figure 1 The steps in the diagram are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order in which these steps are executed; they can be performed in other orders. Furthermore, the above process... Figure 1 At least some of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0070] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and all such modifications and improvements fall within the scope of protection of this application.

Claims

1. A method for automatically connecting transmission tower points in remote sensing images based on a hidden Markov model, characterized in that, The method includes the following steps: Obtain a set of transmission tower locations, which includes information on multiple transmission tower locations, including the longitude, latitude, and altitude of the location of each transmission tower. A spatial index is constructed on the original data of the transmission tower point set, and outliers are filtered out using a density clustering algorithm to obtain the preprocessed transmission tower point set. Based on the preprocessed set of transmission tower points, the observation feature vector of each transmission tower point is extracted; Based on the observed feature vectors, power grid topology, and geographic information and spatial features in remote sensing image data, construct the parameters of the hidden Markov model. Based on the spatial distribution of the transmission tower point set and the power grid topology, the starting point of the transmission line is determined. Starting from the starting point, the next transmission tower point is determined sequentially according to spatial proximity, generating an ordered observation sequence. Based on the ordered observation sequence and the parameters of the hidden Markov model, the optimal state sequence is solved to generate a preliminary connection path. Based on the initial connection path and engineering specifications, compliance adjustments were made to obtain the adjusted path; The adjusted path is simplified and smoothed, and the smoothed path result and geographic information data are loaded into the GIS platform. The path is then registered and displayed by manual dragging and dropping, while the parameters of the Hidden Markov Model are updated to obtain the final optimized path.

2. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, A spatial index is constructed from the original data of the transmission tower point set, and outliers are filtered out using a density clustering algorithm to obtain the preprocessed transmission tower point set, including: We use an R-tree based on a hierarchical index as a spatial index to organize and query spatial data; Based on the density, geographical range, and business needs of the actual transmission tower distribution, determine the neighborhood radius r and the number of points threshold k; Iterate through each transmission tower point p in the transmission tower point set P. Using transmission tower point p as the center and neighborhood radius r as the radius, query all transmission tower points in the neighborhood in the R-tree spatial index and count the number of points in the neighborhood. For each transmission tower point p, if the number of points in its neighborhood is less than the point count threshold k, then the point is marked as an outlier. After all points have been traversed, all marked outliers are removed from the transmission tower point set P to obtain the preprocessed transmission tower point set.

3. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, the observation... The feature vectors include: orientation angle, nearest neighbor distance, local density, gradient, and orientation consistency; Based on the preprocessed set of transmission tower points, the observed feature vector of each transmission tower point is extracted, including: Calculate the orientation angle between each transmission tower point in the preprocessed transmission tower point set, excluding the last transmission tower point, and its next adjacent point; the range of the orientation angle is... ; For the last transmission tower point in the preprocessed transmission tower point set, take the direction angle from the previous transmission tower point to the current point; The search is accelerated by using KD-Tree. The nearest neighbor point of each transmission tower point in the preprocessed transmission tower point set is queried and the Euclidean distance is calculated to obtain the nearest neighbor distance. The number of neighboring points within a neighborhood radius r is used as the local density; The rate of change of elevation per unit distance between adjacent points is used as the elevation gradient; The cosine of the angle between the current direction and the previous direction is taken as the direction consistency.

4. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, Based on the observed feature vectors, power grid topology, and geographic information and spatial features in the remote sensing image data, the parameters of the hidden Markov model are constructed, including: The state space is determined based on the actual connection status and possible connection modes of the transmission tower points; The observation space is determined based on the coordinates of the transmission tower points and the observed feature vectors. All elements of the state transition probability matrix are initialized to preset small non-zero values, or a preset relatively large initial probability is assigned to a common state transition path based on prior knowledge; each element aij in the state transition probability matrix represents the state transition probability from state i to state j; the state transition probabilities satisfy: Where α, β, and γ are weights used to control the changes in orientation angle, the deviation between the nearest neighbor distance and the average distance, and the impact of terrain obstacle costs on the state transition probability; - | is the direction angle corresponding to states i and j. and The absolute value of the interpolation, The nearest distance to transmission tower point i is... The average distance, For the calculation of the transmission tower point using digital elevation model To the transmission tower point The terrain obstacle cost is estimated by analyzing factors such as terrain undulation and obstacle distribution between two points and using a path analysis algorithm. Based on the observed feature vector, a Gaussian mixture model is used to fit the feature distribution to obtain the emission probability; Based on the topology of the power transmission network, the distribution of the starting state is determined, and the initial probability is determined based on the distribution of the starting state.

5. The method for automatically connecting transmission tower points in remote sensing images based on a hidden Markov model according to claim 1, characterized in that, Hidden Markov Model parameters include: state space, observation space, state transition probability matrix, emission probability, and initial probability; Based on the ordered observation sequence and the hidden Markov model parameters, the optimal state sequence is solved to generate a preliminary connection path, including: Based on the initial probability and the first observation vector of the ordered observation sequence ,exist t When =1, the states in the state space j Emit observation vector probability and state j The product of the initial probabilities is used as the state. j Determine the initial state probability and record the state. j Path tracing pointer (Indicates the initial state); when For each state in the state space, under the state transition distance constraint, select the previous state i that satisfies the distance constraint; calculate the transition distance from state i. i to state j State transition probability And combine the state probability of the previous time step. and current state j Launch observation vector launch probability The candidate probability values ​​are obtained as follows: in, N Let S be the size of the state space, and record such that The largest previous state i Index; exist When finding the state index with the highest probability. Determine the joint probability of the optimal state sequence corresponding to the entire observation sequence O; from Begin by sequentially tracing the path source pointer ψ t (j) Backtracking to obtain the optimal state sequence ,in , Until back to The initial state at that time; Based on the optimal state sequence Q* and the ordered observation sequence O, the transmission tower points pᵢ corresponding to each state qᵢ* are connected sequentially to form a preliminary connection path. .

6. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, Based on the initial connection path and engineering specifications, compliance adjustments were made to obtain the adjusted path, including: Based on the key points and priorities of the engineering design, determine the weight coefficients of total route length, gradient penalty cost, and turning angle penalty cost in the cost function of route planning; Based on the preliminary connection path, the Euclidean distance between each adjacent transmission tower point is used to obtain the total line length; Based on the slope of each segment in the initial connection path, and according to the actual terrain slope value, the slope penalty cost of each segment is calculated according to certain penalty rules. The slope penalty costs of all segments are added together to obtain the slope penalty cost. Check the angle of each bend in the preliminary connection path. For bend angles that do not meet the requirements of the transmission line design specifications, calculate the penalty cost for the bend angle based on the degree of deviation from the allowable range. The cost function for route planning is obtained by weighting and summing the total route length, the gradient penalty cost, the turning angle penalty cost, and the weight coefficients of the total route length, the gradient penalty cost, and the turning angle penalty cost. The cost function of the route planning and The algorithm makes local adjustments to the initial connection path to ensure that the span and turning angle meet the preset standard requirements, thus obtaining the adjusted path.

7. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, The adjusted path is simplified and smoothed, and the smoothed path result and geographic information data are loaded into the GIS platform. Registration and display are performed manually by dragging and dropping, while simultaneously updating the Hidden Markov Model parameters to obtain the final optimized path, including: The adjusted path is simplified using the Douglas-Peucker algorithm and smoothed using B-spline curves. The optimized path results are overlaid and displayed in the GIS platform, and manual drag-and-drop adjustment of specific sections is supported; The manually corrected and confirmed path data is added to the HMM training set for incremental updates of model parameters, especially the state transition probability matrix A and the emission probability matrix B, to enhance the model's adaptability.

8. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, Introducing an obstacle avoidance penalty term for the A * Algorithm optimization; the obstacle avoidance penalty is used to avoid rivers, mountains and prohibited construction areas.

9. The method for automatically connecting transmission tower points in remote sensing images based on hidden Markov models according to claim 1, characterized in that, The state space includes, but is not limited to, straight-line extension, left-side obstacle avoidance, right-side obstacle avoidance, clockwise turning connection, and counterclockwise turning connection.

Images