Wind power equipment transportation scene construction method and system

Abstract

The invention provides a wind power equipment transportation scene construction method and system, and relates to the technical field of environment modeling and emulation.The method comprises the steps that environment data in a pre-transportation environment are obtained according to a multi-source sensor, denoising and credibility grading are conducted on the environment data, and graded environment data are obtained; according to the graded environment data, scene distortion factors in the environment data are identified and corrected, and a preliminary static scene is generated according to the corrected environment data; the method comprises the following steps: capturing dynamic change factors in a pre-transportation environment in real time by taking a static scene with geometric and texture information as a reference, identifying a dynamic area influencing a transportation path according to the dynamic change factors, and updating and generating a real-time dynamic scene through the dynamic area. According to the invention, through real-time monitoring and dynamic updating of the environment change, it is ensured that the transportation path and scene always accord with the latest environment condition, the transportation risk caused by the environment change is reduced, and the safety in the transportation process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental modeling and simulation, and in particular to a method and system for constructing a wind power equipment transportation scenario. Background Art

[0002] With the rapid development of the wind power industry, the transportation of wind power equipment has become increasingly complex, especially in harsh environments.

[0003] Conventional wind power equipment transportation relies on static models and simple scenario simulations, which cannot accurately reflect dynamically changing environmental factors. The impact of wind power equipment transportation in a multi-factor environment has not been fully considered, and it is impossible to respond to environmental changes in real time, and it is difficult to effectively integrate in complex multimodal scenarios. Conventional methods mostly rely on historical data or preset models, lack dynamic response to real-time changes, and cannot effectively integrate multi-source data and comprehensively evaluate their impact on transportation. At the same time, it is impossible to automatically update or adjust the scenario, resulting in a large gap between the simulation results and the actual environment, affecting the accuracy of decision-making. Summary of the invention

[0004] The present invention provides a method and system for constructing a wind power equipment transportation scenario, aiming to solve the technical problem of the lack of dynamic response to real-time changes in related technologies, avoid the practical problem of being unable to effectively integrate and synthesize multi-source data, and circumvent the subsequent problem of being unable to automatically update or adjust the scenario.

[0005] In order to achieve the above-mentioned object, the present invention provides a method for constructing a wind power equipment transportation scenario, comprising the following steps: obtaining environmental data in a pre-transportation environment according to a multi-source sensor, and performing denoising and credibility classification on the environmental data to obtain the classified environmental data, specifically comprising: Acquire environmental data in the pre-transportation environment through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data; The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the error areas in the point cloud data are marked and initially corrected by combining the texture features of the real-time image, including: Based on the density clustering algorithm, points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points; by calculating the mean distance from the point to the neighborhood, points whose mean distance is greater than the preset distance threshold are marked as noise points; Calculate the weighted scores of candidate noise points at different scales based on multi-scale spatial parameters, and remove points with scores lower than the preset score threshold; Align the coordinates of the point cloud data with the image data, project the point cloud data onto the image plane, extract the texture features in the image data, and identify the areas with abnormal texture features as error areas; For the marked error areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud; Introduce a credibility assessment model for environmental data and classify the data, including: Assign a credibility score to each set of environmental data, mark the data with a credibility score higher than a preset credibility threshold as high credibility data, and mark the data with a credibility score lower than the preset credibility threshold as low credibility data; Data is stored in layers according to the credibility score. High-credibility data is used first for preliminary static scene construction. Low-credibility data is marked and stored for backup. According to the classified environmental data, a multimodal data adaptive fusion algorithm is used to dynamically match the texture features of point cloud data and real-time image data, and the fusion weight is adjusted according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows:

[0006] In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model:

[0007] In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights, respectively. The goal of the above formula is to fuse multimodal features to generate a static scene with geometric and texture information; Based on the static scene with geometric and texture information, the dynamic change factors in the pre-transportation environment are captured in real time, the dynamic areas that affect the transportation path are identified according to the dynamic change factors, and the real-time dynamic scene is generated by updating the dynamic areas; Perform path optimization calculations based on real-time dynamic scenarios, generate a global optimal path that can be adjusted in real time based on the calculation results, and output the path planning solution through visualization technology.

[0008] As a preferred solution of the present invention, a static scene with geometric and texture information is used as a benchmark to capture dynamic change factors in the pre-transportation environment in real time, identify dynamic areas that affect the transportation path based on the dynamic change factors, and generate a real-time dynamic scene through dynamic area update, specifically including: Real-time monitoring of traffic flow, weather changes and obstacle locations in the pre-transportation environment to identify dynamic factors affecting the transportation route; According to the dynamic change factors in the pre-transportation environment, the dynamic areas that affect the path are identified and divided, and the dynamic areas are continuously tracked to update their positions and states. According to the captured dynamic changes, the dynamic areas in the preliminary static scene are updated, and a real-time dynamic scene is generated based on the updates.

[0009] As a preferred solution of the present invention, according to the dynamic change factors in the pre-transportation environment, the dynamic area affecting the path is identified and divided, the dynamic area is continuously tracked, and its position and state are updated; according to the captured dynamic changes, the dynamic area in the preliminary static scene is updated, and a real-time dynamic scene is generated according to the update, which specifically includes: By collecting environmental data in real time, the dynamic areas that affect the transportation path are identified, and the scope of the dynamic areas is divided and marked according to the nature of the dynamic change factors; For the identified dynamic areas, the system continuously tracks the changes in the dynamic areas through real-time data updates and tracking, and adjusts the boundaries of the dynamic areas according to the updated environmental data; According to the real-time changing status data, the status of the dynamic area is updated, the updated dynamic area information is fed back to the preliminary static scene, the position and shape of the dynamic area in the preliminary static scene are adjusted, and the corresponding area data in the preliminary static scene is updated. By integrating the updated dynamic area with the original static scene data, a real-time dynamic scene is generated, and the status data includes traffic status changes or obstacle position change data.

[0010] The present invention further provides a wind power equipment transportation scenario construction system, comprising the following units: The classification unit is used to obtain environmental data in the pre-transportation environment based on multi-source sensors, and to perform denoising and credibility classification on the environmental data to obtain classified environmental data, specifically including: Acquire environmental data in the pre-transportation environment through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data; The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the error areas in the point cloud data are marked and initially corrected by combining the texture features of the real-time image, including: Based on the density clustering algorithm, points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points; by calculating the mean distance from the point to the neighborhood, points whose mean distance is greater than the preset distance threshold are marked as noise points; Calculate the weighted scores of candidate noise points at different scales based on multi-scale spatial parameters, and remove points with scores lower than the preset score threshold; Align the coordinates of the point cloud data with the image data, project the point cloud data onto the image plane, extract the texture features in the image data, and identify the areas with abnormal texture features as error areas; For the marked error areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud; Introduce a credibility assessment model for environmental data and classify the data, including: Assign a credibility score to each set of environmental data, mark the data with a credibility score higher than a preset credibility threshold as high credibility data, and mark the data with a credibility score lower than the preset credibility threshold as low credibility data; Data is stored in layers according to the credibility score. High-credibility data is used first for preliminary static scene construction. Low-credibility data is marked and stored for backup. A fusion modeling unit is used to dynamically match the texture features of the point cloud data and the real-time image data according to the classified environmental data using a multimodal data adaptive fusion algorithm, and to adjust the fusion weight according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows:

[0011] In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model:

[0012] In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights, respectively. The goal of the above formula is to fuse multimodal features to generate a static scene with geometric and texture information; The dynamic update unit is used to capture the dynamic change factors in the pre-transportation environment in real time based on the static scene with geometric and texture information, identify the dynamic area that affects the transportation path according to the dynamic change factors, and generate a real-time dynamic scene through dynamic area update.

[0013] The beneficial effects of the present invention are: 1. Based on the data of real-time dynamic environmental changes, the wind power equipment transportation scenario is continuously updated. Through dynamic area update technology, it is ensured that the scenario model can reflect the real-time changes in actual transportation and improve the real-time and accuracy of the scenario model.

[0014] 2. Through real-time monitoring and dynamic updating of environmental changes, ensure that the transportation routes and scenarios always meet the latest environmental conditions, reduce transportation risks caused by environmental changes, and improve safety during transportation.

[0015] 3. In a complex and dynamically changing environment, conventional scene construction methods cannot meet the real-time and efficient requirements. The present invention provides a more precise and operational scene model for the transportation of wind power equipment through accurate environmental simulation and dynamic updating.

[0016] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0018] Figure 1A flow chart of a method for constructing a wind power equipment transportation scenario provided in an embodiment of the present invention.

[0019] Figure 2 An overall framework diagram of a wind power equipment transportation scenario construction system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0021] See also Figure 1 , Figure 1 A flowchart of a method for constructing a wind power equipment transportation scenario is provided for an embodiment of the present invention.

[0022] In this embodiment, a method for constructing a wind power equipment transportation scenario includes step S100, step S200, step S300 and step S400.

[0023] Step S100, obtaining environmental data in the pre-transportation environment according to multi-source sensors, and performing denoising and credibility classification on the environmental data to obtain classified environmental data, specifically includes: Environmental data in the pre-transportation environment is obtained through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data.

[0024] It should be noted that multi-source sensors include lidar, drones and meteorological sensors. Lidar, drones and meteorological sensors use device time synchronization to ensure that the collection timestamps of all data sources are consistent, providing a basis for multimodal fusion.

[0025] The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the erroneous areas in the point cloud data are marked and initially corrected in combination with the texture features of the image data.

[0026] It should be noted that the points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points based on the density clustering algorithm; by calculating the mean distance from the point to the neighborhood, the points whose mean distance is greater than the preset distance threshold are marked as noise points.

[0027] It should be further explained that the preset density threshold is used to determine whether the number of points in the neighborhood of each point is lower than the set standard, so as to determine whether the point is a noise point. The preset density threshold is determined by statistically analyzing the normal distribution characteristics of point cloud data under different environmental conditions. The preset density threshold is specifically the density value of the 10th percentile of the density distribution of the point cloud data.

[0028] The preset distance threshold is used to determine whether the average distance between a point and its neighborhood exceeds the normal range. The preset distance threshold is set to the distance mean plus 2 times the standard deviation.

[0029] The weighted scores of candidate noise points at different scales are calculated according to the multi-scale spatial parameters, and the points with scores lower than the preset score threshold are eliminated.

[0030] The preset scoring threshold is specifically the 10th percentile in the scoring distribution.

[0031] Multi-scale spatial parameters refer to a set of parameters used to evaluate and analyze data characteristics at different scales when processing point cloud data. Multi-scale spatial parameters include point density, neighborhood radius, geometric characteristics, point-to-neighborhood distance distribution, and point cloud spatial distribution.

[0032] The coordinates of the point cloud data and the image data are aligned, the point cloud data is projected onto the image plane, the texture features in the image data are extracted, and the areas with abnormal texture features are identified as error areas.

[0033] For the marked erroneous areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud.

[0034] It should be noted that the interpolation algorithm uses nearest neighbor interpolation to recalculate the feature values ​​of the point cloud in the area and calibrate the corrected point cloud.

[0035] Introduce a credibility assessment model for environmental data and classify the data, including: A credibility score is assigned to each set of environmental data, and data with a credibility score higher than a preset credibility threshold is marked as high credibility data, and data with a credibility score lower than the preset credibility threshold is marked as low credibility data.

[0036] Data is stored in layers according to the credibility score, with high-credibility data used first for subsequent scenario construction; low-credibility data is marked and stored for backup.

[0037] It should be noted that the credibility threshold is used to distinguish high-credibility data from low-credibility data; if the credibility score of the data is higher than the threshold, it is regarded as high-credibility data; if the credibility score of the data is lower than the threshold, it is regarded as low-credibility data.

[0038] It should be further explained that the value range of the credibility threshold is specifically 0.8 to 1.

[0039] Step S200: Based on the classified environmental data, a multimodal data adaptive fusion algorithm is used to dynamically match the texture features of the point cloud data with the real-time image data, and the fusion weight is adjusted according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows:

[0040] In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model:

[0041] In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights respectively. The goal of the above formula is to fuse multimodal features and generate a static scene with geometric and texture information.

[0042] It should be further explained that the generation of the preliminary static scene is to synthesize the corrected point cloud data with the image texture information to ensure the seamless connection of all point cloud data and texture information, and finally generate a preliminary static scene model.

[0043] Step S300, based on the static scene with geometric and texture information, real-time capture of dynamic change factors in the pre-transportation environment, identification of dynamic areas that affect the transportation path based on the dynamic change factors, and generation of real-time dynamic scenes through dynamic area update, specifically includes: Real-time monitoring of traffic flow, weather changes and obstacle locations in the pre-transportation environment to identify dynamic factors that affect the transportation route.

[0044] It should be noted that by real-time monitoring of dynamic factors in the pre-transportation environment, including traffic flow, weather changes and obstacle locations, the dynamic areas that affect the transportation path are identified; the dynamic change factors are collected by sensors and combined with environmental parameter analysis to determine the key factors that affect the path, including the following steps: E1. Collect environmental data in real time through multi-source sensors.

[0045] E2. Fuse environmental data from different sources, use Kalman filtering to process traffic flow, weather changes, and obstacle locations, and identify dynamic areas that affect transportation routes.

[0046] According to the dynamic change factors in the pre-transportation environment, the dynamic areas that affect the path are identified and divided, and the dynamic areas are continuously tracked to update their positions and states; according to the captured dynamic changes, the dynamic areas in the preliminary static scene are updated, and the real-time dynamic scene is generated based on the updates, including: It should be noted that the dynamic areas that affect the transportation path are identified and divided into regions according to the nature of the dynamic change factors. The regional division is layered based on the speed and size of environmental changes and the degree of their impact on the transportation path, including the following steps: F1. Based on real-time monitoring data, DBSCAN clustering is used to identify the changing areas. The scope of the dynamic area is defined based on traffic flow density, meteorological change range or obstacle movement trajectory.

[0047] F2. Mark and classify the identified dynamic areas to determine the level of regional dynamics.

[0048] By collecting environmental data in real time, the dynamic areas that affect the transportation path are identified, and the scope of the dynamic areas is divided and marked according to the nature of the dynamic change factors.

[0049] For the identified dynamic areas, the changes in the dynamic areas are continuously tracked through real-time data updates and tracking, and the boundaries of the dynamic areas are adjusted according to the updated environmental data, including the following steps: G1. Use Kalman filtering to continuously track changes in dynamic areas. For each dynamic area, estimate its position change, expansion or contraction.

[0050] G2. Dynamically adjust the shape and position of the area based on the data collected in real time to ensure that the dynamic area in the scene always reflects the real environment changes.

[0051] According to the real-time changing status data, the status of the dynamic area is updated, and the updated dynamic area information is fed back to the preliminary static scene. The position and shape of the dynamic area in the preliminary static scene are adjusted, and the corresponding area data in the preliminary static scene is updated. By integrating the updated dynamic area with the original static scene data, a real-time dynamic scene is generated. The status data includes traffic status changes or obstacle position change data.

[0052] It should be noted that, according to the real-time updated dynamic area information, the position and shape of the dynamic area in the preliminary static scene are adjusted to finally generate a real-time dynamic scene, including the following steps: H1. Feed the updated dynamic area information back to the preliminary static scene and adjust the boundary and position of the original dynamic area.

[0053] H2. Generate a new real-time dynamic scene by combining the updated dynamic area data and the preliminary static scene through the weighted average method.

[0054] Step S400, performing path optimization calculation according to the real-time dynamic scene, generating a global optimal path that can be adjusted in real time according to the calculation results, and outputting the path planning solution through visualization technology, specifically including: Based on real-time dynamic scene data, the changing factors in the transportation path are extracted and input into the path optimization calculation model, including the following steps: J1. Collect traffic flow, weather changes and obstacle location data from real-time dynamic scenes.

[0055] J2. Use Kalman filtering to extract the changing factors that affect the path.

[0056] J3. The extracted change factors are used as input parameters and sent to the path optimization model for calculation.

[0057] Use optimization algorithms to calculate the optimal path based on real-time scene data, dynamically adjust the calculated optimal path based on real-time changing environmental data, and obtain the global optimal path based on dynamic adjustment.

[0058] It should be noted that the Dijkstra path optimization algorithm is used to calculate the optimal path from the starting point to the end point based on real-time scene data. By considering obstacles, traffic flow and weather, each possible path is evaluated and the most suitable path is selected. Real-time changing environmental data will dynamically affect path selection, so it is necessary to introduce a dynamic adjustment mechanism in the optimization algorithm, including the following steps: K1. Calculate the optimal path from the starting point to the end point according to the Dijkstra algorithm, which takes into account all known dynamic change factors.

[0059] K2. During the path calculation process, new environmental data is accessed in real time to adjust the calculated path.

[0060] K3. When new obstacles or traffic changes are encountered during the path calculation process, the path will be updated in real time.

[0061] Based on visualization technology, the global optimal path is superimposed on the pre-transportation scenario for visual mapping.

[0062] It should be noted that the calculated global optimal path is superimposed on the pre-transportation scenario, and the path is displayed through the GIS system. When the path changes, the visual display is immediately updated to ensure that the path planning solution always reflects the latest environmental status.

[0063] At this point, a method for constructing a wind power equipment transportation scenario is completed.

[0064] See also Figure 2 , Figure 2 An overall framework diagram of a wind power equipment transportation scenario construction system is provided for the present invention.

[0065] A wind power equipment transportation scenario construction system includes the following units: The classification unit is used to obtain environmental data in the pre-transportation environment based on multi-source sensors, and to perform denoising and credibility classification on the environmental data to obtain classified environmental data, specifically including: Acquire environmental data in the pre-transportation environment through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data; The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the error areas in the point cloud data are marked and initially corrected by combining the texture features of the real-time image, including: Based on the density clustering algorithm, points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points; by calculating the mean distance from the point to the neighborhood, points whose mean distance is greater than the preset distance threshold are marked as noise points; Calculate the weighted scores of candidate noise points at different scales based on multi-scale spatial parameters, and remove points with scores lower than the preset score threshold; Align the coordinates of the point cloud data with the image data, project the point cloud data onto the image plane, extract the texture features in the image data, and identify the areas with abnormal texture features as error areas; For the marked error areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud; Introduce a credibility assessment model for environmental data and classify the data, including: Assign a credibility score to each set of environmental data, mark the data with a credibility score higher than a preset credibility threshold as high credibility data, and mark the data with a credibility score lower than the preset credibility threshold as low credibility data; Data is stored in layers according to the credibility score. High-credibility data is used first for preliminary static scene construction. Low-credibility data is marked and stored for backup. A fusion modeling unit is used to dynamically match the texture features of the point cloud data and the real-time image data according to the classified environmental data using a multimodal data adaptive fusion algorithm, and to adjust the fusion weight according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows:

[0066] In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model:

[0067] In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights, respectively. The goal of the above formula is to fuse multimodal features to generate a static scene with geometric and texture information; The dynamic update unit is used to capture the dynamic change factors in the pre-transportation environment in real time based on the static scene with geometric and texture information, identify the dynamic area that affects the transportation path according to the dynamic change factors, and generate a real-time dynamic scene through dynamic area update.

[0068] At this point, a wind power equipment transportation scenario construction system has been completed.

[0069] The above is only a specific implementation of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the embodiments of the present invention, which should be included in the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for constructing a wind power equipment transportation scenario, characterized in that: The following steps are involved: The environmental data in the pre-transportation environment is obtained by multi-source sensors, and the environmental data is denoised and credibility graded to obtain graded environmental data, including: Acquire environmental data in the pre-transportation environment through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data; The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the error areas in the point cloud data are marked and initially corrected by combining the texture features of the real-time image, including: Based on the density clustering algorithm, points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points; by calculating the mean distance from the point to the neighborhood, points whose mean distance is greater than the preset distance threshold are marked as noise points; Calculate the weighted scores of candidate noise points at different scales based on multi-scale spatial parameters, and remove points with scores lower than the preset score threshold; Align the coordinates of the point cloud data with the image data, project the point cloud data onto the image plane, extract the texture features in the image data, and identify the areas with abnormal texture features as error areas; For the marked error areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud; Introduce a credibility assessment model for environmental data and classify the data, including: Assign a credibility score to each set of environmental data, mark the data with a credibility score higher than a preset credibility threshold as high credibility data, and mark the data with a credibility score lower than the preset credibility threshold as low credibility data; Data is stored in layers according to the credibility score. High-credibility data is used first for preliminary static scene construction. Low-credibility data is marked and stored for backup. According to the classified environmental data, a multimodal data adaptive fusion algorithm is used to dynamically match the texture features of point cloud data and real-time image data, and the fusion weight is adjusted according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows: In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model: In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights, respectively. The goal of the above formula is to fuse multimodal features to generate a static scene with geometric and texture information; Based on the static scene with geometric and texture information, the dynamic change factors in the pre-transportation environment are captured in real time, the dynamic areas that affect the transportation path are identified according to the dynamic change factors, and the real-time dynamic scene is generated by updating the dynamic areas; Perform path optimization calculations based on real-time dynamic scenarios, generate a global optimal path that can be adjusted in real time based on the calculation results, and output the path planning solution through visualization technology.

2. A method for constructing a wind power equipment transportation scenario according to claim 1, characterized in that: Based on the static scene with geometric and texture information, the dynamic change factors in the pre-transportation environment are captured in real time. The dynamic areas that affect the transportation path are identified according to the dynamic change factors, and the real-time dynamic scene is generated by updating the dynamic areas. Specifically, it includes: Real-time monitoring of traffic flow, weather changes and obstacle locations in the pre-transportation environment to identify dynamic factors affecting the transportation route; According to the dynamic change factors in the pre-transportation environment, the dynamic areas that affect the path are identified and divided, and the dynamic areas are continuously tracked to update their positions and states. According to the captured dynamic changes, the dynamic areas in the preliminary static scene are updated, and a real-time dynamic scene is generated based on the updates.

3. A method for constructing a wind power equipment transportation scenario according to claim 1, characterized in that: According to the dynamic change factors in the pre-transportation environment, the dynamic areas that affect the path are identified and divided, and the dynamic areas are continuously tracked to update their positions and states; according to the captured dynamic changes, the dynamic areas in the preliminary static scene are updated, and the real-time dynamic scene is generated based on the updates, including: By collecting environmental data in real time, the dynamic areas that affect the transportation path are identified, and the scope of the dynamic areas is divided and marked according to the nature of the dynamic change factors; For the identified dynamic areas, the system continuously tracks the changes in the dynamic areas through real-time data updates and tracking, and adjusts the boundaries of the dynamic areas according to the updated environmental data; According to the real-time changing status data, the status of the dynamic area is updated, the updated dynamic area information is fed back to the preliminary static scene, the position and shape of the dynamic area in the preliminary static scene are adjusted, and the corresponding area data in the preliminary static scene is updated. By integrating the updated dynamic area with the original static scene data, a real-time dynamic scene is generated, and the status data includes traffic status changes or obstacle position change data.

4. A wind power equipment transportation scenario construction system, the system is applied to a wind power equipment transportation scenario construction method according to any one of claims 1 to 3, characterized in that: The following units are included: The classification unit is used to obtain environmental data in the pre-transportation environment based on multi-source sensors, and to perform denoising and credibility classification on the environmental data to obtain classified environmental data, specifically including: Acquire environmental data in the pre-transportation environment through multi-source sensors, including point cloud data, real-time images, wind speed and humidity data; The noise points in the point cloud data are removed according to the multi-scale filtering algorithm, and the error areas in the point cloud data are marked and initially corrected by combining the texture features of the real-time image, including: Based on the density clustering algorithm, points whose point density in the point cloud data is lower than the preset density threshold are determined as candidate noise points; by calculating the mean distance from the point to the neighborhood, points whose mean distance is greater than the preset distance threshold are marked as noise points; Calculate the weighted scores of candidate noise points at different scales based on multi-scale spatial parameters, and remove points with scores lower than the preset score threshold; Align the coordinates of the point cloud data with the image data, project the point cloud data onto the image plane, extract the texture features in the image data, and identify the areas with abnormal texture features as error areas; For the marked error areas, the interpolation algorithm of point cloud data is used to fill in the missing points, and the point cloud feature values ​​in the area are recalculated to calibrate the corrected point cloud; Introduce a credibility assessment model for environmental data and classify the data, including: Assign a credibility score to each set of environmental data, mark the data with a credibility score higher than a preset credibility threshold as high credibility data, and mark the data with a credibility score lower than the preset credibility threshold as low credibility data; Data is stored in layers according to the credibility score. High-credibility data is used first for preliminary static scene construction. Low-credibility data is marked and stored for backup. A fusion modeling unit is used to dynamically match the texture features of the point cloud data and the real-time image data according to the classified environmental data using a multimodal data adaptive fusion algorithm, and to adjust the fusion weight according to the deep learning model to generate a static scene containing spatial geometric characteristics and texture features; Project the point cloud data, align the three-dimensional coordinates with the image plane coordinates according to the projection, and obtain the aligned point cloud data. The alignment process is as follows: In the formula, Represents the pixel coordinates of the point cloud data projected onto the image plane, represents the camera intrinsic parameter matrix, Represents the external parameter matrix, including the rotation matrix and translation vectors ; Represents the three-dimensional coordinates of point cloud data. The goal of the above formula is to complete the spatial alignment of point cloud and image, and provide a data benchmark of the coordinate system for feature fusion; Extract geometric features from the aligned point cloud data, use convolutional neural networks to extract texture features from real-time image data, use dynamic matching algorithms to match geometric features with texture features, and dynamically adjust fusion weights based on the deep learning model: In the formula, is the final eigenvalue after fusion, is the point cloud geometric feature, is the image texture feature, are dynamically adjusted fusion weights, respectively. The goal of the above formula is to fuse multimodal features to generate a static scene with geometric and texture information; The dynamic update unit is used to capture the dynamic change factors in the pre-transportation environment in real time based on the static scene with geometric and texture information, identify the dynamic area that affects the transportation path according to the dynamic change factors, and generate a real-time dynamic scene through dynamic area update.

Images