Park intelligent inspection method and system based on digital twinning

Abstract

The invention relates to the technical field of digital twinning, and discloses an intelligent park inspection method and system based on digital twinning. The method comprises the following steps: acquiring image data, point cloud data, inertial navigation data and GPS positioning data, and performing format standardization and timestamp synchronization processing to obtain a multi-dimensional environment data set; performing point cloud space alignment according to image data and the point cloud data to obtain a three-dimensional space frame; performing spatial position calibration according to inertial navigation data and the GPS positioning data to obtain positioning reference data; constructing a digital twinborn model of the park according to a three-dimensional space frame and the positioning reference data; performing path calculation and rehearsal correction according to the digital twinborn model to obtain an inspection track; and according to the inspection track and the digital twinborn model, generating an inspection instruction, so that inspection equipment performs a park inspection task. According to the method, intelligent inspection rehearsal and path optimization can be realized, and the park inspection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a park intelligent inspection method and system based on digital twins. Background Art

[0002] As a key component of modern urban development systems, intelligent park management plays a vital role in improving overall park operational efficiency, ensuring production and operational safety, and promoting optimal resource allocation. Especially against the backdrop of the ongoing advancement of smart cities and industrial upgrades, the scale of parks continues to expand. The increasing complexity of their internal structures, the diversity of their equipment systems, and the dynamic nature of their operational scenarios are placing higher demands on the intelligence and systematization of daily park operations and management. However, in practical applications, the traditional reliance on manual inspections and monitoring-based management models are no longer sufficient to meet the modern park's demand for efficient collaboration and refined operations.

[0003] Currently, the daily operation and management of industrial parks are mostly carried out through manual inspections combined with simple monitoring systems. This model has shown significant shortcomings in coping with complex environmental changes and multi-task scheduling. On the one hand, faced with the dynamically changing operating environment in the industrial park, such as the construction of temporary structures, intensive personnel flow, and frequent changes in equipment status, manual inspections are difficult to update inspection routes or adjust task priorities in a timely manner, and are prone to problems such as path failure and missed detection, which reduces the environmental adaptability to complex environmental changes. On the other hand, when the system faces concurrent task demands from multiple areas at the same time, traditional methods rely on scheduling and linear scheduling mechanisms, lacking the ability to comprehensively coordinate task priorities, equipment status, and path resources, resulting in delayed inspection responses and low resource utilization. In summary, existing industrial park inspection and monitoring methods lack the ability to adapt to dynamic environmental changes, and industrial park inspection efficiency is relatively low. Summary of the Invention

[0004] The present invention provides a digital twin-based intelligent park inspection method and system, which can dynamically update the environmental model based on real-time perception data, automatically adjust the inspection path, and improve the park inspection efficiency.

[0005] In the first aspect, in order to solve the above technical problems, the present invention provides a campus intelligent inspection method based on digital twins, comprising: Acquire image data and point cloud data within the park, acquire inertial navigation data and GPS positioning data of inspection equipment, and perform format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data to obtain a first multidimensional environmental data set; According to the image data and the point cloud data in the first multidimensional environment dataset, performing point cloud spatial alignment based on a preset point cloud registration method to obtain a three-dimensional spatial framework; Performing spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data; Constructing a digital twin model of the park based on the voxel modeling method according to the three-dimensional spatial framework and the positioning reference data; Based on the digital twin model, path calculation and preview correction are performed to obtain the inspection trajectory; According to the inspection trajectory and the digital twin model, an inspection instruction is generated, and the inspection instruction is sent to the inspection device so that the inspection device performs the park inspection task.

[0006] Preferably, performing point cloud spatial alignment based on a preset point cloud registration method according to the image data and the point cloud data in the first multidimensional environment dataset to obtain a three-dimensional spatial framework includes: According to the point cloud data, performing spatial mapping comparison based on feature points in the image data to obtain a registration error; Determine whether the registration error is less than a preset error threshold; if not, perform angle adjustment and coordinate conversion, and then perform the registration error determination again; if yes, perform data aggregation on the image data and the point cloud data to obtain an aligned data set; The reference coordinates of the alignment data set are constructed based on preset reference coordinate rules to obtain a three-dimensional space framework.

[0007] Preferably, performing spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data includes: determining, based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set, whether a signal strength of the GPS positioning data is greater than a preset signal strength threshold, and if so, using the GPS positioning data as primary positioning data; if not, using the inertial navigation data as primary positioning data; According to the main positioning data, coordinate transformation is performed based on the three-dimensional space framework to obtain positioning reference data.

[0008] Preferably, the constructing of a digital twin model of the park based on the voxel modeling method according to the three-dimensional space framework and the positioning reference data includes: Based on the characteristics of voxel grids and regular grids, the three-dimensional space frame is discretized to obtain a grid unit set; Embedding the positioning reference data into the grid unit set to obtain a grid distribution set; According to the grid distribution set, mapping is performed based on the spatial position of the park, and the position is calibrated through coordinate transformation to obtain a digital twin model.

[0009] Preferably, after constructing the digital twin model of the park based on the voxel modeling method, the method further includes: performing format standardization and timestamp synchronization again based on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data at a preset time interval to obtain a second multidimensional environment data set; performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error; It is determined whether the environmental error is greater than a preset environmental error threshold; if so, the second multi-dimensional environmental dataset is updated to the first multi-dimensional environmental dataset.

[0010] Preferably, performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error includes: The environmental error is calculated by the following formula: Where, is the environmental error, is the total number of data points of the first multidimensional environmental dataset and the second multidimensional environmental dataset, The first multidimensional environment dataset data points, The second multidimensional environment dataset data points.

[0011] Preferably, the path calculation and preview correction are performed based on the digital twin model to obtain the inspection trajectory, including: Perform object detection based on the digital twin model to obtain static obstacles and dynamic obstacles; Performing path calculation based on the static obstacles to obtain a first patrol track; Performing a patrol rehearsal based on the first patrol track and correcting the path based on the dynamic obstacle to obtain a second patrol track; According to the second patrol track, a track smoothing process is performed to obtain a patrol track.

[0012] In a second aspect, the present invention provides a digital twin-based park intelligent inspection system, comprising: a data acquisition module, configured to acquire image data and point cloud data within the park, acquire inertial navigation data and GPS positioning data of inspection equipment, and perform format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data to obtain a first multidimensional environmental data set; a point cloud registration module, configured to perform point cloud spatial alignment based on the image data and the point cloud data in the first multidimensional environment dataset and a preset point cloud registration method to obtain a three-dimensional spatial framework; a position calibration module, configured to perform spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data; A model construction module, configured to construct a digital twin model of the park based on the voxel modeling method according to the three-dimensional spatial framework and the positioning reference data; A trajectory generation module is used to perform path calculation and preview correction based on the digital twin model to obtain an inspection trajectory; An instruction generation module is used to generate inspection instructions based on the inspection trajectory and the digital twin model, and send the inspection instructions to the inspection equipment so that the inspection equipment can perform the park inspection task.

[0013] In the third aspect, the present invention also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements any one of the above-mentioned digital twin-based intelligent campus inspection methods.

[0014] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned digital twin-based intelligent campus inspection methods.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention constructs a multi-dimensional environmental dataset and combines voxel modeling with a spatial registration mechanism to achieve modeling and real-time mapping of the park's environmental structure, improving the system's adaptability to complex environmental changes. The system can perceive static and dynamic obstacles in the environment in real time, enabling online path correction and dynamic adjustment of inspection tasks, overcoming the problem of traditional manual inspections being slow to respond to temporary environmental changes.

[0016] (2) Based on the digital twin model, the present invention establishes a simulation environment that can be used for inspection path preview and dynamic simulation. It can complete path feasibility analysis and obstacle prediction before task execution, and supports smoothing and dynamic optimization of inspection trajectories, ensuring the continuity and stability of inspection equipment in high-density environments, avoiding path failure, and improving the reliability and safety of inspection task execution.

[0017] (3) The present invention can achieve unified format standardization and time synchronization processing of data, ensuring the consistency of image, point cloud, navigation and positioning information in spatial and temporal dimensions, providing high-quality data support for subsequent modeling, analysis and instruction generation, improving the integrity and scalability of the system processing capabilities, and meeting the engineering adaptability requirements of the park intelligent inspection system in actual deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a digital twin-based park intelligent inspection method provided by the first embodiment of the present invention; Figure 2 It is a structural diagram of the digital twin-based campus intelligent inspection system provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Reference Figure 1 The first embodiment of the present invention provides a campus intelligent inspection method based on digital twins, including the following steps: S11, acquiring image data and point cloud data within the campus, acquiring inertial navigation data and GPS positioning data of the inspection equipment, and performing format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data and the GPS positioning data to obtain a first multidimensional environmental data set.

[0021] S12: performing point cloud spatial alignment based on the image data and the point cloud data in the first multi-dimensional environment dataset and a preset point cloud registration method to obtain a three-dimensional spatial framework.

[0022] S13: Perform spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environment data set to obtain positioning reference data.

[0023] S14: Construct a digital twin model of the park based on the voxel modeling method according to the three-dimensional space framework and the positioning reference data.

[0024] S15: Perform path calculation and preview correction based on the digital twin model to obtain an inspection trajectory.

[0025] S16: Generate an inspection instruction based on the inspection trajectory and the digital twin model, and send the inspection instruction to the inspection device so that the inspection device performs the park inspection task.

[0026] It is worth noting that intelligent park management, as a key component of the modern urban development system, plays a vital role in improving the overall operational efficiency of the park, ensuring production and operation safety, and promoting the optimal allocation of resources.

[0027] To facilitate understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0028] In step S11, image data and point cloud data within the park are obtained, inertial navigation data and GPS positioning data of the inspection equipment are obtained, and the image data, the point cloud data, the inertial navigation data and the GPS positioning data are format-standardized and time-stamp-synchronized to obtain a first multidimensional environmental data set.

[0029] It is worth noting that in order to achieve a multi-dimensional digital representation of the park's spatial environment, the system needs to acquire multi-source basic data, including image data, point cloud data, inertial navigation data, and GPS positioning data. This data can be obtained through sensing devices installed on inspection equipment or through fixed-point collection terminals deployed within the park, thereby ensuring complete coverage of both dynamic and static environmental information. Specifically, the inspection equipment can be a mobile robot, unmanned vehicle, or drone platform equipped with a high-resolution camera and lidar system. It can operate automatically within a set path or area to collect spatial visual information and geometric structure data of the target area.

[0030] In one possible implementation, image data can be acquired through high-definition cameras mounted on the front or top of inspection equipment, capturing images of the scene wherever the equipment passes. Another possible implementation involves relying on pre-configured intelligent video surveillance nodes within the park to capture visual information within the area through real-time or periodic capture. The image data features high clarity, stable frame rates, and minimal image distortion. The image format is standardized as JPEG or PNG, and is notably accompanied by the device ID and capture timestamp.

[0031] In one feasible approach, point cloud data can be acquired using LiDAR or structured light depth cameras. In mobile acquisition mode, inspection equipment equipped with 3D LiDAR can capture continuous spatial point cloud frames during operation. Each frame contains a large number of 3D coordinate points that reflect the geometric structure of the environment. In static acquisition mode, the system can deploy fixed radar nodes in key areas of the campus for periodic scanning, acquiring spatial topography data with time-series characteristics. This data can be stored in PCL (Point Cloud Library) format, with spatial location tags and acquisition time stamps.

[0032] In one feasible approach, the system acquires inertial navigation data through an inertial measurement unit (IMU), which can be deployed within inspection robots, unmanned vehicles, or wearable devices. Its output includes three-axis acceleration, three-axis angular velocity, and inferred attitude data. The frequency is typically set above 100Hz to ensure time series continuity and trajectory calculation accuracy. Each data set also includes an absolute timestamp and device identification code, enabling alignment and fusion with other data sources.

[0033] In one feasible approach, GPS positioning data can be collected using a multi-mode GNSS module, which can be embedded in mobile inspection equipment or linked to reference base stations deployed on campus (such as RTK differential stations) to achieve centimeter-level positioning accuracy. For example, in areas with severe signal obstruction, such as tunnels or factory interiors, the system can also be configured with UWB or inertial navigation fusion compensation devices to enhance positioning integrity. Positioning data can be structured in a standard NMEA format and converted to the WGS-84 coordinate system to ensure spatial consistency.

[0034] Specifically, to achieve spatiotemporal alignment and multi-source fusion of data, after completing the aforementioned data collection, the system must perform unified format standardization and timestamp synchronization on all sensor data. Standardization operations include image resizing (for example, to 640×480 resolution), point cloud filtering and resampling (e.g., uniform point spacing for Voxel Grid sampling), and IMU attitude data unit normalization (angular velocity to deg / s, linear acceleration to m / s²). Specifically, timestamp synchronization utilizes a unified UTC reference clock. All sensor data is tied to the precise acquisition time and aligned through interpolation and delay correction. This ensures a consistent, time-series, uniformly formatted first-order multidimensional environmental dataset, providing comprehensive data support for subsequent spatial modeling, digital twin construction, and path scheduling analysis.

[0035] In step S12, point cloud space alignment is performed based on the image data and the point cloud data in the first multi-dimensional environment dataset and a preset point cloud registration method to obtain a three-dimensional space framework.

[0036] Preferably, performing point cloud spatial alignment based on a preset point cloud registration method according to the image data and the point cloud data in the first multidimensional environment dataset to obtain a three-dimensional spatial framework includes: According to the point cloud data, performing spatial mapping comparison based on feature points in the image data to obtain a registration error; Determine whether the registration error is less than a preset error threshold; if not, perform angle adjustment and coordinate conversion, and then perform the registration error determination again; if yes, perform data aggregation on the image data and the point cloud data to obtain an aligned data set; The reference coordinates of the alignment data set are constructed based on preset reference coordinate rules to obtain a three-dimensional space framework.

[0037] Specifically, the system needs to extract representative feature points from the image data. These feature points can be obtained using image processing algorithms like SIFT (Scale-Invariant Feature Transform), ORB (Oriented Fast Binary Descriptors), or SURF (Speeded Up Robust Features). These feature points should include distinct corners, edges, or areas of texture variation. Furthermore, point cloud data provides geometric structural information about the same environment. Based on the extracted feature points from the image, the system must find corresponding spatial locations in the 3D point cloud to achieve preliminary spatial mapping and comparison. It's important to note that image feature points are essentially pixel coordinates in 2D space, while point cloud data is represented by (x, y, z) coordinates in 3D space. Therefore, to achieve this mapping, the system must perform coordinate system integration and geometric projection. First, the camera's intrinsic parameter matrix (including focal length and principal point position) and extrinsic parameter matrix (position and pose relative to the point cloud acquisition sensor) are used to back-project the image coordinates into the corresponding 3D spatial directions. Next, based on time synchronization information, the system searches for the point cloud data frame closest to the image frame. A projection cone (camera frustum) based on the current image frame is constructed to define candidate mapping areas. Then, for each image feature point, a projection formula is used to transform the feature point from image coordinates into a 3D ray direction originating from the camera's optical center. Specifically, a back-projection method based on the camera's intrinsic parameter matrix is ​​used to transform the image feature point from a 2D image coordinate system into a ray direction in 3D space. Specifically, a homogeneous coordinate vector is first constructed using the pixel coordinates of the image feature point. This point is then mapped into a 3D direction vector in the camera coordinate system by multiplying it by the inverse of the camera's intrinsic parameter matrix. This vector represents the 3D ray direction originating from the camera's optical center and passing through the image feature point. It can then be normalized for spatial matching and point cloud mapping operations. The point cloud data within the projection cone is then traversed to find the closest point on the ray direction (e.g., calculating the minimum perpendicular distance from each point to the ray). If a point is found that meets a distance threshold (e.g., 1 cm), it is considered the spatial matching point for the image feature point. It is worth noting that to ensure effective matching between image feature points and 3D point cloud data, the system sets a spatial matching distance threshold, which is used to determine whether a point cloud can be considered a corresponding point in 3D space for the image feature point. The distance threshold can be selected based on the calibration accuracy of the image sensor and the ranging error range of the lidar.

[0038] Specifically, in the process of aligning point cloud and image features, the system first establishes a set of spatial corresponding point pairs, assuming that the set is: , where The estimated 3D points are obtained by back-projecting the image feature points; is the actual point that matches it in the point cloud data; is the total number of valid matching points. Specifically, the registration error can be calculated using the Euclidean error method. For each pair of matching points, the distance between their three-dimensional coordinates is calculated: , Reflects the degree of deviation of a single point pair in three-dimensional space; are the coordinates of the source and target points in the three-dimensional coordinate system.

[0039] It's worth noting that registration errors can arise from errors in image feature extraction, such as low light levels and occlusions that lead to misaligned feature point locations; camera and lidar calibration errors, namely projection errors caused by inaccurate internal and external parameters; and point cloud sparsity or occlusion errors, which can lead to mismatching of some 3D structures. The error threshold should be determined based on the actual application scenario. For example, in indoor positioning, a registration RMSE threshold of 2-5 cm is often used; in large-scale campus building modeling, the tolerance can be relaxed to 10 cm.

[0040] Specifically, if the error exceeds the limit, angle adjustment and coordinate transformation are performed. Angle adjustment involves constructing and applying a rotation matrix to correct for deviations in attitude between the image camera's perspective and the point cloud coordinate system. Coordinate transformation involves translation and scaling to correct for origin offsets or scale differences between sensors. After adjustment, the system recalculates the registration error. If the threshold condition is still not met, iterations continue until the required accuracy is met or the maximum number of iterations is reached.

[0041] Specifically, if the error falls below a threshold, the system considers the image and point cloud data aligned and fuses them to form a unified, aligned dataset. This dataset not only preserves the high-precision texture information in the image but also incorporates the depth and spatial structure information provided by the point cloud, forming the foundation for subsequent modeling and positioning. Finally, the system unifies the coordinates of the aligned dataset and performs 3D reconstruction based on pre-set reference coordinate rules, constructing a complete 3D spatial framework.

[0042] It's worth noting that the preset reference coordinate rules can be based on the original park design or a unified coordinate system (such as WGS-84 or UTM projection coordinates). Through coordinate mapping and transformation, all data is normalized to a unified reference. In subsequent steps, the system can further perform voxel modeling, navigation path planning, and multimodal perception analysis tasks based on this framework.

[0043] In step S13, spatial position calibration is performed based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environment data set to obtain positioning reference data.

[0044] Preferably, performing spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data includes: determining, based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set, whether a signal strength of the GPS positioning data is greater than a preset signal strength threshold, and if so, using the GPS positioning data as primary positioning data; if not, using the inertial navigation data as primary positioning data; According to the main positioning data, coordinate transformation is performed based on the three-dimensional space framework to obtain positioning reference data.

[0045] Specifically, the core objective of this step is to calibrate the spatial position of the inspection equipment based on the inertial navigation data and GPS positioning data in the first multidimensional environmental dataset, and output positioning reference data consistent with the three-dimensional spatial framework. During implementation, the system first determines whether the GPS positioning data signal strength meets the usage requirements. In the present invention, the system pre-sets a signal strength threshold, which can be determined by a combination of indicators. It is worth noting that the GPS module's signal-to-noise ratio (SNR) can be used as the primary reference indicator. SNR reflects the strength of satellite signals relative to noise and is a common metric for evaluating GPS data quality. Alternatively, HDOP (Horizontal Dilution of Precision) can be introduced as a supplementary criterion for determining spatial positioning error. A smaller HDOP value indicates a more optimized satellite geometric distribution and a smaller positioning error. For example, the SNR (Signal-to-Noise Ratio) can be set to be greater than 35 dBHz, the HDOP (Horizontal Dilution of Precision) less than 1.5, or the number of visible satellites to be at least six. When the GPS data meets this threshold, it indicates that it has high positioning accuracy and reliability in the current environment, and the system uses GPS data as the primary positioning source. If it does not meet the threshold, the system automatically selects inertial navigation data as the primary positioning basis, and uses the accelerometer and gyroscope data in the IMU to calculate the position, thereby ensuring that position information can be continuously obtained even indoors or in signal-blocked areas.

[0046] Specifically, after the system selects a primary positioning source, its spatial coordinates must be aligned with the existing three-dimensional spatial framework. If GPS data is used as the primary source, its original latitude and longitude format (exemplarily the WGS-84 coordinate system) is converted to a local unified coordinate system, such as the UTM or ENU coordinate system. This conversion process involves using geographic projection formulas (e.g., by approximating the difference between the latitude and longitude points and the central meridian of the current projection zone to obtain accurate projected plane coordinates) or invoking a geographic information database for automatic conversion, ensuring that the location points can be directly projected into the spatial coordinate framework used by the digital twin environment. Specifically, the geographic information database can be based on the WGS-84 geodetic coordinate system (EPSG:4326), which describes spatial position relationships globally based on the Earth ellipsoid model. Conversely, if the system uses IMU data as the primary source, the current position is inferred from the initial known position point by double integration of the three-axis acceleration, and the device attitude is estimated by combining the three-axis angular velocity. Zero-speed update (ZUPT) and filtering mechanisms are also introduced to suppress accumulated errors, ensuring that the inertial guidance results maintain sufficient accuracy over a short period of time. In one feasible approach, the system first acquires the raw data from the IMU at each time step, including linear acceleration and angular velocity along the x, y, and z axes. The system then performs a preliminary attitude estimation step, integrating the angular velocity information to construct the current attitude angle (such as Euler angles or quaternions) and calculates the projection direction of the acceleration in the global coordinate system. Next, the linear acceleration is integrated for the first time to obtain the three-dimensional velocity vector; the velocity is then integrated a second time to obtain the position increment relative to the initial point. It is worth noting that the inertial navigation results must be mapped to the initial attitude and position information to ensure their usability within the three-dimensional space framework.

[0047] Specifically, after the coordinate conversion is completed, the spatial position, attitude, timestamp and error estimate of the main positioning source are integrated into a unified structure of positioning reference data for subsequent path planning and digital twin model construction modules to call. For example, when the inspection vehicle is operating in an open area of ​​the park, the GPS signal strength is high, and the system uses its positioning results to convert into UTM coordinate input; when the equipment enters an underground passage or factory building and other severely obstructed areas, the GPS signal drops to an SNR below 25 dBHz, and the system automatically switches to the IMU-calculated trajectory and connects the position with the previous known point, thereby achieving seamless connection and accuracy assurance of positioning data throughout the inspection process. In summary, step S13 effectively improves the spatial perception capability of the park inspection equipment by dynamically screening and aligning the spatial coordinates of multi-source positioning information, providing accurate and stable spatial position information support for subsequent task execution.

[0048] In step S14, a digital twin model of the park is constructed based on the voxel modeling method according to the three-dimensional space framework and the positioning reference data.

[0049] Preferably, the constructing of a digital twin model of the park based on the voxel modeling method according to the three-dimensional space framework and the positioning reference data includes: Based on the characteristics of voxel grids and regular grids, the three-dimensional space frame is discretized to obtain a grid unit set; Embedding the positioning reference data into the grid unit set to obtain a grid distribution set; According to the grid distribution set, mapping is performed based on the spatial position of the park, and the position is calibrated through coordinate transformation to obtain a digital twin model.

[0050] It's worth noting that the system's core goal is to construct a digital twin of the park using voxel modeling, based on the established 3D spatial framework and positioning benchmark data, to achieve an efficient and computable representation of the park's spatial structure. This process relies on voxel meshing technology, discretizing the continuous 3D space into regular small units and mapping positioning and structural information into them, ultimately establishing a digital modeling system that precisely aligns with the real-world spatial structure. The implementation process primarily involves three key operations: voxel mesh discretization, positioning benchmark data embedding, and spatial mapping calibration, as follows: First, the system discretizes the campus environment into voxels based on a three-dimensional spatial framework. It's worth noting that voxels, or three-dimensional pixels, are a way of dividing three-dimensional space using cubes as basic units and are commonly used to represent building volumes, environmental boundaries, or spatial obstacles. Based on the campus's spatial scale and modeling accuracy requirements, the system sets a preset voxel side length threshold, such as 0.5 meters or 0.2 meters, to control the spatial dimensions of each voxel unit in the X, Y, and Z directions. The choice of voxel size depends on the actual application requirements. If the goal is coarse-grained path planning, 1.0-meter voxels can be used; if the goal is high-precision inspection navigation or detailed structural modeling, a side length of no more than 0.2 meters is recommended. Based on this, the system establishes a regular grid structure, dividing the entire three-dimensional spatial framework into a collection of voxel units. Each unit is identified by its center point coordinates, boundary range, and unique ID, forming the basic spatial mapping unit.

[0051] Then, the system embeds the positioning reference data from step S13 into the grid unit to realize the combination of spatial pose data and environmental structure. The specific method is: for each positioning reference data point ( , , ), the system calculates the voxel length based on its coordinate value Perform integer division to determine the grid cell index (i, j, k) to which it belongs, namely: In the formula 、 The minimum value of the three-dimensional space boundary. Positioning data includes the location, attitude angle, and path trajectory of the inspection equipment. After embedding, each grid cell can record multiple fields of information.

[0052] Finally, the system performs spatial mapping and coordinate calibration operations on the grid set after the positioning data is embedded. Since the voxel grid itself is generated based on regular coordinates, and the park may have tilted areas, non-orthogonal structures, or local fine-tuning, the grid coordinate system needs to be mapped to the park's geographic reference system (such as WGS-84 or the local building blueprint coordinate system). In one feasible way, the system calls a coordinate transformation matrix or a registration algorithm (such as an affine transformation or a rigid body transformation) to convert the positions of all grid cells to a coordinate reference that is consistent with the digital twin model, thereby achieving the final adjustment of spatial coincidence. After the mapping is completed, each voxel unit has a clear geometric position, associated attributes, and contextual information in the digital space, thus forming a complete digital twin model.

[0053] For example, in a typical logistics park, the system divides the park area into a grid of 0.5-meter cubes, forming a basic three-dimensional spatial framework. Inspection equipment and static cameras collect positioning data within the area. The system assigns each location point to a corresponding grid cell based on its spatial trajectory, and records the device's transit time, azimuth, and dwell time within the cell. Once the system completes data collection for the entire area, it reconstructs a digital twin model of the park on a digital platform, complete with visualization, attribute labels, and consistent spatial structure, enabling virtual-reality mapping and subsequent task simulation.

[0054] Preferably, after constructing the digital twin model of the park based on the voxel modeling method, the method further includes: performing format standardization and timestamp synchronization again based on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data at a preset time interval to obtain a second multidimensional environment data set; performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error; It is determined whether the environmental error is greater than a preset environmental error threshold; if so, the second multi-dimensional environmental dataset is updated to the first multi-dimensional environmental dataset.

[0055] Specifically, after the digital twin model is constructed, dynamic correction based on multidimensional environmental data is required to ensure its continued consistency with the actual environment. This is done by collecting new images, point clouds, inertial navigation, and GPS data at preset time intervals, unifying their formats and aligning their time to generate a second multidimensional environmental dataset. This dataset is then compared with the original first multidimensional environmental dataset to assess whether the model's current spatial accuracy has shifted. If the environmental error exceeds a set threshold, the system automatically updates the environmental data to achieve dynamic, synchronous updates of the digital twin model.

[0056] Specifically, the system first recalls the perception system based on a set refresh cycle (e.g., every 10 minutes, every hour, or a trigger period set according to the mission cadence) to acquire a full set of data, including image frame sequences, lidar point cloud data, IMU acceleration and angular velocity information, and GPS position coordinates. To ensure comparability and data integration, the system performs the same data normalization and synchronization on the collected raw data, including image resizing (e.g., 640×480 pixels), point cloud filtering and sampling (e.g., Voxel Grid filtering), IMU unit normalization (acceleration to m / s², angular velocity to deg / s), and time alignment of multiple source data using a unified UTC timestamp. This ultimately creates a second, structurally consistent multidimensional environmental dataset. Subsequently, a difference comparison operation is performed using the first and second multidimensional environmental datasets as input. The core objective is to assess whether the spatial structure reflected by the current model is still consistent with the real-world environment.

[0057] In one possible implementation, the strategy for difference comparison varies slightly depending on the data type. For image data, structural similarity metrics (such as SSIM) or pixel cosine similarity are used to measure image variation. For point cloud data, the ICP algorithm is used to calculate the rigid body alignment error between two point cloud frames, yielding the root mean square error (RMSE). For IMU and GPS data, the Euclidean distance or attitude angle difference between trajectory points can be compared, respectively. To unify the output error metrics, the system performs a weighted normalization on each type of error result to form the final environmental error value.

[0058] Specifically, the environmental error value is compared with a preset environmental error threshold to determine whether to trigger a model update. It's worth noting that the environmental error threshold is determined based on three factors: the spatial accuracy of the sensor (e.g., 0.05m resolution for lidar and ±0.2° / s standard deviation for IMU drift); the tolerance range of the model application (e.g., 0.1m tolerance for inspection path accuracy); and statistical analysis of errors in actual deployment scenarios (e.g., a 95% confidence interval for error obtained through preliminary sampling). If the system detects that the environmental error exceeds the preset environmental error threshold, it immediately executes an environmental data update, replacing the original first multidimensional environmental dataset with the current second multidimensional environmental dataset as the new baseline for subsequent model updates and path simulations. This prevents model lag or failure caused by dynamic changes in the campus (e.g., temporary structures, vehicle repositioning, scene lighting changes, etc.), thereby improving the timeliness and stability of the digital twin platform.

[0059] In another achievable manner, performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error includes: The environmental error is calculated by the following formula: Where, is the environmental error, is the total number of data points of the first multidimensional environmental dataset and the second multidimensional environmental dataset, The first multidimensional environment dataset data points, The second multidimensional environment dataset data points.

[0060] Specifically, the core idea of ​​the formula is to use the average of the point-by-point absolute value differences as a measure of the degree of overall environmental change. It has the advantages of being simple in calculation and sensitive to local mutations. It is particularly suitable for multi-source perception data scenarios with consistent structure and matching dimensions.

[0061] It is worth mentioning that It can represent different types of data items, such as image grayscale values, spatial coordinates of point clouds, acceleration amplitudes of IMUs, or coordinate values ​​of GPS. In particular, the system must normalize different types of data items before performing calculations (such as unifying coordinate units to meters and image values ​​to floating-point numbers between 0 and 1) to ensure that the errors between different physical quantities are comparable. In the case of spatial point cloud data, this calculation method can be used directly to compare the changes in the Euclidean distance between two frames of point clouds at corresponding spatial sampling positions. In the case of image data, it corresponds to pixel-level brightness or texture changes, reflecting changes in lighting, occlusion, or structure. In the comparison of inertial navigation data and GPS trajectories, this formula is equivalent to the average amplitude of the corresponding position change at the sampling moment.

[0062] In step S15, path calculation and preview correction are performed based on the digital twin model to obtain an inspection trajectory.

[0063] Preferably, the path calculation and preview correction are performed based on the digital twin model to obtain the inspection trajectory, including: Perform object detection based on the digital twin model to obtain static obstacles and dynamic obstacles; Performing path calculation based on the static obstacles to obtain a first patrol track; Performing a patrol rehearsal based on the first patrol track and correcting the path based on the dynamic obstacle to obtain a second patrol track; According to the second patrol track, a track smoothing process is performed to obtain a patrol track.

[0064] Specifically, the system first performs object detection within the digital twin model to determine the distribution of static and dynamic obstacles within the current campus environment. Static obstacles can include time-invariant structures such as building walls, pipe supports, and fixed equipment, whose locations are clearly represented in the three-dimensional voxel grid during the modeling process. Dynamic obstacles, on the other hand, include mobile devices, pedestrians, and temporary storage. Their status is updated based on real-time data, determined by comparing differences with the latest image, point cloud, and positioning data. Furthermore, the system can identify dynamic objects within images through background modeling and motion detection algorithms (such as frame difference and optical flow), while simultaneously detecting real-time moving objects in conjunction with changes in point cloud density to construct an obstacle label layer. Once obstacle detection is complete, the system maps all obstacle information into a spatial grid structure, providing traversable area identification for the path planning phase.

[0065] In the second stage, the system performs path calculation based on the starting and target positions of the inspection equipment and refers to the static obstacle layer to generate a preliminary path, namely the first patrol track. Path planning uses heuristic search algorithms such as A-star or Dijkstra algorithms to traverse the three-dimensional voxel grid constructed by the digital twin model. During the path search process, the system will mark the grid cells corresponding to the static obstacle area as inaccessible to ensure that the planned path avoids fixed obstacles. Specifically, the system can also introduce a pass cost function to set the grid cells adjacent to the obstacle as high-cost areas to increase the safety margin of the path. At this time, a pass boundary threshold can be set, for example 0.3 meters, indicating that the path node must be at least 30 centimeters away from the static obstacle to ensure the actual passage safety of the equipment.

[0066] After generating the first patrol track, the system enters the patrol rehearsal and path correction stage. This step simulates the driving process of the patrol equipment on the preliminary path to detect in real time whether there are dynamic obstacles interfering along the way. The system simulates the equipment moving on the path at a fixed speed, and at the same time calls in the latest dynamic obstacle position data for collision prediction. If it is predicted that certain points on the path will spatially overlap with dynamic obstacles, the path correction mechanism is triggered. Exemplarily, the correction method can use the RRT (Rapidly-Exploring Random Tree) algorithm to perform local replanning operations on the affected path segments, thereby generating an alternative path that bypasses dynamic obstacles, which is called the second patrol track. It is worth noting that in order to improve adaptability, the system can set a dynamic interference trigger threshold. For example, if "the distance between the trajectory point and the obstacle is expected to be less than 0.2 meters within 2 seconds", it is considered that interference exists and replanning is required.

[0067] Finally, the system performs trajectory smoothing on the second patrol track to improve path executableness and device motion stability. In one feasible approach, smoothing can be performed using cubic spline interpolation, Bezier curve fitting, or minimum acceleration optimization. Smoothing must minimize the number of path turning points and sudden changes in steering angles while ensuring obstacle avoidance constraints, thereby improving device execution efficiency and reducing energy consumption. In addition, the system can set a smoothing curvature threshold at this stage, such as a curvature change of no more than 0.8 radians per meter, to prevent sharp turns in the path from burdening the device's steering system. The final output inspection trajectory is not only path-coherent and physically feasible, but also has spatial obstacle avoidance and dynamic adaptability.

[0068] In step S16, an inspection instruction is generated according to the inspection trajectory and the digital twin model, and the inspection instruction is sent to the inspection device so that the inspection device performs the park inspection task.

[0069] Based on previously generated inspection trajectories and the park's digital twin model, the system generates and issues inspection instructions, ultimately driving the inspection equipment to perform actual park inspections. The key lies in converting the calculated trajectory information into a sequence of control commands that the inspection equipment can recognize, ensuring stable operation in the real physical environment.

[0070] In one feasible approach, the system first encodes the trajectory based on the geometric structure of the inspection trajectory (including parameters such as the sequence of position points, steering angle, and travel speed). Each trajectory point will be converted into a corresponding control instruction structure, the typical format of which includes: target coordinates (X, Y, Z), steering angle (Yaw), forward speed (v), and execution timestamp (t). For example, during the encoding process, the system can dynamically adjust the control parameters by referring to the environmental characteristics recorded in the digital twin model (such as channel width, ground material, slope angle, etc.). For example, in areas with slopes, the system may reduce the target speed to ensure the stability of the equipment; in narrow areas or corners, the steering angle increment is adjusted to avoid collision risks.

[0071] The system then encapsulates this sequence of control command structures as an inspection command set in a communication protocol format (such as MQTT, ROS Topic, or a custom UDP / TCP message) and sends it to the inspection device terminal via 5G, Wi-Fi, or a dedicated communication link. Upon receiving the commands, the device parses and executes the control actions in timestamp order, enabling point-by-point navigation and path following.

[0072] In summary, in the present invention, the method can construct a unified first multi-dimensional environmental data set by acquiring park image data, point cloud data, inertial navigation data and GPS positioning data, and performing format standardization and timestamp synchronization processing on the above multi-source data; then, based on the feature point matching relationship between the image data and the point cloud data, a point cloud registration operation is performed to generate a three-dimensional space framework, and at the same time, spatial position calibration is performed in combination with inertial navigation and GPS data to obtain positioning reference data; on this basis, the three-dimensional space framework is discretized using a voxel modeling method to construct a digital twin model with a spatial mapping relationship; further, path planning, static and dynamic obstacle detection, inspection rehearsal and trajectory correction are performed based on the model, and finally an optimized inspection trajectory and task instructions are generated. Through the above multi-step fusion processing, the method can realize intelligent inspection rehearsal and path optimization, and improve the efficiency of park inspection.

[0073] Reference Figure 2 The second embodiment of the present invention provides a park intelligent inspection system based on digital twins, including: a data acquisition module, configured to acquire image data and point cloud data within the park, acquire inertial navigation data and GPS positioning data of inspection equipment, and perform format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data to obtain a first multidimensional environmental data set; a point cloud registration module, configured to perform point cloud spatial alignment based on the image data and the point cloud data in the first multidimensional environment dataset and a preset point cloud registration method to obtain a three-dimensional spatial framework; a position calibration module, configured to perform spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data; A model construction module, configured to construct a digital twin model of the park based on the voxel modeling method according to the three-dimensional spatial framework and the positioning reference data; A trajectory generation module is used to perform path calculation and preview correction based on the digital twin model to obtain an inspection trajectory; An instruction generation module is used to generate inspection instructions based on the inspection trajectory and the digital twin model, and send the inspection instructions to the inspection equipment so that the inspection equipment can perform the park inspection task.

[0074] It should be noted that the digital twin-based campus intelligent inspection system provided in an embodiment of the present invention is used to execute all the process steps of the digital twin-based campus intelligent inspection method in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0075] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a program for a digital twin-based intelligent inspection method for a campus. When the processor executes the computer program, the steps in each of the above-mentioned embodiments of the digital twin-based intelligent inspection method for a campus are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the instruction generation module.

[0076] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0077] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0078] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the electronic device and connects various parts of the entire electronic device using various interfaces and lines.

[0079] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0080] If the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0081] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0082] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A digital twin-based intelligent inspection method for a park, characterized in that: include: Acquire image data and point cloud data within the park, acquire inertial navigation data and GPS positioning data of inspection equipment, and perform format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data to obtain a first multidimensional environmental data set; According to the image data and the point cloud data in the first multidimensional environment dataset, performing point cloud spatial alignment based on a preset point cloud registration method to obtain a three-dimensional spatial framework; Performing spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data; Constructing a digital twin model of the park based on the voxel modeling method according to the three-dimensional spatial framework and the positioning reference data; Based on the digital twin model, path calculation and preview correction are performed to obtain the inspection trajectory; According to the inspection trajectory and the digital twin model, an inspection instruction is generated, and the inspection instruction is sent to the inspection equipment to perform the park inspection task.

2. The digital twin-based park intelligent inspection method according to claim 1 is characterized in that: The step of performing point cloud spatial alignment based on the image data and the point cloud data in the first multi-dimensional environment dataset and on a preset point cloud registration method to obtain a three-dimensional spatial framework includes: According to the point cloud data, performing spatial mapping comparison based on feature points in the image data to obtain a registration error; Determine whether the registration error is less than a preset error threshold; if not, perform angle adjustment and coordinate conversion, and then perform the registration error determination again; if so, perform data aggregation on the image data and the point cloud data to obtain an aligned data set; The reference coordinates of the alignment data set are constructed based on preset reference coordinate rules to obtain a three-dimensional space framework.

3. The digital twin-based park intelligent inspection method according to claim 1 is characterized in that: The performing of spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environment data set to obtain positioning reference data includes: determining, based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set, whether a signal strength of the GPS positioning data is greater than a preset signal strength threshold; if so, using the GPS positioning data as primary positioning data; and if not, using the inertial navigation data as primary positioning data; According to the main positioning data, coordinate transformation is performed based on the three-dimensional space framework to obtain positioning reference data.

4. The digital twin-based intelligent inspection method for a park according to claim 1 is characterized in that: The method of constructing a digital twin model of the park based on the voxel modeling method according to the three-dimensional space framework and the positioning reference data includes: Based on the characteristics of voxel grids and regular grids, the three-dimensional space frame is discretized to obtain a grid unit set; Embedding the positioning reference data into the grid unit set to obtain a grid distribution set; According to the grid distribution set, mapping is performed based on the spatial position of the park, and the position is calibrated through coordinate transformation to obtain a digital twin model.

5. The digital twin-based intelligent inspection method for a park according to claim 1 is characterized in that: After constructing the digital twin model of the park based on the voxel modeling method, the following steps are also included: performing format standardization and timestamp synchronization again based on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data at a preset time interval to obtain a second multidimensional environment data set; performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error; It is determined whether the environmental error is greater than a preset environmental error threshold; if so, the second multi-dimensional environmental dataset is updated to the first multi-dimensional environmental dataset.

6. The digital twin-based intelligent inspection method for a park according to claim 5 is characterized in that: The performing a difference comparison based on the first multidimensional environmental dataset and the second multidimensional environmental dataset to obtain an environmental error includes: The environmental error is calculated by the following formula: Where, is the environmental error, is the total number of data points of the first multidimensional environmental dataset and the second multidimensional environmental dataset, The first multidimensional environment dataset data points, The second multidimensional environment dataset data points.

7. The digital twin-based park intelligent inspection method according to claim 1 is characterized in that: The path calculation and preview correction are performed based on the digital twin model to obtain the inspection trajectory, including: Perform object detection based on the digital twin model to obtain static obstacles and dynamic obstacles; Performing path calculation based on the static obstacles to obtain a first patrol track; Performing a patrol rehearsal based on the first patrol track and correcting the path based on the dynamic obstacle to obtain a second patrol track; According to the second patrol track, a track smoothing process is performed to obtain a patrol track.

8. A digital twin-based park intelligent inspection system, characterized by: include: a data acquisition module, configured to acquire image data and point cloud data within the park, acquire inertial navigation data and GPS positioning data of inspection equipment, and perform format standardization and timestamp synchronization processing on the image data, the point cloud data, the inertial navigation data, and the GPS positioning data to obtain a first multidimensional environmental data set; a point cloud registration module, configured to perform point cloud spatial alignment based on the image data and the point cloud data in the first multidimensional environment dataset and a preset point cloud registration method to obtain a three-dimensional spatial framework; a position calibration module, configured to perform spatial position calibration based on the inertial navigation data and the GPS positioning data in the first multi-dimensional environmental data set to obtain positioning reference data; A model construction module, configured to construct a digital twin model of the park based on the voxel modeling method according to the three-dimensional spatial framework and the positioning reference data; A trajectory generation module is used to perform path calculation and preview correction based on the digital twin model to obtain an inspection trajectory; An instruction generation module is used to generate inspection instructions based on the inspection trajectory and the digital twin model, and send the inspection instructions to the inspection equipment so that the inspection equipment can perform the park inspection task.

9. An electronic device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the digital twin-based campus intelligent inspection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the digital twin-based campus intelligent inspection method as described in any one of claims 1 to 7.

Images