Real-time working condition three-dimensional dynamic visual monitoring system for unmanned cleaning vehicle

Abstract

The application discloses a real-time working condition three-dimensional dynamic visual monitoring system for unmanned cleaning vehicles, relates to the technical field of unmanned cleaning vehicle working condition monitoring, and comprises a modeling monitoring module, a region division module, a feature evaluation module, a strategy regulation module and a visual recording module.The region division module extracts a spatial region with model occlusion residues in the case that model occlusion residues exist in continuous scene updating, uniformly divides the spatial region, and generates a plurality of structure calibration zones.The feature evaluation module extracts spatial modeling quality information of each structure calibration zone, analyzes after extraction, evaluates the coverage distortion degree of each structure calibration zone, and classifies each structure calibration zone according to the evaluation result.The application solves the problem that occlusion distortion cannot be recognized and misupdate in continuous modeling, and realizes accurate evaluation and dynamic regulation of local model updating.

Description

Technical Field

[0001] This invention relates to the field of unmanned sweeper operation condition monitoring technology, specifically to a real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers. Background Technology

[0002] With the promotion of intelligent urban sanitation systems, unmanned sweepers are increasingly widely used in urban road cleaning. During operation, sweepers need to continuously perceive their surrounding environment, provide feedback on their own status, and maintain efficient linkage with the cloud platform. To achieve visualization and intelligent supervision of the operation process, the system typically employs 3D modeling technology, image rendering technology, and multi-source sensor data fusion to present the sweeper's location information, operating path, sensor status, and other dynamic operating conditions in real-time using 3D graphics. By integrating data sources such as high-precision maps, LiDAR point clouds, and vehicle-mounted camera images, a 3D virtual environment with spatial fidelity and interactivity can be constructed to support remote monitoring, task scheduling, and system linkage. These systems are largely based on key technologies such as graphics and image processing, virtual simulation modeling, and real-time data synchronization, improving the observability and intelligent management level of unmanned sweeper operation.

[0003] Existing real-time 3D dynamic visualization monitoring technology for unmanned sweepers primarily achieves full-process visual supervision by constructing a digital twin 3D scene, integrating vehicle and environmental sensor data, and utilizing graphics and image processing techniques. This technology first collects real-time location, attitude, environmental perception, and operational status parameters (such as speed, battery level, task status, and sweeping path) of the sweeper using sensors such as LiDAR, cameras, GPS, and IMU. Second, it establishes a 3D spatial model of the actual operating area using high-precision map data, oblique photography, and point cloud data, and digitally reconstructs the real scene using 3D modeling software (such as Blender or Unity). Next, the system dynamically overlays and renders the collected vehicle status data with the scene model, and uses a graphics rendering engine to achieve visual output, generating a real-time updated 3D dynamic interface on the monitoring terminal to intuitively present the sweeper's location, trajectory, and operational status in complex road environments. In addition, the system also includes a data transmission and processing module, responsible for synchronizing vehicle-side data to the cloud via wireless communication networks (such as 4G / 5G). The cloud then performs data parsing, generates visual driving commands, and displays multi-vehicle collaborative data, thereby achieving three-dimensional dynamic visualization monitoring of the entire operation process of the unmanned sweeper. The entire process covers six key stages: data acquisition, modeling and reconstruction, data fusion, graphics rendering, visualization output, and cloud synchronization.

[0004] The existing technology has the following shortcomings:

[0005] During the continuous operation of the unmanned sweeper and the uploading of environmental perception data to the cloud, the system needs to perform real-time incremental updates to the 3D scene model. If the vehicle is in an area with occlusion factors, such as shadows near the vehicle body, road dust or fog, or low-light blind spots, the newly acquired images or point cloud data may have missing or distorted information in local areas. This can lead to spatial structure misalignment or blurred surface texture overlap when superimposed on the existing 3D model. Because these occluded areas cannot provide clear and complete spatial structure information, the system, without recognizing the data quality risks in these areas, still directly uses them to replace the old model content according to the "most recent" strategy, resulting in incorrect model merging or unreasonable overlay. Existing technology cannot dynamically adjust the update and replacement strategy for old model areas based on the degree of local overlay distortion caused by residual model occlusion during continuous scene updates. This results in problems such as breaks, deformations, and texture misalignments in local areas of the 3D model, affecting the monitoring personnel's accurate judgment of the operational space structure, reducing the rationality of the scheduling strategy, and misleading trajectory recording and task backtracking, thus weakening the overall visualization reliability and operational accuracy of the system.

[0006] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers, in order to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers, comprising a modeling and monitoring module, a region division module, a feature evaluation module, a strategy control module, and a visual recording module;

[0009] The modeling and monitoring module uses real-time image and point cloud data generated by the unmanned sweeper during its operation to perform continuous modeling and updating of the 3D scene, and monitors the continuous modeling and updating process of the 3D scene in real time to identify whether there is model occlusion residue in the continuous scene update.

[0010] The region segmentation module extracts the spatial regions with residual model occlusion during continuous scene updates, and divides them evenly to generate several structural calibration regions.

[0011] The feature evaluation module extracts spatial modeling quality information for each structural calibration area, analyzes the extracted information, evaluates the coverage distortion of each structural calibration area, and classifies each structural calibration area based on the evaluation results.

[0012] The strategy control module dynamically controls the update and replacement strategy of the old model region corresponding to each structural calibration region based on the classification results of each structural calibration region.

[0013] The visual recording module displays the dynamic control results in a 3D visualization interface and records the coverage distortion assessment results, classification results, and control results of the update and replacement strategy for each structural calibration area to support subsequent continuous optimization.

[0014] Preferably, in the region division module, when there is residual model occlusion during continuous scene updates, the spatial region with residual model occlusion is extracted. By establishing a regular mesh structure under the 3D modeling reference coordinates, the spatial region with residual model occlusion is uniformly divided in the horizontal and vertical directions according to the preset spatial resolution, forming several sub-regions that are consistent in spatial scale and do not overlap within the coordinate range, and each sub-region is marked as a structure calibration area.

[0015] Preferably, in the feature evaluation module, spatial modeling quality information of each structural calibration area is extracted and preprocessed after extraction; spatial structure alignment information and texture restoration consistency information are extracted from the preprocessed spatial modeling quality information of each structural calibration area, and analyzed after extraction to generate geometric consistency coefficient and texture restoration index of each structural calibration area respectively; based on the generated geometric consistency coefficient and texture restoration index of each structural calibration area, a coverage distortion index of each structural calibration area is generated; a pre-set coverage distortion index threshold range is determined, and after determination, it is compared with the generated coverage distortion index of each structural calibration area; the coverage distortion degree of each structural calibration area is evaluated according to the comparison result, and each structural calibration area is classified according to the evaluation result.

[0016] Preferably, the logic for obtaining the geometric consistency coefficients of each structural calibration region is as follows:

[0017] Spatial structure alignment information is extracted from the spatial modeling quality information of each preprocessed structural calibration area. Specifically, this includes the current average point cloud density value, the historical average point cloud density value, and the Hausdorff distance between the current boundary point set and the historical boundary point set in each structural calibration area, and these values ​​are respectively calibrated as follows: , and , Indicates the first The current average density value of the point cloud in each structural calibration area. Indicates the first The historical point cloud average density value in each structural calibration area Indicates the first Hausdorff distance values ​​between the current set of boundary points and the historical set of boundary points in each structural calibration region. , It is a positive integer;

[0018] The geometric consistency coefficient of each structural calibration zone is calculated using the following formula:

[0019]

[0020] In the formula, For the first The geometric consistency coefficient of each structural calibration region.

[0021] Preferably, the logic for obtaining the texture reproduction index of each structural calibration region is as follows:

[0022] Texture restoration consistency information is extracted from the spatial modeling quality information of each preprocessed structural calibration region. Specifically, this includes the mean pixel brightness of the current frame image, the mean pixel brightness of historical frame images, gray-level covariance, and gray-level gradient variance in each structural calibration region, and these are calibrated as follows: , , and , Indicates the first The average pixel brightness of the current frame image in each structural calibration region. Indicates the first The average pixel brightness of historical frame images in each structural calibration area. Indicates the first Gray-level covariance in each structural calibration region Indicates the first The gray-level gradient variance in each structural calibration region , It is a positive integer;

[0023] The texture reproduction index of each structural calibration region is calculated using the following formula:

[0024]

[0025] In the formula, For the first Texture reproduction index of each structural calibration area.

[0026] Preferably, based on the geometric consistency coefficients of each generated structural calibration region and texture reproduction index The coverage distortion index of each structural calibration region is generated by weighted summation. The specific calculation formula is as follows:

[0027]

[0028] In the formula, For the first Coverage distortion index of each structural calibration region and These are the geometric consistency coefficients for each structural calibration region. and texture reproduction index The non-zero weight coefficients, and .

[0029] Preferably, a pre-defined coverage distortion index threshold range is determined. And after determination, it is compared with the coverage distortion index of each generated structural calibration area. A comparison was performed, and the degree of coverage distortion of each structural calibration area was evaluated based on the comparison results. The structural calibration areas were then classified according to the evaluation results. The specific comparison analysis and classification are as follows:

[0030] like The coverage distortion level of this structure calibration area is high, and this structure calibration area is classified as a high coverage distortion region;

[0031] like The coverage distortion of the calibration area of ​​this structure is moderate, and the calibration area of ​​this structure is classified as a moderate coverage distortion region;

[0032] like The coverage distortion level of the calibration area of ​​this structure is low, and the calibration area of ​​this structure is classified as a low coverage distortion region.

[0033] Preferably, in the strategy control module, the update and replacement strategy of the old model region corresponding to each structural calibration region is dynamically controlled based on the classification results of each structural calibration region, specifically as follows:

[0034] For the structural calibration area classified as a high-coverage distortion area, stop the current replacement operation of the old model in the area, mark the area as a pending area, and then call the data supplementation process and perform multi-frame data overlay processing to obtain the replacement model data;

[0035] For the structural calibration area classified as a medium coverage distortion area, a fusion update operation is performed. By setting the fusion ratio between the old model data and the newly acquired data, the new data is superimposed onto the old model area in a weighted manner.

[0036] For structural calibration areas classified as low-coverage distortion areas, the old model data in the area is directly replaced with the newly acquired data, and the area is marked as an updated area.

[0037] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0038] 1. This invention overcomes the problem of existing systems failing to identify occlusion distortion risks and blindly replacing old models during continuous scene updates by constructing a multi-stage, hierarchical 3D dynamic modeling quality assessment and strategy control mechanism. Compared to traditional update logic based solely on time sequence or data update priority, this scheme introduces a "structural calibration area" as the smallest spatial evaluation unit for the first time. It quantitatively assesses the modeling quality of point cloud and image data using geometric consistency coefficients and texture restoration indices, and then generates an overlay distortion index through weighted synthesis, achieving precise quantification and dynamic perception of local area modeling quality. This mechanism not only improves the accuracy and relevance of model update decisions but also provides a structured quality foundation for subsequent modules such as visualization accuracy control and task scheduling feedback.

[0039] 2. Based on the coverage distortion index, the technical solution divides the structural calibration area into three coverage distortion levels—high, medium, and low—by constructing a threshold interval comparison mechanism. Differentiated update and replacement strategies are then implemented accordingly: replacement is paused and reconstructed in high-distortion areas; weighted fusion updates are performed in medium-distortion areas; and direct replacement is executed in low-distortion areas. This classification strategy effectively solves the local anomaly problems caused by unstable data quality, occlusion blurring, and sensor jitter during the modeling process. This enables the 3D modeling system to possess adaptive recognition and risk avoidance capabilities, significantly improving the structural stability and image coherence of the modeling output. It avoids geometric breaks and texture tears caused by erroneous updates, providing a more reliable data foundation for the operational space perception and remote monitoring of unmanned sweepers.

[0040] 3. This invention also includes a visual recording module that renders and records the evaluation results, classification results, and control strategies for each structural calibration area in real time, ensuring traceability and feedback capabilities throughout the entire evaluation and update process. By dynamically labeling the status of each area through a 3D interface, monitoring personnel can intuitively grasp the modeling status and the location of abnormal areas. Simultaneously, the recorded data supports subsequent model training optimization and adaptive parameter updates. Overall, this technical solution not only enhances the real-time intelligent control capabilities of dynamic modeling but also improves data transparency and optimization closed-loop capabilities during system operation, laying a solid foundation for building a highly reliable and high-fidelity unmanned sweeping operation visualization system. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0042] Figure 1 This is a schematic diagram of the modules of the real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to the present invention.

[0043] Figure 2 This is a system mind map of the real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers, based on the present invention. Detailed Implementation

[0044] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0045] This invention provides, for example Figure 1 and Figure 2 The real-time 3D dynamic visualization monitoring system for unmanned sweepers shown includes a modeling and monitoring module, a region division module, a feature evaluation module, a strategy control module, and a visual recording module.

[0046] The modeling and monitoring module uses real-time image and point cloud data generated by the unmanned sweeper during its operation to perform continuous modeling and updating of the 3D scene, and monitors the continuous modeling and updating process of the 3D scene in real time to identify whether there is model occlusion residue in the continuous scene update.

[0047] Real-time acquisition of image and point cloud data can be achieved through onboard sensor data acquisition software during the operation of the autonomous sweeper. The acquisition and scheduling module continuously calls the driver interfaces of onboard cameras and LiDAR sensors to acquire raw image frames and point cloud frames, which are then synchronously encoded using a timestamp mechanism. Subsequently, the data formatting component performs structural standardization on these frames and pushes them to a buffer queue, ensuring the data has temporal integrity and spatial continuity. Based on this, the system supplies the data stream to the subsequent modeling and update logic module in real time, ensuring that the images and point clouds always reflect the actual working environment of the vehicle.

[0048] Continuous modeling and updating of 3D scenes can be achieved through an incremental modeling workflow built in software. Upon arrival of each new frame of data, the newly acquired image and point cloud data are first converted to a unified coordinate system through pose alignment and coordinate transformation. Subsequently, data compression is performed through voxelization, and a modeling engine based on voxel meshes or TSDF structures is introduced to fuse the current frame with historical models. During the fusion process, only the spatial regions covered by new data are updated, avoiding full reconstruction and improving processing efficiency. The entire modeling process is triggered frame by frame, dynamically reconstructing the 3D model of the work area as the sweeper moves, achieving continuous, local, and real-time scene modeling.

[0049] While performing continuous modeling, the system analyzes the update process in real time through model quality monitoring logic and identifies residual occlusion. This identification process analyzes the regional data characteristics after each modeling update, including indicators such as point cloud density, edge closure, and image texture changes, to construct a local integrity judgment model. Simultaneously, by comparing the geometric differences between the current frame and historical models, it extracts spatial regions with potential occlusion distortion. When local information loss or structural instability is detected, regions with residual occlusion are marked, and an occlusion identification map is generated for subsequent region segmentation and strategy adjustment, forming a closed-loop judgment mechanism.

[0050] This entire process is adopted because the environment in which unmanned sweepers operate is complex and ever-changing. Factors such as occlusion, fog, and insufficient lighting can cause distortion or loss of image and point cloud information in some areas. If the model is updated directly without an effective recognition mechanism, it will lead to distortion of the 3D structure, misjudgment of obstacles or paths, and thus affect the reliability of remote scheduling and task execution. By monitoring the model update status in real time through software, identifying occluded and residual areas, and establishing independent control logic in problem areas, not only can the model quality be improved, but a high-quality data foundation is also laid for subsequent distortion assessment and strategy decision-making, thereby building a dynamic and reliable visual monitoring capability.

[0051] The region segmentation module extracts the spatial regions with residual model occlusion during continuous scene updates, and divides them evenly to generate several structural calibration regions.

[0052] In this embodiment, in the region division module, when there is model occlusion residue in continuous scene updates, the spatial region with model occlusion residue is extracted. By establishing a regular mesh structure under the 3D modeling reference coordinates, the spatial region with model occlusion residue is uniformly divided in the horizontal and vertical directions according to the preset spatial resolution, forming several sub-regions that are consistent in spatial scale and do not overlap in the coordinate range, and each sub-region is marked as a structure calibration area.

[0053] During continuous scene updates, when model occlusion remnants are identified, the spatial regions of these remnants can be extracted using software based on spatial consistency and data integrity rules. Specifically, after each modeling update, the system performs differential analysis on the overlapping areas of the new data and the historical model, calculating the point cloud density gradient, boundary coherence index, and image texture change rate of that region, and comparing these values ​​with a set occlusion threshold. If any index in a spatial unit falls below the confidence standard, that unit is determined to be an occlusion remnant unit. Subsequently, the system continuously detects all spatial units that meet the occlusion remnant conditions in the 3D modeling reference coordinates, and performs connectivity aggregation on these units in space to form closed or semi-closed occlusion remnant spatial regions, thus completing the extraction of the occlusion remnant regions. This process is entirely automated in the software by the data feature calculation and spatial logic aggregation algorithms used in the continuous modeling process, without requiring manual intervention.

[0054] After extracting the spatial regions with residual model occlusion, to facilitate subsequent quantitative assessment and strategy adjustment of regional coverage distortion, a regular mesh structure can be constructed in the 3D modeling reference coordinates using software, and the residual occlusion region can be uniformly divided. Specifically, the system first maps the residual occlusion region to the modeling coordinate system, using the region's boundary as the outer bounding box, and setting a uniform spatial resolution as the segmentation unit size. Then, the region is equally spaced along the horizontal (x-axis) and vertical (y-axis) directions, generating a set of cubic or cuboid sub-units that are completely consistent in spatial scale. The system uniformly registers the spatial number, boundary coordinates, and center point location information of these sub-units, ensuring that each sub-region does not overlap within the coordinate range. Each sub-unit serves as an independent structural calibration area for subsequent feature extraction and evaluation operations. This approach not only improves the granularity of local modeling analysis but also forms a standardized spatial index structure through the segmentation results, making different occluded regions comparable in different cleaning tasks, thus enhancing the versatility and controllability of modeling region processing. Meanwhile, this regular grid structure facilitates rapid localization, region matching, and strategy mapping in subsequent modules such as graphics rendering, algorithm evaluation, and multi-vehicle task allocation, serving as a fundamental step in achieving structured, visual modeling and control. The entire process is automatically completed by the software through coordinate transformation, grid generation, and spatial mapping functions, requiring no manual partitioning and exhibiting high consistency and algorithm compatibility.

[0055] The feature evaluation module extracts spatial modeling quality information for each structural calibration area, analyzes the extracted information, evaluates the coverage distortion of each structural calibration area, and classifies each structural calibration area based on the evaluation results.

[0056] In this implementation, the spatial modeling quality information of each structural calibration area is extracted in the feature evaluation module and preprocessed after extraction. Spatial structure alignment information and texture restoration consistency information are extracted from the preprocessed spatial modeling quality information of each structural calibration area and analyzed after extraction to generate geometric consistency coefficients and texture restoration indices for each structural calibration area. Based on the generated geometric consistency coefficients and texture restoration indices of each structural calibration area, a coverage distortion index for each structural calibration area is generated. A pre-set threshold range for the coverage distortion index is determined and compared with the generated coverage distortion index of each structural calibration area. The degree of coverage distortion of each structural calibration area is evaluated based on the comparison results, and each structural calibration area is classified according to the evaluation results.

[0057] In the feature evaluation module, spatial modeling quality information for each structural calibration area can be extracted using software-based image processing and point cloud analysis techniques. Specifically, after dividing the structural calibration areas, the system automatically traverses the 3D coordinate range of each calibration area and extracts data segments overlapping with that spatial region from the corresponding image frames and point cloud frames. For the image data, the grayscale matrix of the corresponding region is extracted by pixel coordinate mapping for subsequent statistical analysis of brightness mean, gradient changes, and texture structure. For the point cloud data, the set of spatial points within the region is obtained through coordinate filtering, and then the local point cloud density, point distribution range, and boundary point set are calculated. The extraction process relies on an intra-frame indexing system and spatial mapping logic, performing cross-modal data synchronous sampling in the modeling reference coordinate system to ensure spatial alignment and temporal consistency between the image and point cloud data, thereby completely constructing the spatial modeling quality information for each structural calibration area.

[0058] The purpose of preprocessing is to improve the accuracy and stability of subsequent parameter extraction and evaluation. Especially in real-world scenarios with image noise, uneven point cloud sampling, or lighting differences, the raw data often contains outliers, holes, or local anomalies, which can interfere with the judgment of modeling quality. The preprocessing process can be automated via software and mainly includes three steps: First, the extracted image grayscale data is normalized to eliminate the impact of brightness deviation on texture evaluation; second, outlier removal and voxel filtering are performed on the point cloud density data to reduce misjudgments caused by edge noise; finally, consistency compensation is performed on the boundary regions of the image and point cloud, and occluded or missing regions are interpolated or marked to ensure the structural closure of the evaluation region. The entire preprocessing workflow performs closure operations on a structural calibration region basis and generates a normalized dataset, providing a high-quality data foundation for subsequent extraction of spatial structure alignment information and texture restoration consistency information.

[0059] Extracting spatial structure alignment information and texture restoration consistency information from the preprocessed spatial modeling quality information of each structural calibration region can be achieved by using software-constructed multi-dimensional feature extraction logic to perform statistical analysis and spatial comparison operations on point cloud data and image data respectively. The extraction process for spatial structure alignment information includes: voxelizing and reconstructing the point clouds within the same structural calibration region in the current and historical frames; calculating the average density of the point clouds in this region under a unified reference coordinate system and comparing the differences; simultaneously extracting the current and historical boundary point sets and calculating their Hausdorff distance to characterize the boundary alignment error; this information ultimately consists of density change indicators and boundary offset indicators. The extraction process for texture restoration consistency information includes: performing pixel-level operations on the corresponding image blocks of the calibration region in the current and historical image frames at the image level; extracting three types of numerical features: mean brightness difference, gray-level covariance, and gray-level gradient variance, which are used to measure the region's illumination consistency, texture structure correlation, and edge complexity, respectively. The entire information extraction process is based on the inter-frame matching mechanism and spatial mapping matrix for alignment, ensuring that point cloud and image data are processed uniformly in the same spatial block. The extraction results are standardized into parameter-level values ​​using the structural calibration area as the processing unit, providing quantitative input for subsequent index calculation and coverage distortion assessment.

[0060] Determining the pre-defined coverage distortion index threshold range can be achieved through software methods based on statistical and cluster analysis of historical modeling sample data. Specifically, during the initial modeling training phase, the system first collects a large amount of spatial modeling quality information of structural calibration areas from multiple typical operational scenarios, and calculates the corresponding geometric consistency coefficient and texture restoration index for each calibration area, thereby generating a coverage distortion index. Subsequently, the system normalizes these coverage distortion index data and uses clustering algorithms (such as K-means or density clustering) to model and divide the natural distribution of distortion levels, identifying three representative data concentration areas, corresponding to high distortion, medium distortion, and low distortion areas, respectively. Based on this, the system automatically sets the coverage distortion index threshold range corresponding to each level by extracting the statistical boundary values ​​of each cluster interval, and stores it as an evaluation standard. In the formal operation phase, the system loads this threshold configuration and directly uses it to judge the distortion level of each structural calibration area without manual intervention, realizing the function of automatically generating threshold ranges based on data distribution, thereby ensuring that the classification criteria have statistical validity and adaptability.

[0061] In this embodiment, the logic for obtaining the geometric consistency coefficient of each structural calibration region is as follows:

[0062] Spatial structure alignment information is extracted from the spatial modeling quality information of each preprocessed structural calibration area. Specifically, this includes the current average point cloud density value, the historical average point cloud density value, and the Hausdorff distance between the current boundary point set and the historical boundary point set in each structural calibration area, and these values ​​are respectively calibrated as follows: , and , Indicates the first The current average density value of the point cloud in each structural calibration area. Indicates the first The historical point cloud average density value in each structural calibration area Indicates the first Hausdorff distance values ​​between the current set of boundary points and the historical set of boundary points in each structural calibration region. , It is a positive integer;

[0063] The current average point cloud density, historical average point cloud density, and Hausdorff distance between the current and historical boundary point sets in each structural calibration area can all be calculated and obtained in real time by the software during continuous modeling updates. The current average point cloud density is obtained by the system filtering the latest collected point cloud data subset within the spatial range of the structural calibration area in the 3D modeling reference coordinate system during each data update, and calculating the number of points per unit volume. The density is typically measured in points per cubic meter, forming the current average point cloud density for that area. The historical average point cloud density is obtained similarly, but its data source is a cached point cloud snapshot from the previous or a period of stable modeling. The system retrieves the historical point cloud data corresponding to the current structural calibration area location using a timestamp index and then performs the same density statistics method. Boundary point sets are generated by extracting the outermost boundary points from the point cloud subset, typically using α-shape or concave hull algorithms for boundary construction. The Hausdorff distance between the current boundary point set and the historical boundary point set is obtained by calculating the maximum and minimum distances between any point in the current boundary point set and the nearest point in the historical boundary point set. This is achieved in the software through bidirectional point set farthest neighbor distance calculation. All three types of data are based on point cloud clipping and spatial mapping within the spatial range of the structural calibration area. The data is directly extracted from continuous point cloud streams without the need for additional sensors or hardware support, enabling a real-time update, region alignment, and continuous quantization analysis process. These data, in a physical sense, correspond to local structural integrity (density) and spatial coincidence accuracy (boundary distance), respectively, and are core fundamental indicators for evaluating structural consistency and coverage distortion.

[0064] The geometric consistency coefficient of each structural calibration zone is calculated using the following formula:

[0065]

[0066] In the formula, For the first The geometric consistency coefficient of each structural calibration region.

[0067] The formula for calculating the geometric consistency coefficient is used to accurately and continuously measure the degree of geometric structural consistency of each structural calibration area during the continuous modeling and updating process by comparing multi-source spatial structural data. Specifically, the first term in the formula... Indicates the first The squared Hausdorff distance between the current boundary point set and the historical boundary point set of the structural calibration region reflects the maximum deviation of the spatial contour of the region. A larger value indicates a poor overlap between the current model and the historical model edges, and unstable boundaries, which are typical characteristics of structural distortion. Squaring it is to nonlinearly amplify the impact of boundary deviation on overall consistency, so that large deviation regions are strongly penalized by the exponential function. The second term This is the expression for the logarithmically compressed rate of change of point cloud density, where... and These represent the current and historical average point cloud densities, respectively. The density difference reflects the degree of spatial structure filling and geometric detail preservation. The closer the values ​​are, the more stable the modeling density and the stronger the geometric continuity of the region. Taking the logarithm of these values ​​helps to reduce random outliers in density changes while retaining their weight in the overall evaluation. The sum of these two values ​​serves as the negative exponent input to the exponential function, forming a monotonically decreasing function in the range (0,1]. This causes the geometric consistency coefficient in structurally stable regions to approach 1, while the coefficient in regions with large geometric disturbances decreases rapidly. This allows for precise quantification and dynamic grading of the spatial geometric quality of each structural calibration area at the numerical level. The entire calculation process, through the fusion of distance, density, and nonlinear functions, ensures that the evaluation results are robust, continuous, and distinctive, effectively supporting subsequent distortion assessment and strategy control logic.

[0068] No. Geometric consistency coefficient of each structural calibration region This is a quantitative characterization of the spatial structure fidelity of the region during the modeling process, and its value shows a significant negative correlation with the degree of cover distortion. Specifically, when When the value approaches 1, it indicates that during continuous modeling and updating, the current point cloud distribution and historical point cloud density of the calibration area change very little, and the boundary contour shows almost no offset. This indicates that the modeling and updating maintains a high degree of consistency and stability, and the spatial structure is not significantly disturbed. At this point, it can be judged that the coverage distortion of the area is extremely low; conversely, when... A significant decrease indicates a substantial difference between the current point cloud density and the historical density, along with a large deviation in boundary morphology. This reflects a severe geometric mismatch or structural incompleteness in the structure calibration area during the overlay of old and new modeling data, classifying it as a high-coverage distortion region. Therefore, The smaller the value, the more unstable the geometric modeling of the corresponding region, and the higher the degree of coverage distortion; the larger the value, the more continuous and consistent the modeling process, and the lower the risk of distortion. This indicator provides a precise geometric structural dimension basis for assessing the degree of coverage distortion and is a key supporting quantity in the multidimensional distortion assessment system.

[0069] In this embodiment, the logic for obtaining the texture reproduction index of each structural calibration area is as follows:

[0070] Texture restoration consistency information is extracted from the spatial modeling quality information of each preprocessed structural calibration region. Specifically, this includes the mean pixel brightness of the current frame image, the mean pixel brightness of historical frame images, gray-level covariance, and gray-level gradient variance in each structural calibration region, and these are calibrated as follows: , , and , Indicates the first The average pixel brightness of the current frame image in each structural calibration region. Indicates the first The average pixel brightness of historical frame images in each structural calibration area. Indicates the first Gray-level covariance in each structural calibration region Indicates the first The gray-level gradient variance in each structural calibration region , It is a positive integer;

[0071] The average pixel brightness of the current frame image, the average pixel brightness of historical frame images, the gray-level covariance, and the gray-level gradient variance in each structural calibration region can all be calculated and extracted in real time at the image processing level using software. First, after the structural calibration region is divided, the system locates the corresponding image block region in the current image frame based on the spatial range of the calibration region in the modeling coordinate system, and extracts the gray-level values ​​of all its pixels. The average pixel brightness of the current frame image is then obtained by statistically averaging the pixel gray-level values ​​within this region. Similarly, the average brightness of the historical frame image can be obtained by retrieving the image patch at the same coordinate position in the previous frame or a historical stable frame through the time indexing mechanism, and calculating the average of its pixel gray values. The difference between the two reflects the trend of image brightness change. Gray-level covariance. It is calculated based on the joint distribution relationship between corresponding pixel grayscale pairs in the same structural calibration area of ​​the current frame and historical frames. It reflects the similarity of the grayscale distribution pattern of the two image blocks; the higher the value, the more consistent the structural texture. Grayscale gradient variance The calculation requires performing gradient operators (such as Sobel or Scharr) on the current image patch to extract edge response values, statistically analyzing the gradient magnitudes of all pixels, and calculating their variance to reflect the texture complexity and edge change intensity within the region. When performing the above operations, the software can achieve cross-frame data extraction, corresponding pixel pairing, and feature quantity statistics through window convolution, matrix operations, and temporal buffering. All data originates from image content within the structural calibration area, exhibiting high spatial correlation and real-time update capabilities. This data forms the core foundation for quantitatively evaluating overlay distortion at the image level.

[0072] The texture reproduction index of each structural calibration region is calculated using the following formula:

[0073]

[0074] In the formula, For the first Texture reproduction index of each structural calibration area.

[0075] The formula for calculating the texture restoration index is designed to comprehensively reflect the restoration integrity and structural consistency of image content during continuous modeling and updating by fusing and evaluating multiple fundamental attributes at the image level of each structural calibration region. The first term of the formula... This indicates that the current frame image and the historical frame image are at the [number]th ... The logarithmic expression of the difference in mean pixel brightness within each structural calibration area is used to measure the consistency of lighting conditions. A larger difference indicates that the brightness in that area fluctuates drastically due to environmental changes, shadow occlusion, or sensor exposure, resulting in poor image content stability. Compressing the absolute increase of this value using a logarithmic function can suppress the dominance of extreme anomalies on the overall index while preserving its dominant influence. (Second term) This is an inverse expression of the gray-level covariance between the current and historical image patches. Gray-level covariance measures the linear consistency of gray-level distribution between two image patches; a larger value indicates a higher similarity in texture structure. Using its reciprocal form can more strongly negatively lower the exponent when the texture difference is greater. Adding 1 is to prevent division by zero errors and stabilize the function curve. (Third term) This is a logarithmic expression of the gray-level gradient variance, which reflects the severity of image edge changes. An excessively large variance indicates edge distortion, texture tearing, or noise interference in the region, and its adverse effect on the exponent needs to be smoothed using a logarithmic function. The three indicators physically cover the three main visual evaluation dimensions: brightness consistency, texture structure similarity, and texture complexity, and are independent and complementary to each other. Finally, they are weighted and combined as the negative exponent of the exponential function, forming a texture restoration exponent value in the (0,1) interval. When all image features remain good, the exponent approaches 1; if any dimension shows a severe anomaly, the exponent drops rapidly, thus achieving a highly sensitive discrimination of image quality distortion. This calculation logic not only ensures the interpretability and stability of the parameters but also ensures that the evaluation mechanism has good adaptability to complex visual scenes.

[0076] No. Texture reproduction index of each structural calibration area It is a quantitative indicator reflecting the quality of image modeling and restoration in this area, and its value shows a significant negative correlation with the degree of coverage distortion in this area. Specifically, when When the value approaches 1, it indicates that the pixel brightness distribution of the current frame image and the historical frame image in this structural calibration area is highly consistent (small difference in mean brightness), the texture detail structure is well matched (high gray-level covariance), and the edge texture is continuous and stable (moderate gray-level gradient variance). Overall, it shows that the image content is clearly restored, there is very little occlusion interference, and the degree of coverage distortion in this area is extremely low; while when A significant decrease indicates a clear image quality problem in the area, such as abrupt changes in illumination, image blurring, edge tearing, or texture breakage. This manifests as increased brightness difference, decreased gray-level covariance, and increased gradient variance, leading to an increase in the negative exponential term of the exponent and a rapid decrease in the overall exponent. This decrease is a concrete manifestation of the distortion of image overlay information during modeling and updating. Therefore, The lower the value, the worse the image coverage integrity of the calibration area of ​​the structure, the more unstable the modeling quality, and the higher the degree of coverage distortion. It is one of the key quantitative bases for identifying image-level distortion risks.

[0077] In this embodiment, the geometric consistency coefficients of each generated structural calibration region are used as the basis. and texture reproduction index The coverage distortion index of each structural calibration region is generated by weighted summation. The specific calculation formula is as follows:

[0078]

[0079] In the formula, For the first Coverage distortion index of each structural calibration region and These are the geometric consistency coefficients for each structural calibration region. and texture reproduction index The non-zero weight coefficients, and .

[0080] To achieve a comprehensive assessment of the coverage distortion in each structural calibration area, the system calculates the geometric consistency coefficient. With texture reproduction index Then, a weighted summation method is used to generate the coverage distortion index. The implementation method involves processing the two indicators zone by zone in the structural calibration zone dimension using software. and As a quantitative input for this region in two dimensions—geometric modeling quality and image reconstruction quality—two non-zero weight coefficients are introduced. and Adjust its fusion contribution. Weighting coefficients Used to reflect the proportion of geometric consistency in the overall distortion assessment. Used to reflect the degree of influence on image texture quality, both satisfy... The normalization constraint is used to ensure that Numerical stability and interpretability. During system initialization, and The system can be optimized by fitting historical evaluation data based on the relative contributions of geometric and texture factors to the accuracy of monitoring and judgment in typical scenarios. Alternatively, it can be configured based on task type (such as prioritizing structural integrity or image clarity) through machine learning or expert systems. The entire solution process is executed in the calculation loop of the structural calibration area, and the coverage distortion index is output in real time, providing a quantitative basis for subsequent distortion level classification and dynamic adjustment strategies.

[0081] In this embodiment, a pre-set coverage distortion index threshold range is determined. And after determination, it is compared with the coverage distortion index of each generated structural calibration area. A comparison was performed, and the degree of coverage distortion of each structural calibration area was evaluated based on the comparison results. The structural calibration areas were then classified according to the evaluation results. The specific comparison analysis and classification are as follows:

[0082] like The coverage distortion level of this structure calibration area is high, and this structure calibration area is classified as a high coverage distortion region;

[0083] This situation indicates a significant degradation in modeling quality during continuous modeling updates in this area, manifested as discontinuous geometric structures, significant boundary misalignment, or large-area blurring of image textures, lighting distortion, and occlusion / shadowing. These areas are typically caused by limited sensor data acquisition, frequent environmental disturbances, or overlapping modeling conflicts, resulting in current modeling results that fail to accurately reflect the structural characteristics of the actual operational scenario. Without intervention, this may manifest as noticeable breaks, distortions, or ghosting in the 3D visualization interface, severely interfering with the accuracy of monitoring personnel's assessment of cleaning paths, obstacle layouts, and status. It can also mislead trajectory tracing and dispatch command issuance. Therefore, the system should prioritize marking this area as a high-risk model region and suspend its replacement or employ enhanced modeling strategies for repair.

[0084] like The coverage distortion of the calibration area of ​​this structure is moderate, and the calibration area of ​​this structure is classified as a moderate coverage distortion region;

[0085] This indicates that while the region did not exhibit serious problems during modeling updates, it still suffers from a degree of quality inconsistency. This manifests as slight boundary offsets, minor texture damage, or minor mismatches between local brightness and structure. Such areas are typically caused by transient disturbances (such as short-term lighting changes or sensor noise) and are insufficient to damage the model. However, over time, these disturbances can accumulate and evolve into distortion hotspots. Therefore, the system can classify this region as a "needs attention" area of ​​moderate coverage distortion, displaying it with a neutral label in the visualization interface. It will be prioritized for automatic replacement when higher-quality data is updated, thus maintaining the long-term stability and visual clarity of the overall model structure.

[0086] like The coverage distortion level of the calibration area of ​​this structure is low, and the calibration area of ​​this structure is classified as a low coverage distortion region.

[0087] This indicates that the current modeling update results for this area perform well in both geometric structure and image texture, demonstrating high consistency and stability. Specifically, this is reflected in small changes in point cloud density, high boundary overlap, clear texture details, and balanced brightness distribution, indicating that the new data has successfully covered the historical modeling data and can stably reproduce the real-world operational scenario within the visualization system. The existence of such areas improves the overall structural integrity and readability of the model, helping monitoring personnel to intuitively understand the cleaning status and environmental conditions. The system can directly execute a modeling replacement strategy, updating the old model area with new data and marking it as a "high-confidence area" to reduce redundant calculations, free up system resources, and provide stable support for subsequent high-frequency update strategies.

[0088] The strategy control module dynamically controls the update and replacement strategy of the old model region corresponding to each structural calibration region based on the classification results of each structural calibration region.

[0089] In this embodiment, the strategy control module dynamically controls the update and replacement strategy of the old model region corresponding to each structural calibration region based on the classification results of each structural calibration region. Specifically:

[0090] For the structural calibration area classified as a high-coverage distortion area, stop the current replacement operation of the old model in the area, mark the area as a pending area, and then call the data supplementation process and perform multi-frame data overlay processing to obtain the replacement model data;

[0091] For structural calibration areas classified as high-coverage-distortion regions, to prevent new, low-quality data from directly replacing the existing model, the system controls the modeling update process for this region via software. First, upon detecting a coverage distortion index below the minimum threshold during the evaluation phase, the system immediately sets the corresponding old model region to a "frozen" state using a flag variable, preventing the current update operation from overwriting the original model data. Subsequently, the system adds this region to the "pending processing region queue" and triggers a re-acquisition process in the data scheduling engine. This means that in subsequent frames of images and point cloud data, new sets of sensing data are prioritized for this region and cached. To improve modeling quality, after acquisition, the system performs time alignment and brightness normalization on multiple frames of image data within this region, and performs spatial registration and filtering fusion on multiple frames of point cloud data. Finally, it generates a noise-suppressed, structurally continuous replacement model data through voxel stacking or weighted averaging. This processing method avoids erroneous replacements caused by single-frame data occlusion, noise, or local anomalies, effectively improving the model restoration quality in high-distortion regions and ensuring the continuity and reliability of the model structure in the 3D visualization interface. The entire process is automatically controlled by the software at the structural calibration zone level, requiring no external intervention, and has the ability to perform continuous evaluation, on-demand freezing, and intelligent reconstruction.

[0092] For the structural calibration area classified as a medium coverage distortion area, a fusion update operation is performed. By setting the fusion ratio between the old model data and the newly acquired data, the new data is superimposed onto the old model area in a weighted manner.

[0093] For the structural calibration area classified as having moderate coverage distortion, the system implements a fusion update operation through software control logic to gradually introduce new data to optimize model details while preserving the stability of the original modeling. Specifically, the system first acquires newly collected data units, including image patches and point cloud subsets, within the spatial range corresponding to the calibration area, and uses old model data from the historical modeling buffer as a reference. Subsequently, the system performs weighted fusion processing at both the image texture layer and the point cloud geometry layer according to a preset fusion ratio parameter for the moderate coverage distortion level. At the image level, the software performs a weighted average of the brightness value, grayscale gradient, and edge features at corresponding pixel locations to generate transitional image texture patches; at the point cloud level, it uses a weighted superposition of point coordinates and density information within the voxel grid to reconstruct the spatial morphology. Throughout the process, the fusion ratio controls the proportion of new and old data in the output results. For example, setting the proportion of new data to 30% and old data to 70% can mitigate the risk of damage to the overall model caused by occasional occlusion or blurring in the new data. The purpose of this strategy is to balance the continuity and evolution of the model, enabling the system to achieve robust transition modeling even when faced with slight structural inconsistencies. This avoids the system performance overhead caused by frequent rollbacks or repeated modeling, while improving the smoothness of visualization and the controllability of updates. The entire fusion operation is based on spatial consistency indexing and data alignment mechanisms within the structural calibration area, and is automatically completed by the software during the update scheduling phase, exhibiting adaptability and regional independence.

[0094] For structural calibration areas classified as low-coverage distortion areas, the old model data in the area is directly replaced with the newly acquired data, and the area is marked as an updated area.

[0095] For structural calibration areas classified as low-coverage distortion regions, the system directly executes model replacement operations through software decision logic, assuming that the newly acquired data has been evaluated as high-quality, low-error, and reliable. Specifically, the system first locates the precise range of the current structural calibration area in the 3D modeling coordinate system using spatial mapping relationships, and extracts the corresponding image patches and point cloud data subsets for that area in the new data update channel. Subsequently, the software automatically calls the update interface to remove the old modeling data for that area from the image texture cache and point cloud voxel structure, performs a data overwrite operation, and completely writes the new data into the corresponding image cache and spatial point set structure, achieving data-level replacement updates. Simultaneously, the system updates the structural calibration area in the status flag table to "updated area." This status is used in subsequent scheduling decisions to indicate that it will no longer participate in the replacement evaluation process of the current round, thereby reducing redundant computation and resource waste. Direct replacement is performed in low-coverage distortion areas because these areas have reached stable thresholds in both geometric structure and texture restoration. New data in these areas can be considered to have a realistic representation of the scene. Direct replacement not only simplifies the processing flow but also improves overall modeling efficiency and real-time visualization. Furthermore, the state marking mechanism avoids redundant processing, improves the system's throughput in high-frequency data update scenarios, and enables a low-intervention, high-trust adaptive update strategy. The entire operation is performed on a unit-by-unit basis based on the structural calibration area and is entirely driven automatically by the software scheduling process, requiring no external control.

[0096] The visual recording module displays the dynamic control results in a 3D visualization interface and records the coverage distortion assessment results, classification results, and control results of the update and replacement strategy for each structural calibration area to support subsequent continuous optimization.

[0097] To display and record the dynamic control results in a 3D visualization interface, this can be achieved by integrating a modeling engine and a data tagging management mechanism into the software. Specifically, the system first uses a spatial indexing function to bind each structural calibration area to its corresponding coverage distortion index, distortion level classification result, and control strategy result, based on the spatial location of the structural calibration area in the 3D modeling coordinate system. The 3D visualization interface uses a graphics rendering engine (such as OpenGL or Unity) to dynamically load the current modeling state. Colored overlay layers are rendered in the model view, unit by unit, using a preset color gradient (such as red-yellow-green) based on the coverage distortion level for intuitive labeling. An information window pops up when the mouse hovers over or selects an area, displaying the coverage distortion index value, classification level (such as high coverage distortion area), and the type of strategy executed (such as pause update, merge replacement, or direct overlay). This allows monitoring personnel to intelligently interpret and judge each area without leaving the model view.

[0098] Meanwhile, to support continuous system optimization and scheduling backtracking, the system also constructs a structured record table in software to record the evaluation indicators, classification labels, and control instructions for each structural calibration area in each frame or each round of modeling updates in a time-series manner. Specifically, after each evaluation and strategy execution, the system automatically uses the unique spatial number of the calibration area as the primary key index, along with key information such as the coverage distortion index, classification results, and strategy type for that round, and writes it to a database or high-performance log cache, forming a time-series data snapshot. These records will be used for subsequent modeling behavior optimization analysis, strategy feedback adjustment, and statistical attribution processing of abnormal areas, further enhancing the system's adaptive and intelligent evolution capabilities. Through this approach, the software not only effectively expresses visual information but also constructs a highly structured modeling control closed loop, providing strong support for the reliable operation of the system in high-frequency, dynamically changing scenarios.

[0099] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0100] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0101] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0102] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0104] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0105] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers, characterized in that, It includes a modeling and monitoring module, a region division module, a feature evaluation module, a strategy control module, and a visual recording module; The modeling and monitoring module uses real-time image and point cloud data generated by the unmanned sweeper during its operation to perform continuous modeling and updating of the 3D scene, and monitors the continuous modeling and updating process of the 3D scene in real time to identify whether there is model occlusion residue in the continuous scene update. The region segmentation module extracts the spatial regions with residual model occlusion during continuous scene updates, and divides them evenly to generate several structural calibration regions. The feature evaluation module extracts spatial modeling quality information of each structural calibration area and preprocesses it after extraction. From the preprocessed spatial modeling quality information of each structural calibration area, spatial structure alignment information and texture restoration consistency information are extracted and analyzed. Geometric consistency coefficient and texture restoration index of each structural calibration area are generated respectively. Based on the generated geometric consistency coefficient and texture restoration index of each structural calibration area, the coverage distortion of each structural calibration area is evaluated, and each structural calibration area is classified according to the evaluation results. The logic for obtaining the geometric consistency coefficients of each structural calibration region is as follows: Spatial structure alignment information is extracted from the spatial modeling quality information of each preprocessed structural calibration area. Specifically, this includes the current average point cloud density value, the historical average point cloud density value, and the Hausdorff distance between the current boundary point set and the historical boundary point set in each structural calibration area, and these values ​​are respectively calibrated as follows: , and , Indicates the first The current average density value of the point cloud in each structural calibration area. Indicates the first The historical point cloud average density value in each structural calibration area Indicates the first Hausdorff distance values ​​between the current set of boundary points and the historical set of boundary points in each structural calibration region. , It is a positive integer; The geometric consistency coefficient of each structural calibration zone is calculated using the following formula: In the formula, is the th Geometric consistency coefficient of each structural calibration region; The various knots The logic for obtaining the texture reproduction index of the calibration region is as follows: Texture restoration consistency information is extracted from the spatial modeling quality information of each preprocessed structural calibration region. Specifically, this includes the mean pixel brightness of the current frame image, the mean pixel brightness of historical frame images, gray-level covariance, and gray-level gradient variance in each structural calibration region, and these are calibrated as follows: , , and , Indicates the first The average pixel brightness of the current frame image in each structural calibration region. Indicates the first The average pixel brightness of historical frame images in each structural calibration area. Indicates the first Gray-level covariance in each structural calibration region Indicates the first The gray-level gradient variance in each structural calibration region , It is a positive integer; The texture reproduction index of each structural calibration region is calculated using the following formula: In the formula, For the first Texture reproduction index of each structural calibration area; The strategy control module dynamically controls the update and replacement strategy of the old model region corresponding to each structural calibration region based on the classification results of each structural calibration region. The visual recording module displays the dynamic control results in a 3D visualization interface and records the coverage distortion assessment results, classification results, and control results of the update and replacement strategy for each structural calibration area to support subsequent continuous optimization.

2. The real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to claim 1, characterized in that, In the region segmentation module, when there are residual model occlusions during continuous scene updates, the spatial regions with residual model occlusions are extracted. By establishing a regular mesh structure under the 3D modeling reference coordinates, the spatial regions with residual model occlusions are uniformly divided in the horizontal and vertical directions according to the preset spatial resolution, forming several sub-regions that are consistent in spatial scale and do not overlap within the coordinate range. Each sub-region is then labeled as a structure calibration area.

3. The real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to claim 2, characterized in that, In the feature evaluation module, spatial modeling quality information of each structural calibration area is extracted and preprocessed. Spatial structure alignment information and texture restoration consistency information are extracted from the preprocessed spatial modeling quality information of each structural calibration area and analyzed after extraction to generate geometric consistency coefficients and texture restoration indices for each structural calibration area. Based on the generated geometric consistency coefficients and texture restoration indices of each structural calibration area, a coverage distortion index for each structural calibration area is generated. A pre-set threshold range for the coverage distortion index is determined and compared with the generated coverage distortion index of each structural calibration area. The degree of coverage distortion of each structural calibration area is evaluated based on the comparison results, and each structural calibration area is classified according to the evaluation results.

4. The real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to claim 3, characterized in that, Based on the geometric consistency coefficients of each generated structural calibration region and texture reproduction index The coverage distortion index of each structural calibration region is generated by weighted summation. The specific calculation formula is as follows: In the formula, For the first Coverage distortion index of each structural calibration region and These are the geometric consistency coefficients for each structural calibration region. and texture reproduction index The non-zero weight coefficients, and .

5. The real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to claim 4, characterized in that, Determine the pre-set coverage distortion index threshold range And after determination, it is compared with the coverage distortion index of each generated structural calibration area. A comparison was performed, and the degree of coverage distortion of each structural calibration area was evaluated based on the comparison results. The structural calibration areas were then classified according to the evaluation results. The specific comparison analysis and classification are as follows: like The coverage distortion level of this structure calibration area is high, and this structure calibration area is classified as a high coverage distortion region; like The coverage distortion of the calibration area of ​​this structure is moderate, and the calibration area of ​​this structure is classified as a moderate coverage distortion region; like The coverage distortion level of the calibration area of ​​this structure is low, and the calibration area of ​​this structure is classified as a low coverage distortion region.

6. The real-time three-dimensional dynamic visualization monitoring system for unmanned sweepers according to claim 5, characterized in that, In the strategy control module, the update and replacement strategy for the old model region corresponding to each structural calibration region is dynamically controlled based on the classification results of each structural calibration region. Specifically: For the structural calibration area classified as a high-coverage distortion area, stop the current replacement operation of the old model in the area, mark the area as a pending area, and then call the data supplementation process and perform multi-frame data overlay processing to obtain the replacement model data; For the structural calibration area classified as a medium coverage distortion area, a fusion update operation is performed. By setting the fusion ratio between the old model data and the newly acquired data, the new data is superimposed onto the old model area in a weighted manner. For structural calibration areas classified as low-coverage distortion areas, the old model data in the area is directly replaced with the newly acquired data, and the area is marked as an updated area.

Images