Urban dynamic target filtering system and method based on historical prior voxels

Abstract

The present application relates to the technical field of data processing, and is a city dynamic target filtering system and method based on historical prior voxels, specifically comprising: acquiring real-time point cloud feature data and adjacent load vibration spectrum features of a region to be inspected; preprocessing the above data, and continuing to analyze the difficulty of adhesion decoupling of the current region; at the same time, analyzing the stability of the city municipal asset structure; constructing a dynamic filtering efficiency evaluation model to evaluate the comprehensive efficiency of the filtering action; repeatedly acquiring filtering comprehensive efficiency evaluation values under all candidate filtering modes, screening the optimal parameters as a basic execution strategy, and restoring the city municipal asset structure under the basic execution strategy. The present application solves the problem of excessive resection in the filtering process of dynamic targets in the prior art, which leads to missed detection or false positives in the semantic recognition link due to incomplete features.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and is a system and method for filtering urban dynamic targets based on historical prior voxels. Background Technology

[0002] With the deepening of refined management in smart cities, the use of 3D LiDAR for the inspection of urban municipal facilities has become mainstream. Dynamic target filtering based on historical prior voxel libraries is a core step in obtaining high-precision static background maps. Existing technologies face the problem of scene adaptability. Temporary facilities such as water-filled barriers widely used in urban roads, although recorded as statically occupied in historical prior libraries, are easily affected by strong ground vibrations and lateral airflows caused by frequent passage of heavy-duty buses or dump trucks in the real and complex urban environment due to their non-absolutely rigid physical characteristics. This results in centimeter-level micro-drifts or physical tilts in the barriers, producing offset features in the real-time point cloud that do not conform to the prior model. At the same time, due to the extremely compact space on urban roads, dynamic targets such as pedestrians or non-motorized vehicles usually move close to the edge of the barriers. Under the condition of limited radial distance and angular resolution of the sensor, the newly added background point cloud caused by these displacements and the point cloud of the dynamic targets exhibit extremely high neighborhood coupling in 3D space. This makes it impossible for traditional spatial clustering algorithms or voxel dilation filtering logic to achieve effective decoupling at the topological level, resulting in severe topological adhesion. In this situation, when existing inspection systems perform dynamic target filtering, they often fail to determine boundaries, inadvertently removing the edge geometric features of the construction site while eliminating pedestrian anthropomorphic features, resulting in background erosion and over-cutting. This damage to background integrity not only leads to misjudgments of violations such as exceeding construction area limits, but also poses a deeper problem: because the construction site shares spatial extension with municipal facilities such as manhole covers and storm drain grates below, excessive cutting during the filtering process directly destroys the edge topological features of these critical assets, causing missed detections or false alarms in the semantic recognition stage due to feature incompleteness. Summary of the Invention

[0003] The technical problem to be solved by the present invention is that the excessive removal in the process of filtering dynamic targets in the prior art leads to missed detections or false alarms in the semantic recognition stage due to feature incompleteness. The present invention proposes a dynamic target filtering system and method for cities based on historical prior voxels.

[0004] To achieve the above objectives, the technical solution of the urban dynamic target filtering method based on historical prior voxels of the present invention includes the following steps:

[0005] S1: Obtain real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time obtain the vibration spectrum features of adjacent loads and the topological features of urban municipal assets.

[0006] S2: Based on the preset parameters of the current candidate filtering mode, the data obtained in S1 is preprocessed, and the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features are imported into the background topological adhesion decoupling model to analyze the adhesion decoupling difficulty of the current region.

[0007] S3: Import the topological features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model to analyze the stability of the urban municipal asset structure under the filtering action.

[0008] S4: Construct a dynamic filtering efficiency evaluation model, import the results of adhesion decoupling difficulty analysis and the results of urban municipal asset structure stability analysis into the dynamic filtering efficiency evaluation model, and evaluate the comprehensive efficiency of the filtering action.

[0009] S5: Repeat steps S1-S4 to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

[0010] Preferably, step S2 includes:

[0011] S21: Preprocess the data obtained in S1 according to the preset parameters of the current candidate filtering mode, specifically including:

[0012] Based on a preset echo intensity threshold, the real-time point cloud feature data is initially segmented to generate dynamic target candidate masks.

[0013] Based on a preset Euclidean clustering neighborhood radius, spatial clustering is performed on the point cloud in the candidate mask of the dynamic target to determine the precise boundary between the dynamic target and the background, and the number of topological overlap points between the dynamic target and the drifting background is quantified. ;

[0014] The correction is performed by multiplying the preset compensation efficiency coefficient with the vibration spectrum characteristics of the adjacent load by frequency points, and the corrected background deformation displacement is obtained by frequency domain integral transformation method.

[0015] S22: Obtain the echo intensity characteristic spatial consistency fluctuation coefficient based on echo intensity spatial distribution data under the current environment. Simultaneously, the deviation coefficients of the voxel spatiotemporal creep constraint field were obtained based on environmental vibration frequency spectrum data. ;

[0016] S23: Analyze the difficulty of decoupling adhesion in the current region. The formula for calculating the difficulty of decoupling adhesion is:

[0017] ;

[0018] AD represents the difficulty of decoupling adhesion. The echo intensity characteristic spatial uniformity fluctuation coefficient, represents the deviation coefficient of the voxel spatiotemporal creep constraint field. This represents the number of topological overlap points between the dynamic target and the drifting background. This represents the total number of points in the local critical domain.

[0019] Preferably, in step S23, the construction process of the background topological adhesion decoupling model includes the following specific steps:

[0020] First, based on the spatial distribution data of echo intensity, the complexity of identifying material fluctuations in the current region is analyzed, and the spatial consistency fluctuation coefficient of echo intensity characteristics is obtained. ;

[0021] Then, based on environmental vibration frequency spectrum data, the vibration coupling deformation of the semi-static construction enclosure was analyzed to obtain the deviation coefficient of the voxel spatiotemporal creep constraint field. .

[0022] Preferably, in step S3, the stability of the urban municipal asset structure under the filtering action is analyzed, including:

[0023] S31: Extract the prior feature vectors of municipal facilities involved in the current filtering action and the real-time background occlusion mask;

[0024] S32: Based on the prior feature vectors of municipal facilities and real-time background occlusion mask, a stability analysis model for the protection of urban municipal assets is constructed to analyze the structural damage risk of key municipal facilities and obtain the stability SW of urban municipal asset structures.

[0025] Preferably, in step S32, the process of constructing the urban municipal asset characteristic protection stability analysis model includes:

[0026] First, based on the real-time point cloud feature data and historical prior voxel distribution data under the current inspection trajectory, the degree of background topological occlusion interference caused by the dynamic target filtering action is analyzed, and the analysis results of the degree of topological interference caused by the filtering action to the topology of urban municipal assets are obtained.

[0027] Then, based on the current filter mask boundary and the topological characteristics of urban municipal assets, the degree of geometric feature loss of the mistakenly damaged urban municipal assets is analyzed, and the analysis results of the degree of loss of the edge geometric features of urban municipal assets are obtained.

[0028] Preferably, in step S4, constructing a dynamic filtering efficiency evaluation model includes: extracting the current adhesion decoupling difficulty analysis results and the stability analysis results of the urban municipal asset structure obtained from the analysis, and evaluating the overall filtering efficiency;

[0029] The formula for evaluating the overall filtering effectiveness is as follows: ; This is the overall filtration efficiency evaluation value under the current filtration parameters.

[0030] Preferably, step S5 includes the following specific steps:

[0031] S51: Traverse the set of voxels affected by vibration and obtain the overall filtering performance evaluation value of all candidate filtering modes. Select the set of parameters with the highest overall filtering performance evaluation value as the basic execution strategy.

[0032] S52: Under the basic execution strategy, activate the topology self-healing module under the shadow mask, automatically compare the missing background surfaces before and after the dynamic target mask under the same viewpoint, and if it is found that the missing surfaces belong to the part of the feature vector of the urban municipal asset topology structure, then use historical prior voxels to perform forced closure and completion to restore the urban municipal asset structure.

[0033] In addition, the urban dynamic target filtering system based on historical prior voxels of the present invention includes the following modules:

[0034] The module includes a data acquisition module, an adhesion and decoupling analysis module, an urban municipal asset stability analysis module, a filtering efficiency evaluation module, and a topology self-healing regulation module.

[0035] The data acquisition module is used to acquire real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time acquire adjacent load vibration spectrum features and urban municipal asset topology features.

[0036] The adhesion and decoupling analysis module preprocesses the data acquired by the data acquisition module according to the preset parameters of the current candidate filtering mode, and imports the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features into the background topology adhesion and decoupling model to analyze the adhesion and decoupling difficulty of the current region.

[0037] The asset stability analysis module is used to import the topological structure features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model, and to analyze the stability of the urban municipal asset structure under the filtering action.

[0038] The filtering efficiency evaluation module is used to construct a dynamic filtering efficiency evaluation model. It imports the results of the adhesion decoupling difficulty analysis and the results of the stability analysis of the urban municipal asset structure into the dynamic filtering efficiency evaluation model to evaluate the comprehensive efficiency of the filtering action.

[0039] The topology self-healing control module is used to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

[0040] A storage medium storing instructions that, when read by a computer, cause the computer to execute the urban dynamic target filtering method based on historical prior voxels.

[0041] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the above-described method for filtering urban dynamic targets based on historical prior voxels.

[0042] Compared with existing technologies, the technical effects of this invention are as follows: Firstly, in terms of both the accuracy of topological adhesion decoupling in extremely compact urban spaces and the integrity protection of municipal asset edge structures, this invention significantly improves the ability to separate dynamic targets from offset backgrounds. By introducing echo intensity feature spatial consistency verification, it successfully solves the technical difficulty of existing algorithms that easily misjudge newly added point clouds generated by drift and closely moving dynamic targets as the same connected domain when faced with centimeter-level micro-displacements of fences caused by heavy-duty vehicle vibrations. This achieves accurate separation of pedestrians from offset backgrounds at the material property level, eliminating background over-cutting and structural erosion caused by filtering. Secondly, this invention greatly enhances the protection of key municipal asset topological features. While filtering out dynamic interference, it can automatically identify and restore the geometric shapes of facilities such as manhole covers and storm drain grates below and at the edges of fences using voxel spatiotemporal creep constraint fields and topological self-healing mechanisms. This effectively prevents false alarms and missed detections of assets due to incomplete background features, providing a high-fidelity static base map for subsequent refined inspections. Finally, this invention significantly reduces the risk of misjudgment in urban management decisions. The system can keenly detect extremely subtle construction coordinate deviations or violations of area limits, and can clearly distinguish the essential difference between physical facility drift and newly added illegal occupancy, effectively ensuring the safe inspection of urban infrastructure. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0044] Figure 1 This is a flowchart illustrating the urban dynamic target filtering method based on historical prior voxels of the present invention.

[0045] Figure 2 This is a schematic diagram of the urban dynamic target filtering system based on historical prior voxels according to the present invention. Detailed Implementation

[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0047] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0049] Example 1:

[0050] like Figure 1 As shown, the urban dynamic target filtering method based on historical prior voxels in this embodiment of the invention includes the following specific steps:

[0051] S1: Obtain real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time obtain the vibration spectrum features of adjacent loads and the topological features of urban municipal assets.

[0052] S2: Based on the preset parameters of the current candidate filtering mode, the data obtained in S1 is preprocessed, and the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features are imported into the background topological adhesion decoupling model to analyze the adhesion decoupling difficulty of the current region.

[0053] In step S2, the difficulty of decoupling adhesion in the current region is analyzed, including:

[0054] S21: Preprocess the data obtained in S1 according to the preset parameters of the current candidate filtering mode, specifically including:

[0055] Based on a preset echo intensity threshold, the real-time point cloud feature data is initially segmented to generate dynamic target candidate masks.

[0056] Based on a preset Euclidean clustering neighborhood radius, spatial clustering is performed on the point cloud in the candidate mask of the dynamic target to determine the precise boundary between the dynamic target and the background, and the number of topological overlap points between the dynamic target and the drifting background is quantified. ;

[0057] The correction is performed by multiplying the preset compensation efficiency coefficient with the vibration spectrum characteristics of the adjacent load by frequency points, and the corrected background deformation displacement is obtained by frequency domain integral transformation method.

[0058] S22: Obtain the echo intensity characteristic spatial consistency fluctuation coefficient based on echo intensity spatial distribution data under the current environment. Simultaneously, the deviation coefficients of the voxel spatiotemporal creep constraint field were obtained based on environmental vibration frequency spectrum data. ;

[0059] S23: Construct a background topological adhesion decoupling model, analyze the adhesion decoupling difficulty of the current region, and the formula for calculating the adhesion decoupling difficulty is as follows:

[0060] ;

[0061] AD represents the difficulty of decoupling adhesion. The echo intensity characteristic spatial consistency (EIH) fluctuation coefficient, The deviation coefficient of the voxel spatiotemporal creep constraint field (VCCF) is given. This represents the number of topological overlap points between the dynamic target and the drifting background. This represents the total number of points in the local critical domain.

[0062] It should be noted that the material scattering intensity represents the separability of spectral features. Using the historical prior reflection intensity distribution as a benchmark, it measures whether the current voxel deviates from the known background material. The higher the deviation, the greater the score fluctuation and the more difficult the decoupling. Spatiotemporal vibration creep represents the stability of the geometric dimension, reflecting the voxel drift caused by vehicle vibration or equipment deformation.

[0063] It should also be noted that, For the reason and The interference coupling term that constitutes the difficulty of decoupling adhesion is determined by... In or By calculating the partial derivatives and analyzing their signs, we can see that when any one of the interference sources is fixed, if the other interference source suddenly increases, the difficulty of decoupling will inevitably increase. When the adhesion remains unchanged, the difficulty of decoupling AD increases with... It increases with the increase of, and similarly, when When the adhesion remains unchanged, the difficulty of decoupling AD increases with... It increases with the increase of;

[0064] Furthermore, in this embodiment, it should also be noted that, considering and These are independent sources of interference, and in the actual implementation of the solution, there will be... or The situation where the interference level of one source increases while the interference level of another source decreases is equivalent to the following: when the interference level of one source decreases, even if the interference level of the other source increases, the difficulty of decoupling the adhesion is reduced compared to the initial state. This is because, according to practical engineering principles, adhesion decoupling can still be performed based on the dimension with less interference. Therefore, the difficulty of adhesion decoupling is actually reduced. For example:

[0065] If the material is complex (such as a reflective surface): the target can be separated by relying more on geometric features (shape, position);

[0066] If only geometric drift (such as vibration) exists: the target can be separated by relying more on material characteristics (reflection intensity);

[0067] If and only if the interference levels of both dimensions increase: features in both dimensions become unreliable, all reliable features are lost, and adhesion and separation become extremely difficult. This is precisely what this embodiment aims to achieve. One of the situations that warrants punishment.

[0068] In step S23, the construction process of the background topological adhesion decoupling model includes the following specific steps:

[0069] First, based on the spatial distribution data of echo intensity, the complexity of identifying material fluctuations in the current region is analyzed, and the spatial consistency fluctuation coefficient of echo intensity characteristics is obtained. ;

[0070] For example, in this embodiment, the formula for calculating the spatial consistency of echo intensity characteristics is: ;

[0071] Where n is the number of categories for classifying material strength characteristics. Let be the ratio of the number of feature points of type i to the total number of points. This represents the average real-time reflection intensity. The standard reflectance of the corresponding material. The standard deviation of historical reflection intensity is represented by this term. By introducing an information entropy model, the disorder of point cloud reflection intensity is quantified, and a normalization term is used to eliminate the magnitude interference from different material backgrounds, comprehensively covering the complex factors of point clouds under reflection patterns and real-time environmental interference.

[0072] Then, based on environmental vibration frequency spectrum data, the vibration coupling deformation of the semi-static construction enclosure was analyzed to obtain the deviation coefficient of the voxel spatiotemporal creep constraint field. .

[0073] For example, in this embodiment, the formula for calculating the deviation coefficient is:

[0074] ;

[0075] Where M represents the number of vibration monitoring cycles for the construction site enclosure. The background deformation displacement is the value of the m-th monitoring. This is the average displacement.

[0076] y is the extremum amplification factor. , These are the maximum background deformation displacement and the minimum deformation displacement, respectively. It should be noted that in this embodiment, the mean difference is eliminated by the coefficient of variation, and the hyperbolic tangent function is used to characterize the amplification effect of extreme value shifts (such as sudden severe shocks of heavy-duty vehicles, heavy-duty buses, or heavy-duty dump trucks).

[0077] S3: Import the topological features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model to analyze the stability of the urban municipal asset structure under the filtering action.

[0078] In step S3, the stability of the urban municipal asset structure under the filtering action is analyzed, including:

[0079] S31: Extract the prior feature vectors of municipal facilities involved in the current filtering action and the real-time background occlusion mask;

[0080] S32: Based on the prior feature vectors of municipal facilities and real-time background occlusion mask, a stability analysis model for the protection of urban municipal assets is constructed to analyze the structural damage risk of key municipal facilities and obtain the stability SW of urban municipal asset structures.

[0081] For example, in this embodiment, the formula for calculating the stability of the urban municipal asset structure includes: assessing the overall stability of the urban municipal asset structure based on the degree of topological interference generated by the filtering action and the degree of loss of the urban municipal assets; taking the negative of the sum of the degree of topological interference oc generated by the filtering action and the result of the loss analysis gc as the exponential term of the natural exponential function (i.e., exp()) to obtain the stability SW of the urban municipal asset structure.

[0082] It should be noted that the exponential function used in this embodiment can effectively identify the stability dispersion difference under different processing intensities. When the filtering action is light and the edge of the urban municipal assets is well maintained, the stability remains at a high level. When the filtering action of erroneous removal causes a significant loss of edge features of urban municipal assets (such as manhole covers), the stability decreases, reflecting that the current filtering strategy has the risk of background overcutting.

[0083] In step S32, the process of constructing the urban municipal asset characteristic protection stability analysis model includes:

[0084] First, based on the real-time point cloud feature data and historical prior voxel distribution data under the current inspection trajectory, the degree of background topological occlusion interference caused by the dynamic target filtering action is analyzed, and the analysis results of the degree of topological interference caused by the filtering action on the topology of urban municipal assets (i.e., background) are obtained.

[0085] For example, in this embodiment, a formula is provided for calculating the degree of topological interference caused by the filtering action to the topology of urban municipal assets (i.e., the background):

[0086] ;

[0087] Where oc represents the degree of topological interference generated by the filtering action; The standard deviation of the background point cloud feature matching frequency between frames under the current filtering mask; T is the total time it takes for the inspection equipment to pass through the target area; This is the sensor's rated scanning frequency; Let be the effective projected area of ​​the background remaining after the mask removes the dynamic target at time t; The standard projected area of ​​the background at this location in the historical prior data.

[0088] It should be noted that this embodiment comprehensively analyzes the impact of filter frequency fluctuations and background visibility holes on the degree of topological interference in the protection of urban municipal assets. Firstly, the feature-matched frequency standard deviation is used. This reflects the stability of the filtering process. When the algorithm frequently changes the mask boundaries between consecutive frames, The increase indicates that the filtering behavior severely interferes with the consistency of the background topology and disrupts the rhythm of the background point cloud.

[0089] Secondly, the time integral of the projected area deviation is introduced. This is used to measure the cumulative effect of background holes caused by the filtering mask, by dividing by the total duration T and the standard product. This embodiment quantifies the level of background visibility impairment. It adopts a product form to combine spatial graphic deviation with temporal processing frequency fluctuation, which can accurately identify interference situations such as pedestrians close to manhole covers, where excessive filtering causes severe voids or instantaneous flickering in the background urban municipal asset features.

[0090] Then, based on the current filter mask boundary and the topological characteristics of urban municipal assets, the degree of geometric feature loss of the mistakenly damaged urban municipal assets is analyzed, and the analysis results of the degree of loss of the edge geometric features of urban municipal assets are obtained.

[0091] For example, in this embodiment, the filtering loss is analyzed from three dimensions: vector missingness, structural continuity, and surface topological degradation, and a formula for calculating the degree of loss of geometric features is provided:

[0092] ;

[0093] Where gc represents the degree of geometric feature loss of urban municipal assets; H represents the total number of topological vector segments of urban municipal assets cut by the edge of the filter mask. The missing length of the h-th segment of the vector being sliced; This is the curvature weight coefficient of this segment in the historical prior library (reflecting the importance of the edge). The maximum allowed single topology vector missing length for the asset type of the city's municipal assets; The topological discontinuity (average surface roughness) of the current urban municipal assets after the edges have been filtered out. The maximum permissible structural discontinuity threshold for this type of urban municipal asset.

[0094] It should be noted that the segmented vector refers to the part of the vector trajectory that falls inside the filtering area after the original continuous topological evolution path of the municipal asset is truncated or pruned by the filtering boundary due to the overlap between the coordinate domain of the filtering mask and the skeleton vector domain of the background municipal asset during the execution of dynamic target filtering task. It essentially represents the geometric components that are incorrectly stripped from the background static features due to over-cutting during filtering. The length acquisition strategies include: using the Euclidean distance formula for geometrically linear assets and using integration for geometrically curved assets.

[0095] It should be noted that the lack of vectors at the edges of urban municipal assets is the core of the loss. This embodiment, based on geometric topology, uses the sum of squares of the missing length and curvature weight as the main source of loss, reflecting the devastating impact of missing key geometric features (such as the arc boundary of manhole covers) on the identification of urban municipal assets.

[0096] The maximum allowed single-item topological vector missing length is achieved by summing H segments and dividing by the asset type of the city's municipal assets. This achieves the normalization of geometric loss, enabling it to reflect the average relative loss under different urban municipal assets.

[0097] Furthermore, an exponential term is introduced. This reflects the amplification effect of topological discontinuities. When filtering actions cause jagged edges or severe damage to urban municipal assets, and the discontinuities... Approaching or exceeding the safety threshold At this time, the loss score will increase exponentially, reflecting the damage to the long-term stability of urban municipal assets caused by structural destruction. This model can scientifically quantify the degree of topological damage to urban municipal assets caused by overcutting.

[0098] S4: Construct a dynamic filtering efficiency evaluation model, import the results of adhesion decoupling difficulty analysis and the results of urban municipal asset structure stability analysis into the dynamic filtering efficiency evaluation model, and evaluate the comprehensive efficiency of the filtering action.

[0099] In step S4, a dynamic filtering efficiency evaluation model is constructed, including: extracting the adhesion decoupling difficulty analysis results and the stability analysis results of the urban municipal asset structure, and evaluating the overall filtering efficiency.

[0100] The formula for evaluating the overall filtering effectiveness is as follows: ; This is the overall filtration efficiency evaluation value under the filtration parameters.

[0101] It should be noted that this embodiment evaluates the overall performance by calculating the ratio of SW to AD and combining it with the hyperbolic tangent function. The calculation results are between 0 and 1. The larger the value, the better the stripping effect on dynamic targets under the premise of ensuring the integrity of the background urban municipal assets. In this embodiment, the hyperbolic tangent function is selected. By utilizing its nonlinear compression characteristics, the difference in its effectiveness under different environmental fluctuations is highlighted. This avoids the interference of a single extreme sample (such as a large-scale complex adhesion) on the determination of the overall filtering confidence. The optimal parameters for balancing clean filtering and protection of urban municipal assets are identified, so as to accurately balance the filtering accuracy while ensuring steady-state operation.

[0102] S5: Repeat steps S1-S4 to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

[0103] Step S5 includes the following specific steps:

[0104] S51: Traverse the set of voxels affected by vibration and obtain the overall filtering performance evaluation value of all candidate filtering modes. Select the set of parameters with the highest overall filtering performance evaluation value as the basic execution strategy.

[0105] For example, in this embodiment, multiple sets of filtering parameters for complex dynamic targets in the city are preset. Each set of parameters (i.e., candidate filtering modes) includes, but is not limited to, the echo intensity threshold, the Euclidean clustering neighborhood radius, and the compensation efficiency coefficient for ground vibration. The system traverses the voxel key domains affected by heavy traffic vibrations. For each set of candidate parameters, it substitutes them into the aforementioned steps S2 to S4 to perform difficulty analysis, stability assessment, and efficiency calculation. The set of parameters with the highest comprehensive filtering efficiency assessment value is selected as the filtering instruction for the current frame to ensure that while filtering out dynamic targets such as pedestrians and vehicles, the miscut rate of urban municipal asset features such as road surfaces and manhole covers is reduced.

[0106] It should be noted that the parameters of the candidate filtering mode directly affect the adhesion decoupling difficulty analysis in step S2. Specifically, the echo intensity threshold is used to binarize the point cloud according to the reflection intensity, thus affecting the separation of the dynamic target from the background, and the Euclidean clustering neighborhood radius is used to perform spatial clustering of the dynamic target, thus affecting the number of topological overlap points between the dynamic target and the drifting background. Ultimately, this affects the difficulty of decoupling adhesion; the compensation efficiency coefficient is used to correct the background deformation displacement data caused by ground vibration, thereby affecting the deviation coefficient of the voxel spatiotemporal creep constraint field.

[0107] S52: Under the basic execution strategy, activate the topology self-healing module under the shadow mask, automatically compare the missing background surfaces before and after the dynamic target mask under the same viewpoint, and if it is found that the missing surfaces belong to the part of the feature vector of the urban municipal asset topology structure, then use historical prior voxels to perform forced closure and completion to restore the urban municipal asset structure.

[0108] For example, this embodiment provides a strategy for restoring the structure of urban municipal assets. Specifically, the system first obtains the spatial bounding box of the dynamic target mask in the current coordinate system, compares it with the background missing surfaces generated by the filtering action within the bounding box, and determines whether there are topological breaks caused by occlusion by calculating the distribution density of the background point cloud in the vertical normal direction. Then, the coordinates of the identified background missing surfaces are used to perform real-time spatial indexing with the urban municipal asset topological structure feature vector obtained in step S1. In this embodiment, the KD-Tree algorithm is used for implementation. If the spatial coordinates of the missing surfaces are found to be in the historical database, the system will perform the indexing. Within the boundary of fixed urban municipal assets (such as the half-circle edge of a manhole cover, drainage pipe ports, etc.), it is determined that there is necessary topological loss in the area. Finally, the geometric vector features of the corresponding coordinate segments in the historical prior voxel library are read, and the RANSAC algorithm is used to align the historical voxel surface with the edge of the current remaining urban municipal assets in the optimal pose. The missing topological holes are forcibly closed and filled. It should be noted that in this embodiment, the reflection intensity and normal vector of the historical voxel are injected into the current hole, so as to restore the complete geometric shape of the damaged urban municipal assets while filtering out dynamic targets, and ensure the accuracy of subsequent automated damage detection.

[0109] Example 2:

[0110] like Figure 2 As shown in the figure, the urban dynamic target filtering system based on historical prior voxels according to an embodiment of the present invention includes the following modules:

[0111] The module includes a data acquisition module, an adhesion and decoupling analysis module, an urban municipal asset stability analysis module, a filtering efficiency evaluation module, and a topology self-healing regulation module.

[0112] The data acquisition module is used to acquire real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time acquire adjacent load vibration spectrum features and urban municipal asset topology features.

[0113] The adhesion and decoupling analysis module preprocesses the data acquired by the data acquisition module according to the preset parameters of the current candidate filtering mode, and imports the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features into the background topology adhesion and decoupling model to analyze the adhesion and decoupling difficulty of the current region.

[0114] The asset stability analysis module is used to import the topological structure features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model, and to analyze the stability of the urban municipal asset structure under the filtering action.

[0115] The filtering efficiency evaluation module is used to construct a dynamic filtering efficiency evaluation model. It imports the results of the adhesion decoupling difficulty analysis and the results of the stability analysis of the urban municipal asset structure into the dynamic filtering efficiency evaluation model to evaluate the comprehensive efficiency of the filtering action.

[0116] The topology self-healing control module is used to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

[0117] Example 3:

[0118] This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0119] The processor executes the aforementioned urban dynamic target filtering method based on historical prior voxels by calling the computer program stored in memory.

[0120] The electronic device can vary considerably depending on its configuration or performance. It may include one or more Central Processing Units (CPUs) and one or more memories, wherein the memory stores at least one computer program that is loaded and executed by the processor to implement the urban dynamic target filtering method based on historical prior voxels provided in the above-described method embodiments. The electronic device may also include other components for implementing its functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Details will not be elaborated upon in this embodiment.

[0121] Example 4:

[0122] This embodiment proposes a computer-readable storage medium on which an erasable and rewritable computer program is stored.

[0123] When the computer program runs on the computer device, it causes the computer device to execute the above-described method for filtering urban dynamic targets based on historical prior voxels.

[0124] For example, computer-readable storage media can be read-only memory (ROM), random access memory (RAM), compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage devices.

[0125] 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.

[0126] It should be understood that determining B based on A does not mean determining B solely based on A; it also means determining B based on A and / or other information.

[0127] 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. A 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 flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired network and / or wireless network. A 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. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).

[0128] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, 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 through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0129] 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.

[0130] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for filtering urban dynamic targets based on historical prior voxels, characterized in that, The method includes: S1: Obtain real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time obtain the vibration spectrum features of adjacent loads and the topological features of urban municipal assets. S2: Based on the preset parameters of the current candidate filtering mode, the data obtained in S1 is preprocessed, and the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features are imported into the background topological adhesion decoupling model to analyze the adhesion decoupling difficulty of the current region. The construction process of the background topological adhesion decoupling model includes the following specific steps: First, based on the spatial distribution data of echo intensity, the complexity of identifying material fluctuations in the current region is analyzed, and the spatial consistency fluctuation coefficient of echo intensity characteristics is obtained. ; Then, based on environmental vibration frequency spectrum data, the vibration coupling deformation of the semi-static construction enclosure was analyzed to obtain the deviation coefficient of the voxel spatiotemporal creep constraint field. ; S3: Import the topological features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model to analyze the stability of the urban municipal asset structure under the filtering action. S4: Construct a dynamic filtering efficiency evaluation model, import the results of adhesion decoupling difficulty analysis and the results of urban municipal asset structure stability analysis into the dynamic filtering efficiency evaluation model, and evaluate the comprehensive efficiency of the filtering action. S5: Repeat steps S1-S4 to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

2. The urban dynamic target filtering method based on historical prior voxels according to claim 1, characterized in that, Step S2 includes: S21: Preprocess the data obtained in S1 according to the preset parameters of the current candidate filtering mode, specifically including: Based on a preset echo intensity threshold, the real-time point cloud feature data is initially segmented to generate dynamic target candidate masks. Based on a preset Euclidean clustering neighborhood radius, spatial clustering is performed on the point cloud in the candidate mask of the dynamic target to determine the precise boundary between the dynamic target and the background, and the number of topological overlap points between the dynamic target and the drifting background is quantified. ; The correction is performed by multiplying the preset compensation efficiency coefficient with the vibration spectrum characteristics of the adjacent load by frequency points, and the corrected background deformation displacement is obtained by frequency domain integral transformation method. S22: Obtain the echo intensity characteristic spatial consistency fluctuation coefficient based on echo intensity spatial distribution data under the current environment. Simultaneously, the deviation coefficients of the voxel spatiotemporal creep constraint field were obtained based on environmental vibration frequency spectrum data. ; S23: Construct a background topological adhesion decoupling model, analyze the adhesion decoupling difficulty of the current region, and the formula for calculating the adhesion decoupling difficulty is as follows: ； AD represents the difficulty of decoupling adhesion. The echo intensity characteristic spatial uniformity fluctuation coefficient, represents the deviation coefficient of the voxel spatiotemporal creep constraint field. This represents the number of topological overlap points between the dynamic target and the drifting background. This represents the total number of points in the local critical domain.

3. The urban dynamic target filtering method based on historical prior voxels according to claim 2, characterized in that, In step S3, the stability of the urban municipal asset structure under the filtering action is analyzed, including: S31: Extract the prior feature vectors of municipal facilities involved in the current filtering action and the real-time background occlusion mask; S32: Based on the prior feature vectors of municipal facilities and real-time background occlusion mask, a stability analysis model for the protection of urban municipal assets is constructed to analyze the structural damage risk of key municipal facilities and obtain the stability SW of urban municipal asset structures.

4. The urban dynamic target filtering method based on historical prior voxels according to claim 3, characterized in that, In step S32, the process of constructing the urban municipal asset characteristic protection stability analysis model includes: First, based on the real-time point cloud feature data and historical prior voxel distribution data under the current inspection trajectory, the degree of background topological occlusion interference caused by the dynamic target filtering action is analyzed, and the analysis results of the degree of topological interference caused by the filtering action to the topology of urban municipal assets are obtained. Then, based on the current filter mask boundary and the topological characteristics of urban municipal assets, the degree of geometric feature loss of the mistakenly damaged urban municipal assets is analyzed, and the analysis results of the degree of loss of the edge geometric features of urban municipal assets are obtained.

5. The urban dynamic target filtering method based on historical prior voxels according to claim 4, characterized in that, In step S4, a dynamic filtering efficiency evaluation model is constructed, including: extracting the adhesion decoupling difficulty analysis results and the stability analysis results of the urban municipal asset structure, and evaluating the overall filtering efficiency. The formula for evaluating the overall filtering effectiveness is as follows: ; This is the overall filtration efficiency evaluation value under the current filtration parameters.

6. The urban dynamic target filtering method based on historical prior voxels according to claim 5, characterized in that, Step S5 includes the following specific steps: S51: Traverse the set of voxels affected by vibration and obtain the overall filtering performance evaluation value of all candidate filtering modes. Select the set of parameters with the highest overall filtering performance evaluation value as the basic execution strategy. S52: Under the basic execution strategy, activate the topology self-healing module under the shadow mask, automatically compare the background missing surfaces before and after the dynamic target candidate mask under the same view. If the missing surfaces are found to be part of the feature vector of the urban municipal asset topology structure, then use historical prior voxels to perform forced closure and completion to restore the urban municipal asset structure.

7. A city dynamic target filtering system based on historical prior voxels, used to implement the city dynamic target filtering method based on historical prior voxels as described in any one of claims 1-6, characterized in that, The system includes: The module includes a data acquisition module, an adhesion decoupling analysis module, an urban municipal asset stability analysis module, a filtering efficiency evaluation module, and a topology self-healing regulation module. The data acquisition module is used to acquire real-time point cloud feature data and historical prior voxel distribution data of the area to be inspected, and at the same time acquire adjacent load vibration spectrum features and urban municipal asset topology features. The adhesion and decoupling analysis module preprocesses the data acquired by the data acquisition module according to the preset parameters of the current candidate filtering mode, and imports the preprocessed real-time point cloud feature data and adjacent load vibration spectrum features into the background topology adhesion and decoupling model to analyze the adhesion and decoupling difficulty of the current region. The asset stability analysis module is used to import the topological structure features and real-time point cloud feature data of urban municipal assets into the urban municipal asset feature protection stability analysis model, and to analyze the stability of the urban municipal asset structure under the filtering action. The filtering efficiency evaluation module is used to construct a dynamic filtering efficiency evaluation model. It imports the results of the adhesion decoupling difficulty analysis and the results of the stability analysis of the urban municipal asset structure into the dynamic filtering efficiency evaluation model to evaluate the comprehensive efficiency of the filtering action. The topology self-healing control module is used to obtain the comprehensive filtering performance evaluation value of all candidate filtering modes, select the set of parameters with the highest comprehensive filtering performance evaluation value as the basic execution strategy, and activate the topology self-healing module under the shadow mask under the basic execution strategy to restore the urban municipal asset structure.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the urban dynamic target filtering method based on historical prior voxels as described in any one of claims 1-6.

9. An electronic device, characterized in that, include: Memory, used to store instructions; A processor is configured to execute the instructions, causing the device to perform operations that implement the urban dynamic target filtering method based on historical prior voxels as described in any one of claims 1-6.

Images