Method and apparatus for repairing map data

Abstract

This disclosure provides a method and apparatus for repairing map data, relating to the fields of geographic information data and autonomous driving processing technology. The method for repairing map data includes: determining at least one anomalous region from the map data, comprising anomalous point clouds, based on the terrain attribute data of each point cloud in the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point clouds; determining the anomalous type of each of the at least one anomalous region based on the geometric features of the point clouds in each of the at least one anomalous region and the terrain attribute data of each anomalous point cloud within the anomalous region; and repairing the anomalous region according to a repair strategy matching the anomalous type.

Description

Technical Field

[0001] This disclosure relates to the fields of geographic information data processing and autonomous driving technology, and more specifically, to a method and device for repairing map data. Background Technology

[0002] In various autonomous driving scenarios, terrain and environmental data are the core foundation for achieving precise operations and safe operation. For example, in open-pit mines, as important sites for mineral resource extraction, terrain data is the core basis for mine design, production scheduling, safety monitoring, and ecological restoration. Similarly, the scheduling and route planning of autonomous vehicles in logistics parks and ports, as well as the road condition perception and safety avoidance of autonomous passenger vehicles, all rely heavily on accurate terrain and environmental data. Terrain data can be acquired through methods such as LiDAR (Light Detection and Ranging), drone aerial photography, and total station measurements. However, due to the influence of complex environments, the acquired terrain data often exhibits anomalies such as missing data, noise, and distortion (e.g., anomalies caused by frequent terrain changes in mining environments, and anomalies caused by obstacles in mines, logistics facilities, and ports). Therefore, effective restoration of terrain data such as maps is necessary to meet the requirements of high-precision data applications.

[0003] However, the relevant terrain data restoration methods are difficult to adapt to complex terrain features and have problems such as reliance on a single data source, crude anomaly identification, and simplistic restoration logic. Summary of the Invention

[0004] In view of this, this disclosure provides a method and apparatus for repairing map data.

[0005] One aspect of this disclosure provides a method for repairing map data, comprising: determining at least one anomalous region comprising anomalous point clouds from the map data based on terrain attribute data of each point cloud in the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point clouds; determining anomaly type of each of the at least one anomalous region based on the point cloud geometric features of each of the at least one anomalous region and the terrain attribute data of each anomalous point cloud in the anomalous region; and repairing the anomalous region according to a repair strategy matching the anomaly type.

[0006] Another aspect of this disclosure provides a map data repair apparatus, comprising: an anomaly region determination module, configured to determine at least one anomaly region including anomaly point clouds from the map data based on the terrain attribute data of each point cloud in the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point clouds; an anomaly type determination module, configured to determine an anomaly type of at least one anomaly region based on the geometric features of the point clouds of each of the at least one anomaly regions and the terrain attribute data of each anomaly point cloud in the anomaly region; and a repair module, configured to repair the anomaly region according to a repair strategy matching the anomaly type.

[0007] Another aspect of this disclosure provides an electronic device comprising: one or more processors; and a memory for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the method as described above.

[0008] Another aspect of this disclosure provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the method described above.

[0009] Another aspect of this disclosure provides a computer program product including computer-executable instructions that, when executed, are used to implement the method as described above.

[0010] In the embodiments of this disclosure, based on the terrain attribute data of the point cloud, multiple spatially discrete anomalous point clouds can be identified in the map data. Based on these multiple anomalous point clouds, anomalous regions with clearly defined spatial extents can be determined, achieving precise localization of data anomalies from points to surfaces. This facilitates subsequent anomaly type identification and anomalous region repair. Furthermore, by combining the point cloud geometric features and terrain data of the overall point cloud within the anomalous region, the distribution of the point cloud and the continuity of the terrain can be comprehensively considered to accurately identify the anomaly type (missing anomaly, noise anomaly, distortion anomaly) of the anomalous region composed of each anomalous point cloud. Subsequently, differentiated repair strategies are formulated for each anomaly type, such as missing anomalies, noise anomalies, and distortion anomalies, to perform repair. This approach can effectively adapt to complex terrain changes (such as the "slope-step-platform" scenario in mining operations), thereby significantly improving the accuracy and reliability of the repaired map data. Attached Figure Description

[0011] The above and other objects, features and advantages of this disclosure will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0012] Figure 1 The illustration shows application scenarios of map data repair methods, apparatus, devices, media, and program products according to embodiments of this disclosure;

[0013] Figure 2 A flowchart of a method for repairing map data according to an embodiment of this disclosure is shown;

[0014] Figure 3 A flowchart illustrating the determination of multiple anomalous point clouds based on an improved isolated forest algorithm according to an embodiment of the present disclosure is shown.

[0015] Figure 4 A block diagram of a map data repair apparatus according to an embodiment of the present disclosure is shown; and

[0016] Figure 5 A block diagram of an electronic device suitable for implementing a method for repairing map data according to an embodiment of the present disclosure is shown. Detailed Implementation

[0017] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0018] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein, indicate the presence of said features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components. All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0019] In the embodiments of this disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security. In the embodiments of this disclosure, user authorization or consent has been obtained before acquiring or collecting user personal information.

[0020] This disclosure relates to the fields of geographic information data processing and autonomous driving technology, and can address the high-precision map construction needs in autonomous driving and autonomous driving application scenarios. Examples of such applications include, but are not limited to, smart mines, logistics parks, ports, airport shuttles, sanitation cleaning, public transportation connections, smart agriculture, last-mile delivery, and Robotaxi (autonomous taxis).

[0021] To facilitate understanding, some of the technical terms mentioned in this article are explained here:

[0022] The UTM (Universal Transverse Mercator Coordinate System) is a widely used geographic coordinate system that divides the Earth's surface into 60 regions, each 6 degrees wide, using the transverse Mercator projection. This coordinate system is suitable for regional mapmaking and navigation because it can convert the Earth's curved surface into a planar coordinate system, thus simplifying the calculation process.

[0023] The WGS84 coordinate system (World Geodetic System 1984) is the standard coordinate system of the Global Positioning System (GPS). It uses the Earth's center of mass as the origin and employs a reference ellipsoid to describe the Earth's shape. The WGS84 coordinate system provides a globally unified framework for representing location information, including longitude, latitude, and altitude.

[0024] Kriging interpolation, also known as spatial local interpolation, is a statistical spatial interpolation method. It is based on regionalized variables and uses a variogram to make linear, unbiased, and optimal estimates of unknown sample points. It is suitable for processing spatially correlated data.

[0025] According to one embodiment of this disclosure, taking a mining operation scenario in an unmanned driving scenario as an example, the complex environmental impact of a mine typically includes dust obstruction, large equipment obstruction, sensor offset caused by blasting vibration, extreme weather interference, and terrain structure offset caused by terrain operations. Under environmental influences, the collected terrain data often exhibits missing anomalies (caused by obstruction), noise anomalies (caused by reflection, dust misjudgment), and distortion anomalies (caused by sensor attitude offset, blasting, etc., resulting in instantaneous changes in terrain).

[0026] Existing terrain data restoration methods suffer from the following shortcomings: 1. Dependence on a single data source: Restoration is based solely on LiDAR point clouds or single image data, failing to fully utilize the complementarity of multi-source data, resulting in low restoration accuracy in complex occlusion scenarios; 2. Crude anomaly identification: Fixed threshold methods (such as elevation difference thresholds) are often used for a "one-size-fits-all" approach to identifying anomaly areas, neglecting the specific characteristics of mine terrain and causing mixed classification of anomaly types, leading to insufficient recognition accuracy (false positive rate often exceeding 15%). This method is ill-suited to the characteristics of open-pit mines with large slope variations and dramatic terrain undulations, easily resulting in "false positives" or "missed positives"; 3. Simplified restoration logic: All anomaly types are treated with uniform interpolation or smoothing, failing to consider the different causes of various anomalies, leading to a disconnect between the restored data and actual terrain features, especially significant errors in key areas such as slopes and steps. Therefore, there is an urgent need for an algorithm that can adapt to the complex terrain of open-pit mines, integrate the advantages of multi-source data, accurately identify and hierarchically restore anomalies, to meet the high-precision data application needs of mines.

[0027] This disclosure provides a method, apparatus, device, medium, and program product for repairing map data. The method for repairing map data includes: determining at least one anomalous region from the map data, including anomalous point clouds, based on the terrain attribute data of each point cloud in the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point clouds; determining the anomalous type of each of the at least one anomalous region based on the geometric features of the point clouds of each of the at least one anomalous region and the terrain attribute data of each anomalous point cloud in the anomalous region; and repairing the anomalous region according to a repair strategy matching the anomalous type.

[0028] Figure 1 The illustration shows application scenarios of map data repair methods, apparatus, devices, media, and program products according to embodiments of the present disclosure.

[0029] like Figure 1 As shown, the application scenario 100 according to an embodiment of this application may include a vehicle 101, multiple terminal devices such as a first terminal device 102, a second terminal device 103, and a third terminal device 104; the system architecture 100 also includes a network 105 and a server 106. The network 105 is used as a medium to provide a communication link between the vehicle 101, the first terminal device 102, the second terminal device 103, the third terminal device 104, and the server 106. The network 105 may include various connection types, such as wired and / or wireless communication links, etc.

[0030] Vehicle 101 interacts with server 106 via network 105 to receive or send messages, etc. For example, vehicle 101 can send real-time collected data, updated maps, updated map features, etc., to server 106 via network 105. Similarly, users can use a first terminal device or a third terminal device to interact with server 106 via network 105 to receive or send messages, etc.

[0031] Vehicle 101 can be any type of vehicle capable of performing various mining operations, such as a data collection vehicle, a transport vehicle, or a dumping vehicle. The embodiments of this disclosure do not limit the specific type of vehicle 101. Point cloud data can be collected by vehicle 101 or other devices and sent to the first terminal device to the third terminal device or the server 106.

[0032] The first terminal device 102, the second terminal device 103, or the third terminal device 104 can be various electronic devices with a display screen and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0033] Server 106 can be a server that provides various services, such as a backend management server that supports the websites browsed or information edited by vehicle 101 and the first to third terminal devices (for example only). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to each terminal device.

[0034] It should be noted that the map data repair method provided in this application embodiment can generally be executed by server 106 and / or vehicle 101, terminal devices 102-104. Accordingly, the map data repair device provided in this application embodiment can generally be set in server 106 and / or vehicle 101, terminal devices 102-104.

[0035] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0036] Figure 2 A flowchart of a method for repairing map data according to an embodiment of this disclosure is shown.

[0037] like Figure 2 As shown, the map data repair method includes operations S210~S230.

[0038] In operation S210, based on the terrain attribute data of each point cloud in the map data, at least one anomalous region including the anomalous point cloud is determined from the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point cloud.

[0039] In operation S220, based on the point cloud geometric features of at least one abnormal region and the terrain attribute data of each abnormal point cloud within the abnormal region, the anomaly type of at least one abnormal region is determined.

[0040] During operation S230, the abnormal area is repaired according to the repair strategy that matches the abnormality type.

[0041] According to one embodiment of this disclosure, the map data repair method is used to repair open-pit mine terrain data, which can meet the high-precision map construction needs of application scenarios such as unmanned driving and precise navigation of autonomous mining trucks in open-pit mines.

[0042] According to one embodiment of this disclosure, the map data is, for example, open-pit mine terrain data, which can be point cloud data in a UTM coordinate system. Terrain attribute data can be used to characterize the terrain at the location of the point cloud; the terrain may include, for example, steps, slopes, and platforms. The terrain attribute data of the point cloud includes, for example, the x-axis coordinate, y-axis coordinate, z-axis coordinate, slope value, and elevation value. Taking a mining operation scenario as an example, during the mining process, each horizontal layer forms a stepped structure in space; each step is called a step, and its height is called the step height. A slope refers to the inclined surface of the step facing the goaf; the degree of inclination of the surface at the location of each point cloud on this inclined surface relative to the horizontal plane is also called the slope value. A platform (or flat plate) refers to the space on the horizontal plane between the bottom and top lines of the slope of the step.

[0043] Elevation values ​​are usually not limited by terrain. The difference in coordinates of the point cloud at each location in the vertical direction relative to a certain reference plane (such as the WGS84 reference plane) is called the elevation value, which is usually directly represented by the Z-axis coordinate of the point cloud.

[0044] Anomaly point clouds can be understood as distorted data points in map data. These distorted data points may be caused by measurement errors or environmental interference, and may deviate significantly from mining design parameters and / or actual terrain trends. For example, anomaly point clouds may exhibit significant deviations in elevation values, slope values, or planar positions from mining design parameters, or may show a significant discrepancy with the spatial distribution trend of surrounding terrain points, potentially distorting the true form of key terrain features such as steps and slopes. Mining design parameters refer to the standard design parameters in mine design drawings, which can be understood as the overall planning drawings used for mine operations, describing various design terrain data of the mine, such as the design height of steps and the design slope of slopes.

[0045] An abnormal area can be understood as a region in map data that includes abnormal point clouds and / or other point clouds (normal point clouds). For example, various anomaly detection methods, such as difference comparison and anomaly point identification, can be used to determine abnormal point clouds based on the terrain attribute data of each point cloud, and the area covered by the abnormal point cloud is regarded as an abnormal area.

[0046] Point cloud geometric features are used to characterize the shape or location characteristics of point cloud spatial distribution, such as point cloud density and number of points.

[0047] Anomaly types can include, for example, missing anomalies, noise anomalies, and distortion anomalies. In one embodiment, the point cloud geometric features and terrain data features of the anomaly region can be combined to determine whether each anomaly region conforms to the characteristics of multiple anomaly types, thereby identifying the anomaly region that conforms to the characteristics of the corresponding anomaly type as the corresponding anomaly type.

[0048] Missing anomalies are typically caused by equipment occlusion or scanning blind spots. For example, if the point cloud density in an anomalous area is significantly lower than that in the surrounding normal area, and there are large areas of data gaps within the anomalous area, then the anomaly type of this area can be identified as a missing anomaly. Noise-related anomalies are typically caused by dust interference, equipment reflections, etc. For example, if the point cloud density in an anomalous area is comparable to that in the surrounding normal area, but the terrain attribute data (such as elevation values) of the point cloud in this anomalous area shows irregular and drastic jumps, and there are a large number of discrete points that deviate from the main distribution and are spatially discontinuous, then the anomaly type of this area can be identified as a noise-related anomaly. Distortion-related anomalies are typically caused by sensor calibration errors or instantaneous changes in terrain due to blasting vibrations. For example, if the point cloud distribution in an anomalous area is continuous, but the terrain surface it represents (such as the slope surface) has a systematic and continuous deviation from the design surface or the surrounding normal surface, then the anomaly type of this area can be identified as a distortion-related anomaly.

[0049] For the aforementioned anomaly types, different repair strategies can be formulated for missing anomalies, noise anomalies, and distortion anomalies, and the anomaly areas can be repaired according to the repair strategy matched to the anomaly type. Furthermore, based on the anomaly types, further refined repair strategies can be formulated for steps, slopes, and platforms based on the specific terrain scene characteristics of different terrains. Repair strategies include, but are not limited to: removing anomalous point clouds from the anomaly area, correcting anomalous point clouds, and generating new point clouds using other point clouds in the anomaly area.

[0050] In the embodiments of this disclosure, based on the terrain attribute data of the point cloud, multiple spatially discrete anomalous point clouds can be identified in the map data. Based on these multiple anomalous point clouds, anomalous regions with clearly defined spatial extents can be determined, achieving precise localization of data anomalies from points to surfaces. This facilitates subsequent anomaly type identification and anomalous region repair. Furthermore, by combining the point cloud geometric features and terrain data of the overall point cloud within the anomalous region, the distribution of the point cloud and the continuity of the terrain can be comprehensively considered to accurately identify the anomaly type (missing anomaly, noise anomaly, distortion anomaly) of the anomalous region composed of each anomalous point cloud. Subsequently, differentiated repair strategies are formulated for each anomaly type, such as missing anomalies, noise anomalies, and distortion anomalies, enabling repair. This approach effectively adapts to complex terrain changes (such as the "slope-step-platform" scenario in mining operations), thereby significantly improving the accuracy and reliability of the repaired map data.

[0051] Regarding the map data in operation S210, in one embodiment, the map data can be obtained based on the following operations: 1. Multi-source data acquisition and input: GIS (Geographic Information System) software acquires LiDAR point cloud data (containing X, Y, Z three-dimensional coordinates) and UAV orthophoto data (containing elevation data) of the target area of ​​the open-pit mine, and simultaneously imports the existing mining design drawings of the open-pit mine to extract mining design parameters such as bench design height and slope design gradient; 2. Coordinate normalization: The LiDAR point cloud data and UAV image data are uniformly converted to the UTM coordinate system to eliminate the deviation caused by the difference in coordinate systems; for example, the LiDAR point cloud data and UAV image data can be uniformly converted to the WGS84 coordinate system first, and then the corresponding UTM partition can be selected according to the area where the mine is located, and then the aforementioned data can be uniformly converted to the UTM coordinate system; 3. Data deduplication and fusion: Voxel filtering is applied to the LiDAR point cloud data. The data is deduplicated using a GridFilter, with voxel sizes set to 0.5m × 0.5m × 0.5m (to meet the resolution requirements of open-pit mine terrain). Elevation data from UAV imagery is extracted and fused with the deduplicated LiDAR point cloud data based on z-coordinate values ​​(elevation values) to construct a multi-source fusion dataset D (i.e., the map data mentioned above). During the fusion process, a weighted fusion strategy can be used for different data sources. For example, for elevation values ​​at the same spatial location (error ≤ 0.3m), the weight of the LiDAR data source can be set to 0.7, and the weight of the UAV data source to 0.3, and the fused elevation value can be calculated using a weighted average. In this embodiment, fusing multi-source terrain data avoids information insufficiency caused by occlusion from a single data source.

[0052] According to embodiments of this disclosure, determining at least one anomalous region including anomalous point clouds from map data based on terrain attribute data of each point cloud in map data includes: determining multiple anomalous point clouds from map data based on terrain attribute data of each point cloud in map data; and clustering the multiple anomalous point clouds, and determining the region in map data corresponding to the clustering result as at least one anomalous region.

[0053] In one embodiment, at least one anomalous region, including anomalous point clouds, can be identified from map data through a combination of preliminary screening and fine screening.

[0054] During the initial screening process, multiple anomalous point clouds can be identified from the map data based on the terrain attribute data of each point cloud. For example, anomalous point clouds deviating from design specifications can be identified by comparing the terrain attribute data of each point cloud with the mining design parameters in the mining design drawings. Alternatively, anomaly detection algorithms (such as statistical anomaly detection algorithms or machine learning-based anomaly detection algorithms) can be used to filter out anomalous point clouds from the map data based on their terrain attribute data.

[0055] As an example, dynamic or static thresholds can be set to classify point clouds whose elevation and / or slope values ​​deviate significantly from mining design parameters (such as bench design height and slope design gradient) as anomalous point clouds. As another example, based on the elevation and / or slope values ​​of point clouds in map data, unsupervised anomaly detection algorithms can be used to classify point clouds that significantly deviate from mining design parameters and / or surrounding terrain trends as anomalous point clouds.

[0056] In the fine-tuning process, to further classify the initially selected discrete anomalous point clouds into meaningful anomalous regions and filter out possible sporadic noise, multiple anomalous point clouds can be clustered, and the region in the map data corresponding to the clustering results can be identified as at least one anomalous region. Various clustering algorithms can be used. For example, based on the DBSCAN clustering algorithm, spatially adjacent anomalous point clouds can be grouped into the same cluster, while spatially isolated anomalous point clouds that cannot be assigned to any cluster are marked as noise. Furthermore, rule constraints (such as minimum number of point clouds within a cluster, elevation difference between point clouds within a cluster and normal point clouds) can be applied to the clustering results for filtering, excluding clusters containing noise.

[0057] According to embodiments of this disclosure, accurate identification of abnormal point clouds and abnormal regions in map data is a core prerequisite for effective map data repair. It can be achieved by first screening and then fine screening to identify abnormal point clouds and abnormal regions, thereby enabling accurate positioning and classification of abnormal point cloud data.

[0058] According to embodiments of this disclosure, determining multiple anomalous point clouds from map data based on the terrain attribute data of each point cloud in the map data includes: acquiring standard terrain attribute data corresponding to at least one terrain region, wherein the map data includes at least one terrain region, and each terrain region includes multiple point clouds; determining multiple anomalous point clouds from the map data based on the comparison results between the terrain attribute data of each point cloud and the standard terrain attribute data of the terrain region to which each point cloud belongs; preferably, determining the terrain fluctuation range using the standard terrain attribute data, and determining multiple anomalous point clouds from the map data based on the comparison results between the terrain attribute data of each point cloud and the terrain fluctuation range of the terrain region to which each point cloud belongs.

[0059] A terrain region can include stepped areas, slope areas, and platform areas, and each terrain region contains multiple point clouds. Typically, multiple point clouds within a terrain region belong to the same terrain. Standard terrain attribute data can be standard values ​​corresponding to the terrain attribute data. For example, standard terrain attribute data can include the designed height of steps and the designed slope of slopes extracted from mine design drawings.

[0060] In one example, for each point cloud in any terrain region, the elevation / slope values ​​in the point cloud's terrain attribute data can be directly compared with the corresponding step design height / slope design slope to directly identify anomalous point clouds. The comparison results can include the absolute values ​​of the differences between the elevation / slope values ​​and the corresponding step design height / slope design slope, such as elevation difference and slope difference. By ensuring that each point cloud meets the condition that its elevation difference is greater than the corresponding elevation difference threshold and / or its slope difference is greater than the corresponding slope difference threshold, the point cloud can be identified as an anomalous point cloud.

[0061] In a preferred embodiment, for the complex terrain scenario of "slope-step-platform" in open-pit mines, the constraints on anomalous point clouds can be relaxed. For example, this can be achieved by designing corresponding terrain fluctuation ranges (such as dynamic elevation fluctuation ranges and slope fluctuation ranges). For instance, the dynamic elevation fluctuation range can be set to [1.8H, 1.2H] based on the step design height H (e.g., relaxing the construction error by ±20%) to dynamically adapt to steps of different design dimensions while filtering out abnormal elevation jumps. If the elevation values ​​of any point cloud fall outside [1.8H, 1.2H], it is identified as an anomalous point cloud. This allows for a 1-2m deviation in the actual step height due to blasting, while also filtering out anomalous elevation values ​​above 50m caused by equipment obstruction. Alternatively, the slope fluctuation range can be set to [α-5°, α+5°] based on the slope design slope α (e.g., relaxing the error by ±5°) to adapt to the reasonable fluctuation range of open-pit mine slope slopes while avoiding misjudging normal slope slopes as distortions. For example, if a slope is designed to be 38°, its actual slope value may vary between 33° and 43° due to rainwater erosion. In this case, the slope value falls within the slope fluctuation range and is regarded as a normal point cloud.

[0062] In one specific embodiment, the traditional Isolation Forest algorithm can be used to filter out anomalous point clouds based on terrain attribute data. Isolation Forest is a machine learning algorithm for anomaly detection that isolates anomalous point clouds by constructing a series of random decision trees (also known as isolation trees). However, the traditional Isolation Forest algorithm applies equal weight to all data points, making it unsuitable for the complex terrain variations of open-pit mines, such as slopes, steps, and platforms. Therefore, this paper further improves the Isolation Forest algorithm to achieve anomalous point cloud filtering while adapting to the terrain.

[0063] Figure 3 A flowchart illustrating the determination of multiple anomalous point clouds based on an improved isolated forest algorithm according to an embodiment of this disclosure is shown. Figure 3 As shown, in operation S301, the anomaly weight of each point cloud is determined based on the matching result between the terrain attribute data and weight conditions. In operation S302, the path length of each point cloud in the map data is determined using the Isolation Forest algorithm. The segmentation criterion of the Isolation Forest algorithm is determined based on the terrain attribute data, and the path length is determined based on the number of times each point cloud is segmented into isolated anomaly point clouds. In operation S303, the anomaly score of each point cloud is determined based on its anomaly weight and path length. In operation S304, multiple anomaly point clouds are identified from the map data based on the comparison result between the anomaly scores and anomaly thresholds.

[0064] According to one embodiment of this disclosure, the isolated forest algorithm in step S302 above is an improved isolated forest algorithm (which can be understood as an isolated forest algorithm that introduces abnormal weights).

[0065] Regarding the above operation S301, in one embodiment, the weight condition can be predetermined. For example, the weight condition can be: when the terrain attribute data matches a certain condition interval among multiple condition intervals, the weight corresponding to that condition interval is determined as the abnormal weight of the point cloud.

[0066] It should be noted that the point clouds used here include: individual point clouds in map data, or point clouds obtained after comparing with the standard terrain attribute data mentioned above.

[0067] Regarding the above operation S302, during the sampling phase of the isolated forest, samples can be directly extracted multiple times and randomly from the point cloud. A point cloud (such as) Take 256), each time draw Each point cloud is considered as a set of point clouds for one segmentation. Then, in the segmentation phase, for each point cloud set, a bisection method is used to spatially divide all point clouds in the set according to the segmentation criterion, until only one point cloud remains on one side of the bisection. The segmentation criterion includes a segmentation dimension and a segmentation threshold. The segmentation dimension can be randomly selected from the terrain attribute data, such as any one of x, y, z, slope value, or elevation value (or, the elevation difference determined based on the difference between the elevation value of the point cloud and the standard terrain attribute data). The segmentation threshold is randomly determined from the point cloud set. The bisection method can be: under the segmentation dimension, point clouds within the set whose segmentation value under that segmentation dimension is less than the segmentation threshold are classified as left, and those whose value is greater than the segmentation threshold are classified as right. After continuing the above segmentation operation on multiple point cloud sets, for each point cloud in the multiple point cloud sets (duplicate point clouds may exist between multiple point cloud sets), the number of times it was classified as an anomalous point cloud in each segmentation is used as the path length. An anomaly score can be obtained by combining the path lengths of multiple segmentations of the same point cloud (if multiple segmentations exist) and the anomaly weight of that point cloud.

[0068] For operation S303, for any point cloud x in multiple point cloud sets, if it is selected into t point cloud sets, then calculate its average path length E[h(x)] after t segmentations, and the initial anomaly score. The calculation formula is:

[0069] (1)

[0070] (2)

[0071] (3)

[0072] In equations (1) to (3), For the standardized path length, that is The average path length of each sample, H(n), is the harmonic number, n = -1. ≈0.5772 is Euler's constant.

[0073] Determining the initial outlier score Then, taking the i-th point cloud as an example, the i-th anomaly score is... ,in, For the abnormal weights of the point cloud, for The original outlier score.

[0074] For operation S304, the anomaly threshold τ can be determined according to actual needs, such as setting it to any value between 0.8 and 1.0, where the anomaly score of the i-th point cloud satisfies... If the anomaly score is greater than or equal to τ, then the i-th point cloud is determined to be an anomaly point cloud (requiring warning and repair); otherwise, ... <τ, is determined to be a normal point cloud (no processing required).

[0075] In one specific embodiment, taking an anomaly threshold τ of 0.8 as an example, the following is an exemplary scenario for anomaly determination combined with terrain: In scenario 1: Anomaly identification needs to meet the requirement of accurate identification of minor anomalies in slope areas. For areas with steep slopes (e.g., 30°-45°) near the edge of steps, due to the higher terrain risk, the anomaly weight of the point cloud in this area is higher (e.g., (Values ​​can range from 1.8 to 2.25). For example, for a point cloud in this area with an abnormally small elevation deviation (e.g., only 0.8m), the initial anomaly score is about 0.6. After anomaly weighting correction, the anomaly score can be between 1.08 and 1.35. At this point, the anomaly score significantly exceeds the anomaly threshold τ, thus ensuring that the point cloud can be accurately and reliably identified as an anomalous point cloud, meeting the need for priority identification of minor slope anomalies and prevention of landslide risks.

[0076] In Scenario 2: Anomaly detection needs to meet the reasonable fault tolerance requirement of slight noise in the platform area. Within a flat platform area (e.g., slope ≤ 5° and far from the edge of a step), the terrain is stable, and the anomaly weight of the point cloud in this area is low (e.g., (1.0 can be taken). For example, for a point cloud in this area, its elevation deviation is slight (e.g., 1.2m), and the initial anomaly score is about 0.7. After the anomaly weight correction, the anomaly score can be 0.7. At this time, the anomaly score is less than the anomaly threshold τ, so the point cloud is not judged as an anomaly point cloud, which meets the platform's requirement of relaxing the judgment of slight noise and not affecting mining.

[0077] In scenario 3: Anomaly detection needs to meet the hierarchical judgment requirements of the transition zone at the edge of the step. In the transition zone near the edge of the step (e.g., a gentle slope within 5-20 meters of the edge), the terrain risk is between the two mentioned above, and the anomaly weight of the point cloud in this area is relatively moderate (e.g., (1.2 can be taken). Therefore, the abnormal score after the abnormal weight correction can maintain a reasonable correction scale and meet the hierarchical judgment requirements between steps and platforms.

[0078] In the embodiments of this disclosure, after determining the anomaly weights of each point cloud based on the terrain attribute data, the anomaly weights are weighted and fused with the segmented path lengths. This ensures that the determination of the final anomaly score does not affect the randomness of the segmentation (it does not interfere with the segmentation stage and the path length calculation stage), and can also accurately determine the anomaly point clouds that are adapted to the mine terrain using the anomaly weights, thereby achieving accurate and differentiated anomaly judgment.

[0079] According to embodiments of this disclosure, the path length of each point cloud in map data is determined using the Isolation Forest algorithm, including: determining the sampling weight of each point cloud based on its anomaly weight and the sum of the anomaly weights of all point clouds in the map data; using the Isolation Forest algorithm, sampling the map data multiple times according to the sampling probability distribution adjusted by the sampling weights to obtain multiple point cloud sets for segmentation; and determining the path length of each point cloud in the map data based on the number of times each point cloud in each point cloud set is segmented into isolated anomaly point clouds.

[0080] Traditional isolated forests use equal-probability random sampling during the sampling phase, meaning that all points are assigned the same sampling probability. However, this cannot match the risk differences in mine terrain. It will result in high-risk points on the edge of slopes / steps and low-risk points inside the platform being sampled with the same probability, leading to insufficient learning of anomalies in high-risk areas by the algorithm.

[0081] In this regard, a specific embodiment of the present disclosure applies anomaly weights to the sampling stage of the isolated forest algorithm, adjusting the sampling probability distribution of data points so that points with larger anomaly weights have a higher probability of being selected into the point cloud set, thereby strengthening the learning weights of samples in high-risk areas and making the algorithm more adaptable to the characteristics of mine terrain.

[0082] For point clouds Its sampling weight It can be determined based on the following formula:

[0083] (4)

[0084] In formula (4), is the number of point clouds in the map data, is the sum of the anomaly weights of all point clouds.

[0085] Multiple random samplings can be performed on the map data according to the sampling probability distribution adjusted by the sampling weights. Each time, samples are randomly selected to construct a point cloud set. The operation of determining the anomaly score using the Isolation Forest algorithm in this embodiment can be referred to above.

[0086] According to an embodiment of the present disclosure, the terrain attribute data includes a slope value and an elevation value, and the weight conditions include a slope weight condition and an elevation value weight condition: according to the matching result between the terrain attribute data of each point cloud and the weight conditions, the anomaly weight of each point cloud is determined, including: for each point cloud, according to the matching result between the slope value and the slope weight condition, the first weight is determined; according to the matching result between the elevation value and the elevation value weight condition, the second weight is determined; and according to the first weight and the second weight, the anomaly weight is determined.

[0087] In one embodiment, the slope weight condition can be determined based on the slope value S of the point cloud, and the elevation value weight condition can be determined based on the distance L from the point cloud to the edge of the step.

[0088] For example, the anomaly weight W can be determined based on the following operations. For the first weight W1: when the slope value S ≤ 30°, W1 = 1.0; when 30° < S ≤ 45°, W1 = 1.2 - 1.5. In this slope value range, the larger the slope value, the higher the first weight, and the two can change according to a linear relationship. For the second weight W2: when the distance L from the point cloud to the edge of the step ≤ 5m, W2 = 1.5; when 5m < L ≤ 20m, W2 = 1.2; when L > 20m, W2 = 1.0. The above design of the second weight satisfies the rule that the closer to the edge of the step, the higher the weight. Then, the product of the first weight and the second weight can be determined as the anomaly weight, such as W = W1 × W2 (the value range of the anomaly weight can be 1.0 - 2.25). Thus, the greater the anomaly weight of a point, the greater the probability of being determined as an anomaly in the Isolation Forest model.

[0089] In the embodiment of the present disclosure, by combining the weight fusion of two dimensions of the slope value and the elevation value, it is possible to adapt to the characteristics of the step area, slope area, and platform area in the mine scene, perform accurate anomaly recognition, and thereby improve the recognition accuracy of subsequent anomaly point clouds and anomaly areas.

[0090] According to embodiments of this disclosure, determining the anomaly type of at least one anomaly region based on its respective point cloud geometric features and terrain attribute data of each anomaly point cloud within the anomaly region includes: determining the anomaly type of at least one anomaly region based on the matching results between point cloud geometric features and geometric conditions, and / or the matching results between terrain continuity features and continuity conditions determined using terrain attribute data of multiple anomaly point clouds within the anomaly region. Preferably, the point cloud geometric features include point cloud density, and the terrain attribute data includes slope value and elevation value; determining the anomaly type of each of at least one anomalous region includes: if the point cloud density of the anomalous point cloud in the anomalous region is less than a first density threshold, and the difference between the point cloud density of the anomalous point cloud and the point cloud density of the adjacent normal region is greater than or equal to a second density threshold, the anomaly type of the anomalous region is determined to be a missing anomaly; if the point cloud density of the anomalous point cloud in the anomalous region is greater than or equal to a third density threshold, and the standard deviation determined using the elevation values ​​of multiple anomalous point clouds in the anomalous region is greater than a first continuity threshold, the anomaly type of the anomalous region is determined to be a noise anomaly; if the point cloud density of the anomalous point cloud in the anomalous region is greater than or equal to a fourth density threshold, and the difference in slope value between two adjacent anomalous point clouds in the anomalous region is greater than a second continuity threshold, the anomaly type of the anomalous region is determined to be a distortion anomaly.

[0091] Point cloud geometric features include point cloud density, and geometric conditions include point cloud density conditions. Terrain attribute data can include slope and elevation values. By utilizing the slope and elevation values ​​of multiple anomalous point clouds within an anomalous region, continuous terrain features can be determined.

[0092] For example, the anomaly type of an anomaly region can be determined based on the matching results between the point cloud density corresponding to the anomaly region and the point cloud density condition. As another example, the anomaly type of an anomaly region can be determined based on the matching results between the terrain continuity features corresponding to the anomaly region and the continuity condition. Yet another example is that the anomaly type of an anomaly region can be determined based on the matching results between the point cloud density corresponding to the anomaly region and the point cloud density condition, as well as the matching results between the terrain continuity features corresponding to the anomaly region and the continuity condition.

[0093] For example, if the point cloud density of the abnormal region is less than the first density threshold (e.g., 0.1 points / m²), and the density difference between the abnormal region and the neighboring normal region (point cloud density ≥ 0.5 points / m²) is greater than or equal to the second density threshold (e.g., 0.4 points / m²), the abnormality type of the abnormal region can be determined to be a missing anomaly.

[0094] For example, if the point cloud density in the anomalous area is greater than or equal to the third density threshold (e.g., 0.5 points / m²) and the standard deviation of the point cloud elevation in the anomalous area is greater than the first continuity threshold (e.g., 0.8m), the terrain continuity is considered to be abnormal terrain fluctuation, and the anomalous type of the anomalous area can be determined to be noise-type anomalous.

[0095] Preferably, the DBSCAN clustering algorithm can be used to verify whether the number of clusters of noise points meets the corresponding category characteristics. For example, it can be determined whether the number of clusters is greater than 5. If it is greater than 5, it indicates that the point cloud is scattered and has no continuous terrain features, so as to avoid misjudging the "uneven transition area of ​​steps after blasting" (elevation standard deviation 1.0m but with continuous slope) as noise.

[0096] For example, if the point cloud density in the anomalous area is greater than or equal to the fourth density threshold (e.g., 0.3 points / m², which is used to exclude missing points), and there is a slope difference between adjacent points in the anomalous area that is greater than the second continuity threshold (e.g., 15°), then the terrain continuity is considered to be an abnormal slope transition, and the anomalous type of the anomalous area can be determined to be a distortion-type anomalous.

[0097] Preferably, terrain feature lines (such as step edge lines and slope direction lines) can be further extracted through edge detection algorithms. If the terrain feature lines show "breakage or offset" in the area (such as the step edge line suddenly offset by 3m), it is judged as distortion, thus avoiding misjudging "artificially excavated steep slopes" (slope difference of 20° but continuous feature lines) as distortion.

[0098] In the embodiments of this disclosure, whether a region is abnormal is determined from the spatial geometric distribution dimension of the point cloud by using point cloud geometric features, and further determined from the rationality design dimension of environmental characteristics by combining terrain continuity. Thus, targeted and accurate anomaly identification of terrain regions can be achieved.

[0099] According to embodiments of this disclosure, an abnormal region is repaired according to a repair strategy matching the anomaly type, including: when the abnormal region is a missing anomaly or a distortion anomaly, the abnormal region is repaired according to a repair strategy matching the anomaly type and the terrain region to which the abnormal region belongs; when the abnormal region is a noise anomaly, the abnormal region is repaired according to a repair strategy matching the anomaly type; preferably, the abnormal region includes multiple sub-regions belonging to different terrain regions, and when the abnormal region is a missing anomaly or a distortion anomaly, each sub-region is repaired according to a repair strategy matching the anomaly type and the terrain region to which each sub-region belongs.

[0100] According to the repair strategy that matches the anomaly type, the abnormal point cloud can be removed from the abnormal region before repair, or the abnormal point cloud in the abnormal region can be repaired directly.

[0101] In one embodiment, when the anomalous area is a noise-type anomaly, it can be repaired according to a repair strategy matching noise-type anomalies. In another embodiment, when the anomalous area is a missing anomaly, it can be repaired according to a repair strategy matching both the missing anomaly and the terrain region to which it belongs. For example, if the anomalous area is a missing anomaly and the terrain region to which it belongs is a slope region, then the anomalous area can be repaired according to a repair strategy matching both the missing anomaly and the slope region. In yet another embodiment, when the anomalous area is a distortion-type anomaly, it can be repaired according to a repair strategy matching both the distortion-type anomaly and the terrain region to which it belongs. For example, if the anomalous area is a distortion-type anomaly and the terrain region to which it belongs is a platform region, then the anomalous area can be repaired according to a repair strategy matching both the distortion-type anomaly and the platform region.

[0102] In a preferred embodiment, the anomalous region comprises multiple sub-regions belonging to different terrain regions. When the anomalous region is a missing or distorted anomalous region, each sub-region is repaired according to a repair strategy matching the anomalous type and the terrain region to which each sub-region belongs.

[0103] For example, if the anomaly is a missing anomaly, and the anomaly includes a first sub-region belonging to the slope region and a second sub-region belonging to the platform region, then the first sub-region can be repaired according to the repair strategy that matches the missing anomaly and the slope region; the second sub-region can be repaired according to the repair strategy that matches the missing anomaly and the platform region.

[0104] In the embodiments of this disclosure, customized repair is carried out for each anomaly type by using a variety of repair strategies that match the anomaly type and / or the terrain region to which it belongs. This can better adapt to the defects of various terrain regions and improve the integrity of point cloud repair within the anomaly region.

[0105] According to embodiments of this disclosure, when the abnormal region is a missing anomaly or a distorted anomaly, repairing the abnormal region according to a repair strategy matching the anomaly type and the terrain region to which the abnormal region belongs includes: when the abnormal region is a distorted anomaly, removing abnormal point clouds from the abnormal region to obtain a preliminary repair region; fitting the preliminary repair region with the terrain boundary lines identified from the preliminary repair region as boundary constraints to obtain the repaired target region; when the abnormal region is a missing anomaly, removing abnormal point clouds from the abnormal region to obtain a preliminary repair region; determining the interpolation weight of each point cloud based on the correspondence between each point cloud in the preliminary repair region and the key point clouds in the terrain design map, as well as the correspondence between each point cloud and the terrain boundary lines; and using an interpolation algorithm, determining the repaired target region based on the elevation values ​​and interpolation weights of each point cloud in the preliminary repair region.

[0106] When the abnormal region is a noise-type anomaly, the abnormal region is repaired according to the repair strategy for the anomaly type. This includes: determining the size of the filtering window centered on each point cloud based on the slope range of the slope value of each point cloud in the abnormal region, wherein the size of the filtering window is different for multiple point clouds; determining the terrain similarity between the point cloud and multiple point clouds other than the point cloud in the filtering window based on the difference in slope value and aspect value between the point cloud and multiple point clouds other than the point cloud in the filtering window based on the terrain similarity; adjusting the elevation weight between the point cloud and multiple point clouds other than the point cloud in the filtering window based on the terrain similarity; determining the filtered elevation value of the point cloud based on the adjusted elevation weight, distance weight, and the individual elevation value of multiple point clouds other than the point cloud in the filtering window, so that the repaired target region is composed of point clouds with filtered elevation values.

[0107] The restoration of open-pit mine terrain data needs to balance "accuracy adaptation to the scenario" and "feature preservation requirements." Specifically, bench platforms need to be matched with mining design parameters first (to avoid affecting the calculation of mining elevation), slope areas need to ensure slope continuity (related to slope stability analysis), and platform areas need to balance smoothness and data accuracy (to adapt to equipment travel path planning). To meet the differentiated needs of open-pit mine scenarios, this disclosure embodiment constructs an intelligent layered restoration system that deeply binds "anomaly type - terrain scene features - restoration strategy" for three types of terrain data anomalies in open-pit mines: missing anomalies, noise anomalies, and distortion anomalies.

[0108] For example, to address anomalies in slope point cloud data (such as laser scan occlusion, noise interference, contour distortion caused by abrupt terrain changes, and elevation deviation), after removing anomalous point clouds from the anomaly areas to obtain a preliminary repair area, edge detection algorithms are used to identify terrain boundary lines, such as key slope feature lines (e.g., step edge lines, slope direction lines), from the preliminary repair area. Using these terrain boundaries as necessary constraints (strong constraints), a triangulation network is constructed layer by layer based on design parameters, and the elevation is corrected.

[0109] The specific implementation process is as follows: After identifying the edge lines of the steps (the boundary between the upper and lower edges of the steps) and the slope direction lines (the main outline of the slope's extension direction), the automatically extracted terrain boundary lines can be manually verified and corrected by combining on-site measured data and design drawings to eliminate extraction deviations. Then, based on the natural stratification of the slope, the step levels, and the design slope segment division, the point cloud is split by region to obtain point clouds used to construct triangular meshes at different levels. During the construction process, the terrain boundary lines after deviation elimination are used as the edges of the triangular mesh construction. The edges of each layer of the triangular mesh are constrained to fit the terrain boundary lines after deviation elimination, while controlling the density and shape of the triangular mesh units and ensuring they do not exceed the terrain boundary lines. This results in a hierarchical triangular mesh, ensuring that the mesh structure is highly compatible with the local terrain features of the slope, while avoiding unreasonable mesh shapes such as elongated triangles.

[0110] For the constructed triangular network, the least squares method is used to fit the elevation surface of the point cloud within each triangular network to obtain the repair areas for each layer. Then, the layered repair areas are superimposed in the order of natural layer, step layer, and designed slope section to obtain the repaired target area. Alternatively, for local areas where there are still deviations after fitting, the Kriging interpolation algorithm is used to interpolate effective point clouds to correct the deviations in those local areas, thus obtaining the repair areas for each layer.

[0111] For example, missing anomalies can be directly corrected through interpolation. During the correction process, precise correction can be achieved using topographic design maps (i.e., mine design maps). Topographic design maps typically include a key point cloud and continuous topographic points located on the topographic boundary line. The key point cloud, also known as design-related points, refers to measured / design points on the boundary of the missing area (i.e., the anomaly area of ​​the missing anomaly) that directly correspond to key parameters in the formal open-pit mine design drawings. For example, the key point cloud includes the design elevation points of the platform, the design slope toe / apex points of the slope, and the boundary feature points between the platform and the slope. Continuous topographic points refer to measured points within a certain range around the missing area (usually within a 5-10m radius, adjustable according to the mine's topographic resolution) where the topography is continuous and free of abnormal noise / abrupt changes. Continuous topographic points include topographic points at the edge of the missing area that are not identified as anomalies, and continuous sampling points of adjacent normal topography. Based on the characteristics of the key point cloud and continuous topographic points, the interpolation weight of the key point cloud can be set to 0.8, and the interpolation weight of the continuous topographic points to 0.2.

[0112] Before interpolation, expert experience can be used to determine the correspondence between each point cloud in the preliminary repair area and the key point clouds and terrain boundaries in the anomaly area. This helps determine whether each point cloud is a key point cloud or a continuous terrain point on the terrain design map, thereby determining the interpolation weight of each point cloud. Then, the Kriging interpolation algorithm can be used to interpolate the elevation values ​​of the newly added point clouds based on the elevation values ​​of each point cloud and the interpolation weights. The newly added point clouds are then combined with the point clouds in the preliminary repair area to form the repaired target area.

[0113] For example, the mutation function of the Kriging interpolation algorithm can be corrected using interpolation weights, and the elevation values ​​of the newly added point cloud can be obtained by interpolation based on the corrected mutation function. The corrected mutation function is shown in the following formula:

[0114] (5)

[0115] In equation (5), To determine the elevation of the newly added point cloud, These are the Kriging interpolation coefficients (which can be set according to actual needs). To initially repair the interpolation weights of point cloud i in the region, Let i be the elevation value of point cloud i. The total number of point clouds participating in the interpolation.

[0116] In other embodiments, for the elevation value of the newly added point cloud obtained through interpolation, the elevation difference between this elevation value and the corresponding point cloud on the terrain design map can be determined. If this elevation difference exceeds a corresponding threshold, it can be adjusted by adjusting the Kriging interpolation coefficients or the point clouds involved in the interpolation and their number, until the elevation difference between the newly added point cloud obtained through interpolation and the corresponding point cloud does not exceed the corresponding threshold. As an example, for a scenario where the missing area is a step or platform area, if the elevation difference between the interpolated elevation value and the corresponding point cloud on the terrain design map is greater than 0.3m, then re-interpolation is performed. As another example, for a scenario where the missing area is a slope area, the slope value of the newly added point cloud can be calculated, and the deviation between this slope value and the slope value of the corresponding point cloud on the terrain design map can be calculated. If this deviation exceeds a corresponding threshold (e.g., 1°), it can be adjusted by adjusting the Kriging interpolation coefficients or the point clouds involved in the interpolation and their number.

[0117] For noisy anomalies, a bilateral filter can be used to denoise the point clouds within the noisy anomaly region. Preferably, the denoising can be performed on the anomalous point clouds within the anomaly region. During the filtering process, the size of the filtering window for the anomalous point cloud can be dynamically adjusted based on the slope range of the slope where the anomalous point cloud is located. Then, the elevation weights are optimized based on the terrain similarity between the anomalous point cloud and multiple point clouds in the filtering window other than the anomalous point cloud, achieving accurate noise removal while preserving key terrain features such as step edges and slope gradients to the greatest extent possible.

[0118] For example, a large filter window is used for gentle slopes (slope less than 15°) to expand the filtering range and improve noise removal; a medium filter window is used for medium slopes (slope between 15° and 45°); and a small filter window is used for steep slopes (slope greater than 45°) to avoid over-smoothing of step edges and abrupt slope changes. Specifically, if the slope value of an anomaly point cloud falls within the slope range corresponding to a gentle slope, the size of the filter window for that anomaly point cloud is determined to be 7×7; if it falls within the range of a medium slope, the size is 5×5; and if it falls within the range of a steep slope, the size is 3×3. The slope can be determined by the sum of the squares of the partial derivatives of the z-coordinate of the anomalous point cloud with respect to the x and y coordinates. For example, the slope value can be determined by the following formula (6):

[0119] (6)

[0120] in, and These are the sum of squares of the elevation partial derivatives of the z-coordinate of the anomalous point cloud in the x and y coordinates, respectively.

[0121] For any anomalous point cloud within an anomalous region, the filtering window for that anomalous point cloud is centered on it. This anomalous point cloud can be called the point cloud to be filtered. Other point clouds (including only normal point clouds, or all point clouds within the filtering window except the point cloud to be filtered) are called the window's neighborhood point clouds. The terrain similarity between the point cloud to be filtered and the window's neighborhood point clouds can be determined based on the difference between their slope and aspect values. For example, terrain similarity... The calculation formula is as follows:

[0122] (7)

[0123] In equation (7), , These are the slope values ​​of the point cloud to be filtered and the point cloud in the window's neighborhood, respectively. , These are the slope aspect values ​​of the point cloud to be filtered and the point cloud in the window's neighborhood, respectively. The terrain similarity value ranges from 0 to 1. The closer the terrain similarity is to 1, the more similar the terrain is between the point cloud to be filtered and the point cloud in the window's neighborhood.

[0124] Understandably, the higher the terrain similarity, the more similar the terrain between two point clouds, and the more this part of the point cloud needs to be retained during filtering. Therefore, elevation weights can be positively correlated with terrain similarity; for example, the higher the terrain similarity, the greater the elevation weight. Alternatively, elevation weights can be stratified using a similarity threshold. For instance, when terrain similarity is ≥0.8, the elevation weight of the neighboring point clouds within the window is increased by 20% based on the baseline elevation weight to strengthen the contribution of similar terrain point clouds to the filtering results and preserve the overall terrain trend. When terrain similarity is <0.5, the elevation weight of the neighboring point clouds within the window is decreased by 30% based on the baseline elevation weight to weaken the interference of dissimilar terrain points (such as noise points or feature abrupt changes) and avoid distortion of the filtering results. When terrain similarity is between 0.5 and 0.8, the baseline elevation weight is used as the elevation weight of the neighboring point clouds within the window.

[0125] In addition, during the filtering process, the distance weights of the bilateral filter can be designed based on empirical values ​​so that the two are negatively correlated. For example, the greater the distance between the window's neighborhood point cloud and the point cloud to be filtered, the smaller the distance weight.

[0126] After determining the elevation weight and distance weight, the filtered elevation value of each point cloud to be filtered can be determined using the following adaptive bilateral filtering elevation calculation formula:

[0127] (8)

[0128] In equation (8), Point cloud to be filtered Filtered elevation values For filtering window, Distance weights To utilize the elevation weights determined by terrain similarity, Point cloud for window neighborhood The elevation value.

[0129] In one embodiment, the filtered elevation values ​​of each point cloud to be filtered can be determined through multiple rounds of iterative calculations, such as 2-3 times, with each iteration using the elevation value obtained in the previous round. The number of iterations can be determined according to actual needs to avoid excessive iterations that could lead to feature blurring. In another embodiment, after filtering, the filtering effect can be verified using two metrics: noise removal rate and feature retention rate. The noise removal rate can be determined by calculating the proportion of the number of index point clouds before and after filtering (usually points with an elevation difference exceeding 3 times the standard deviation), and the feature retention rate can be determined by comparing the positional deviations of step edge points and slope toe / slope peaks before and after filtering. If the filtered noise removal rate is <75% and / or the feature retention rate is <98%, filtering can be performed again after readjusting the elevation weight optimization ratio or the size of the filtering window.

[0130] Preferably, after repairing the abnormal area, the accuracy of the repaired target area can be verified and optimized. For example, at least 20 field sampling points (measured with a total station, accuracy ≤ ±0.05m) can be selected from the repaired target area through field sampling. The elevation error ΔZ_repair between the target area and the field sampling points is calculated. If the proportion of field sampling points with ΔZ_repair ≤ 0.5m is ≥ 90%, the repair is considered qualified; otherwise, the corresponding repair parameters (such as interpolation weights, filtering windows, etc.) are adjusted and the repair is repeated. Alternatively, the terrain continuity can also be verified. For example, the slope transition difference ΔS between the target area and the adjacent normal area is calculated. If the proportion of areas with ΔS ≤ 5° is ≥ 95%, the terrain is considered continuous; otherwise, a secondary smoothing process is performed on the transition area between the target area and the adjacent normal area.

[0131] In the embodiments of this disclosure, the data of the target area after repair for various anomaly types can be exported as LAS format (point cloud) or GeoTIFF format (digital elevation model DEM) for subsequent map applications.

[0132] Based on the above description, the embodiments of this disclosure provide a map data repair method with the following technical effects: 1. Multi-source data fusion improves basic accuracy: By fusing LiDAR point cloud and UAV image data and combining them with mine design parameters, the problem of insufficient information caused by occlusion from a single data source is solved; 2. High anomaly identification accuracy: The improved isolated forest algorithm introduces terrain feature weights (i.e., anomaly weights mentioned above), which are adapted to the characteristics of large slope undulations in open-pit mines, resulting in high anomaly identification accuracy and low misjudgment rate; 3. Layered repair adapts to different anomalies: Repair logic is designed separately for three types of anomalies: missing, noise, and distortion. Especially in key areas such as steps and slopes, the elevation error after repair is ≤0.5m, meeting the high-precision application requirements of mines; 4. Strong practicality: The algorithm process is clear, compatible with existing GIS software, and the output data format supports mainstream application scenarios such as mine development planning and slope stability analysis.

[0133] Figure 4 A block diagram of a map data repair apparatus according to an embodiment of the present disclosure is shown. Figure 4As shown, the map data repair apparatus 400 includes an abnormal region determination module 410, an abnormal type determination module 420, and a repair module 430. The abnormal region determination module 410 is used to determine at least one abnormal region including abnormal point clouds from the map data based on the terrain attribute data of each point cloud in the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point clouds. The abnormal type determination module 420 is used to determine the abnormal type of each of the at least one abnormal region based on the geometric features of the point clouds of each of the at least one abnormal region and the terrain attribute data of each abnormal point cloud within the abnormal region. The repair module 430 is used to repair the abnormal regions according to a repair strategy matching the abnormal type. The operation of the apparatus portion of this embodiment is similar to the operation of the method portion described above, and the specific operations will not be repeated here.

[0134] Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be implemented by dividing them into multiple modules. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as hardware circuitry, such as a Field-Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a System-on-Chip, a System-on-a-Substrate, a System-on-Package, an Application-Specific Integrated Circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, one or more of the modules, submodules, units, and subunits according to embodiments of the present disclosure can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0135] Figure 5 A block diagram of an electronic device suitable for implementing a method for repairing map data according to an embodiment of the present disclosure is shown. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein. Figure 5As shown, an electronic device 500 according to an embodiment of the present disclosure includes a processor 501, which can perform various appropriate actions and processes according to a program stored in a ROM (Read-Only Memory) 502 or a program loaded from a storage portion 508 into a RAM (Random Access Memory) 503. The processor 501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 501 may also include onboard memory for caching purposes. The processor 501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0136] RAM 503 stores various programs and data required for the operation of electronic device 500. Processor 501, ROM 502, and RAM 503 are interconnected via bus 504. Processor 501 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 502 and / or RAM 503. It should be noted that the programs may also be stored in one or more memories other than ROM 502 and RAM 503. Processor 501 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0137] According to embodiments of this disclosure, the electronic device 500 may further include an input / output (I / O) interface 505, which is also connected to a bus 504. The electronic device 500 may also include one or more of the following components connected to the input / output (I / O) interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the input / output (I / O) interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 510 as needed so that computer programs read from it can be installed into the storage section 508 as needed.

[0138] According to embodiments of this disclosure, the method flow according to embodiments of this disclosure can be implemented as a computer software program. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by processor 501, it performs the functions defined in the system of embodiments of this disclosure. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0139] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0140] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples include, but are not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503 described above.

[0141] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of this disclosure. When the computer program product is run on an electronic device, the program code enables the electronic device to implement the map data repair method provided in the embodiments of this disclosure. When the computer program is executed by processor 501, it performs the functions defined in the system / device of the embodiments of this disclosure. According to embodiments of this disclosure, the systems, devices, modules, units, etc., described above can be implemented by computer program modules.

[0142] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 509, and / or installed from a removable medium 511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0143] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on a user's computing device, partially on a user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions. Those skilled in the art will understand that the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways, even if such combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments of this disclosure may be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0145] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A method for repairing map data, comprising: Based on the terrain attribute data of each point cloud in the map data, at least one abnormal region including abnormal point clouds is determined from the map data, wherein the terrain attribute data is used to characterize the terrain at the location of the point cloud. Based on the point cloud geometric features of at least one of the anomalous regions and the terrain attribute data of each of the anomalous point clouds within the anomalous region, determine the anomalous type of each of the at least one of the anomalous regions; and The abnormal region is repaired according to a repair strategy that matches the abnormality type.

2. The method according to claim 1, wherein, The step of determining at least one anomalous region, including anomalous point clouds, from the map data based on the terrain attribute data of each point cloud in the map data includes: Based on the terrain attribute data of each point cloud in the map data, multiple anomalous point clouds are identified from the map data; and Cluster the multiple anomalous point clouds and determine at least one anomalous region in the map data that corresponds to the clustering results.

3. The method according to claim 2, wherein, The step of identifying multiple anomalous point clouds from the map data based on the terrain attribute data of each point cloud includes: Obtain standard terrain attribute data corresponding to at least one terrain region, wherein the map data includes at least one terrain region, and each terrain region includes multiple point clouds; Based on the comparison results between the terrain attribute data of each point cloud and the standard terrain attribute data of the terrain region to which each point cloud belongs, multiple abnormal point clouds are identified from the map data. Preferably, the terrain fluctuation range is determined using the standard terrain attribute data, and multiple abnormal point clouds are determined from the map data based on the terrain attribute data of each point cloud and the comparison results between the terrain fluctuation ranges of the terrain regions to which each point cloud belongs.

4. The method according to claim 2 or 3, wherein, The step of identifying multiple anomalous point clouds from the map data based on the terrain attribute data of each point cloud includes: The abnormal weights of each point cloud are determined based on the matching results between the terrain attribute data and the weight conditions of each point cloud. The path length of each point cloud in the map data is determined using the isolated forest algorithm, wherein the segmentation criterion of the isolated forest algorithm is determined based on the terrain attribute data, and the path length is determined according to the number of times each point cloud is segmented into isolated anomalous point clouds. Based on the anomaly weight and path length of each point cloud, an anomaly score is determined for each point cloud; and Based on the comparison results between the anomaly scores and anomaly thresholds of each point cloud, multiple anomalous point clouds are identified from the map data; Preferably, the point cloud used in the isolated forest algorithm includes: anomaly point clouds determined from the map data based on a comparison between the terrain attribute data of each point cloud and the standard terrain attribute data of the terrain region to which each point cloud belongs.

5. The method according to claim 4, wherein, The method of using the isolated forest algorithm to determine the path length of each point cloud in the map data includes: The sampling weight of each point cloud is determined based on the anomaly weight of each point cloud and the sum of the anomaly weights of each point cloud in the map data. Using the isolated forest algorithm, the map data is sampled multiple times according to the sampling probability distribution adjusted by the sampling weights to obtain multiple point cloud sets for segmentation; The path length of each point cloud in the map data is determined based on the number of times each point cloud in each point cloud set is segmented into isolated abnormal point clouds.

6. The method according to claim 4 or 5, wherein, The terrain attribute data includes slope value and elevation value, and the weighting conditions include slope weighting conditions and elevation value weighting conditions. The step of determining the anomaly weight of each point cloud based on the matching result between the terrain attribute data and weight conditions of each point cloud includes: For each point cloud, a first weight is determined based on the matching result between the slope value and the slope weight condition; a second weight is determined based on the matching result between the elevation value and the elevation weight condition; and an anomaly weight is determined based on the first weight and the second weight.

7. The method according to any one of claims 1 to 6, wherein, The step of determining the anomaly type of at least one of the at least one anomalous regions based on the point cloud geometric features of each of the at least one anomalous region and the terrain attribute data of each anomalous point cloud within the anomalous region includes: Based on the matching results between the geometric features and geometric conditions of the point cloud, and / or the matching results between the continuous features and continuous conditions of the terrain determined by the terrain attribute data of multiple anomalous point clouds within the anomalous region, at least one anomalous type of each of the anomalous regions is determined. Preferably, the point cloud geometric features include point cloud density, and the terrain attribute data includes slope and elevation values; determining the anomaly type of each of the at least one of the anomaly regions includes: If the point cloud density of the abnormal point cloud in the abnormal region is less than a first density threshold, and the difference between the point cloud density of the abnormal point cloud and the point cloud density of the adjacent normal region is greater than or equal to a second density threshold, the abnormality type of the abnormal region is determined to be a missing anomaly. If the point cloud density of the abnormal point cloud in the abnormal region is greater than or equal to the third density threshold, and the standard deviation determined by the elevation values ​​of the multiple abnormal point clouds in the abnormal region is greater than the first continuity threshold, then the abnormality type of the abnormal region is determined to be a noise-type abnormality. If the point cloud density of the anomalous point cloud in the anomalous region is greater than or equal to the fourth density threshold, and the difference in slope between two adjacent anomalous point clouds in the anomalous region is greater than the second continuity threshold, then the anomalous type of the anomalous region is determined to be a distortion-type anomalous region.

8. The method according to any one of claims 1 to 7, wherein, The repair of the abnormal region according to the repair strategy matching the abnormality type includes: If the abnormal region is a missing anomaly or a distortion anomaly, the abnormal region is repaired according to a repair strategy that matches the anomaly type and the terrain region to which the abnormal region belongs. If the abnormal region is a noise-type abnormality, the abnormal region shall be repaired according to a repair strategy that matches the abnormality type. Preferably, the abnormal region includes multiple sub-regions belonging to different terrain regions. When the abnormal region is a missing anomaly or a distortion anomaly, each sub-region is repaired according to a repair strategy that matches the anomaly type and the terrain region to which each sub-region belongs.

9. The method according to claim 8, wherein, In cases where the anomalous region is a missing or distorted anomalous region, the anomalous region is repaired according to a repair strategy that matches the anomalous type and the terrain region to which the anomalous region belongs. This includes: In the case that the abnormal region is a distortion-type anomaly, the abnormal point cloud is removed from the abnormal region to obtain a preliminary repair region; the preliminary repair region is fitted with the terrain boundary line identified from the preliminary repair region as a boundary constraint to obtain the repaired target region. In the case that the abnormal area is a missing anomaly, the abnormal point cloud is removed from the abnormal area to obtain a preliminary repair area; based on the correspondence between each point cloud in the preliminary repair area and the key point cloud in the terrain design map, as well as the correspondence between each point cloud and the terrain boundary line, the interpolation weight of each point cloud is determined; using an interpolation algorithm, the target area after repair is determined based on the elevation value and interpolation weight of each point cloud in the preliminary repair area. When the abnormal region is a noise-type abnormality, the abnormal region is repaired according to a repair strategy for the abnormality type, including: In the case that the abnormal region is a noise-type abnormality, the size of the filtering window centered on each point cloud is determined according to the slope range in which the slope value of each point cloud in the abnormal region is located, wherein the sizes of the filtering windows of multiple point clouds are different. For each point cloud, the terrain similarity between the point cloud and the multiple point clouds in the filter window (excluding the point cloud itself) is determined based on the difference in slope value and the difference in aspect value between the point cloud and the multiple point clouds in the filter window. The elevation weights between the point cloud and multiple point clouds other than the point cloud in the filtering window are adjusted based on the terrain similarity. Based on the adjusted elevation weights and distance weights between the point cloud and multiple point clouds other than the point cloud in the filtering window, as well as the individual elevation values ​​of the multiple point clouds other than the point cloud in the filtering window, the filtered elevation value of the point cloud is determined, so that the repaired target area is formed by the point clouds with filtered elevation values.

10. An electronic device, comprising: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images