A measurement method, device, storage medium and electronic device

Abstract

The application provides a measurement method, device, storage medium and electronic equipment, and relates to the technical field of measurement. The measurement method comprises the following steps: obtaining point cloud data and a target image of a target area; determining a first abnormal object and a second abnormal object in the point cloud data based on the target image; and generating measurement data based on the first abnormal object and the second abnormal object; wherein the first abnormal object represents an interference point caused by foreign matter reflection in the target area, the second abnormal object represents an interference area formed by the foreign matter, and the first density of the first abnormal object is less than the second density of the second abnormal object.

Description

Technical Field

[0001] This invention relates to the field of measurement technology, and in particular to a measurement method, apparatus, storage medium, and electronic device. Background Technology

[0002] LiDAR point cloud technology is widely used in many fields such as topographic surveying and environmental monitoring. However, in actual measurement environments, various interference factors often exist, such as floating objects on the water surface, vegetation cover areas, or other temporary obstacles. These interference factors can lead to inaccuracies in the collected point cloud data. Related technologies typically use filtering or noise reduction methods to address these issues. However, in complex environments where interference objects are unevenly distributed and have varying shapes, these technologies struggle to distinguish between different types of abnormal data, directly affecting the accuracy of the measurement data. Summary of the Invention

[0003] The main objective of this invention is to provide a measurement method, apparatus, storage medium, and electronic device, thereby overcoming the shortcomings of the prior art.

[0004] The first aspect of the present invention provides a measurement method, comprising:

[0005] Obtain point cloud data and target image of the target area;

[0006] Based on the target image, determine the first and second anomalous objects in the point cloud data;

[0007] Measurement data is generated based on the first abnormal object and the second abnormal object;

[0008] Wherein, the first abnormal object represents an interference point caused by reflection of a foreign object in the target area, the second abnormal object represents an interference area formed by the foreign object, and the first density of the first abnormal object is less than the second density of the second abnormal object.

[0009] According to an embodiment of the present invention, determining the first anomalous object and the second anomalous object in the point cloud data based on the target image includes:

[0010] The target image is converted to a color space to obtain a color image;

[0011] Candidate interference regions in the target image are determined based on preset color threshold ranges and texture thresholds;

[0012] The point cloud data is mapped to the color space corresponding to the target image to obtain the mapped point cloud;

[0013] Based on the density space of the mapped point cloud within the candidate interference region, the mapped point cloud is divided into the first abnormal object and the second abnormal object.

[0014] According to an embodiment of the present invention, dividing the mapped point cloud into a first anomalous object and a second anomalous object based on the density space of the mapped point cloud within the candidate interference region includes:

[0015] Taking any point in the mapped point cloud as the center, determine a spherical neighborhood with a preset radius;

[0016] Calculate the spatial density of the center in the spherical neighborhood;

[0017] If the spatial density is greater than or equal to a first preset density threshold, the mapped point cloud is identified as the first abnormal object.

[0018] If the spatial density is less than or equal to a second preset density threshold, the mapped point cloud is identified as the second abnormal object.

[0019] According to an embodiment of the present invention, after obtaining the point cloud data of the target area and the target image, the method further includes:

[0020] First pose information and second pose information are obtained, and the first pose information and the second pose information represent the spatial position and attitude angle of the target device when collecting point cloud data and target image of the target area;

[0021] Based on the first pose information and the second pose information, the temporal correspondence and spatial transformation relationship between the point cloud data and the target image are determined;

[0022] Based on the time correspondence and the spatial transformation relationship, spatiotemporal registration is performed on the point cloud data and the target image.

[0023] According to an embodiment of the present invention, the method further includes:

[0024] The motion characteristics of the foreign object are obtained, and the motion characteristics characterize the dynamic change pattern of the foreign object under the action of wind and water flow.

[0025] A height change prediction model for the foreign object is constructed based on the aforementioned motion characteristics;

[0026] The height change prediction model is used to compensate for height anomalies in the point cloud data caused by the dynamic swinging of foreign objects.

[0027] According to an embodiment of the present invention, after compensating for the height anomalies in the point cloud data caused by the dynamic swaying of foreign objects based on the height change prediction model, the method further includes:

[0028] If abnormal regions still exist after compensating the point cloud data, the spatial scale of the abnormal regions and the valid point cloud data within the abnormal regions are obtained.

[0029] If the spatial scale of the abnormal region is greater than a preset scale threshold, a surface fitting model is generated based on the effective point cloud data, the surface fitting model is solved to obtain a first height, and the abnormal region is compensated based on the first height.

[0030] If the spatial scale of the abnormal region is less than or equal to a preset scale threshold, the valid point cloud data is interpolated to obtain a second height, and the abnormal region is compensated based on the second height.

[0031] According to an embodiment of the present invention, generating measurement data based on the first abnormal object and the second abnormal object includes:

[0032] In the point cloud data, the point cloud data corresponding to the first abnormal object is removed and the point cloud data corresponding to the second abnormal object is reconstructed to obtain the measurement data.

[0033] A second aspect of the present invention provides a measuring device, comprising:

[0034] The data acquisition module is used to acquire point cloud data and target images of the target area.

[0035] An anomaly determination module is used to determine a first anomaly object and a second anomaly object in the point cloud data based on the target image;

[0036] The data generation module is used to generate measurement data based on the first abnormal object and the second abnormal object;

[0037] Wherein, the first abnormal object represents an interference point caused by reflection of a foreign object in the target area, the second abnormal object represents an interference area formed by the foreign object, and the first density of the first abnormal object is less than the second density of the second abnormal object.

[0038] A third aspect of the present invention provides an electronic device comprising:

[0039] One or more processors;

[0040] Memory, used to store one or more programs.

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement the measurement method as described above.

[0042] A fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the measurement method described above.

[0043] By employing the technical solution of this invention, point cloud data and target images of the target area are acquired. Then, based on the target image, abnormal objects in the point cloud data are identified and categorized into first abnormal objects and second abnormal objects. These different abnormal objects are effectively distinguished by density differences. This allows the system to adopt targeted processing strategies based on the different properties of the abnormal objects, thereby improving the accuracy of the measurement data.

[0044] Compared to filtering or noise reduction methods in related technologies that rely on a single data source, this invention can more accurately locate abnormal areas through image-assisted recognition. By classifying abnormal objects into low-density first abnormal objects and high-density second abnormal objects, the system can distinguish different types of interference and then take more targeted processing measures, thereby reducing the situation of erroneous deletion of effective data and retention of invalid data in complex environments. It is particularly suitable for measurement environments where interference is unevenly distributed and has various shapes. Attached Figure Description

[0045] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the present invention with reference to the accompanying drawings:

[0046] Figure 1 The flowchart of a measurement method provided in a typical embodiment of the present invention is illustrated schematically.

[0047] Figure 2 The schematic diagram illustrates a structural block diagram of a measuring device provided in a typical embodiment of the present invention;

[0048] Figure 3 The diagram illustrates the structure of an electronic device provided in a typical embodiment of the present invention. Detailed Implementation

[0049] Hereinafter, embodiments of the present invention will 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 invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention 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 concept of the invention.

[0050] Figure 1A flowchart illustrating a measurement method provided by an embodiment of the present invention is shown, the measurement method including steps S101 to S103.

[0051] Step S101: Obtain point cloud data and target image of the target area. The target area refers to a specific lake / wetland area with dense aquatic vegetation requiring high-precision mapping. In this embodiment, the target area can be understood as an aquatic environment with high density and uneven distribution of aquatic vegetation, exhibiting dynamic swaying characteristics due to wind and water flow. In such environments, the water surface is often covered by a large amount of aquatic vegetation, with the stems and leaves floating on or slightly below the water surface, easily causing reflection interference to the lidar signal and forming anomalies in the mapping data. For example, the target area includes, but is not limited to: key water areas of concern for ecological monitoring, such as lakes within nature reserves, aquatic plant reserves, and key areas for water quality monitoring; lake shallows, reservoir siltation observation areas, and waterway dredging planning areas required for water conservancy engineering planning; and typical aquatic vegetation growth areas and wetland ecosystem change monitoring areas required for environmental research. After determining the target area, point cloud data and target image of the target area need to be collected. The point cloud data of the target area refers to a set of three-dimensional spatial points collected in the target lake / wetland area using lidar technology. In this embodiment of the invention, point cloud data can be understood as a set of three-dimensional coordinate points generated by a lidar scanning device emitting a laser beam and receiving reflected signals. Each point contains spatial location information and attributes such as reflection intensity. Point cloud data visually represents the three-dimensional information of the target area, including water surface topography, aquatic plant distribution, and underwater landforms. It is worth noting that the point cloud data of the target area is used to construct a high-precision water surface topography model and is the fundamental data for underwater landform mapping and water surface morphology analysis. Point cloud data can accurately reflect changes in water surface elevation and underwater topographic structure, but it can generate a large number of abnormal reflection points in areas with dense aquatic plants, requiring identification and processing using specific algorithms. Therefore, the spatial distribution characteristics, density variations, and elevation fluctuations of point cloud data are important bases for identifying aquatic plant areas and the actual water surface. However, relying solely on point cloud data is insufficient to effectively distinguish between the water surface and aquatic plants; therefore, target images of the target area need to be introduced. Target images refer to two-dimensional image data of the same target area captured by a high-definition visual camera within the synchronous or near-synchronous time of lidar scanning. In this embodiment of the invention, the target image can be understood as a digital image acquired by a visible light camera or a multispectral camera mounted on a mapping platform, which records visual feature information such as color and texture of the target area.

[0052] Specifically, target images are used to assist in identifying anomalous objects in point cloud data, particularly in areas covered by aquatic plants. Because aquatic plants exhibit distinct color and texture features in the visible light spectrum, target images can provide rich visual features not found in point cloud data. Furthermore, these visual features, through image processing techniques, can effectively distinguish between water surface and aquatic plant areas, providing crucial auxiliary information for identifying anomalous objects in point cloud data.

[0053] Despite acquiring both point cloud and image data simultaneously, a series of problems remain to be solved when conducting high-precision mapping in densely vegetationd lake wetlands. First, due to the high density and uneven distribution of aquatic plants, which sway dynamically under wind and water flow, the stems and leaves of these plants float on or slightly below the water surface when using lidar technology. This causes significant reflection interference to the lidar signal, creating anomalies in the mapping data. These anomalies are mixed with actual water surface reflections, making accurate differentiation difficult based solely on the geometric features of the point cloud data. Furthermore, while image data can be introduced to assist in identifying aquatic plant areas, point cloud and image data originate from different sensors, inevitably leading to temporal and spatial inconsistencies during acquisition. These inconsistencies manifest in two main ways: firstly, lidar and cameras are typically mounted on the same mapping platform, but their positions and orientations differ, resulting in different coordinate systems and viewing angles for the two sets of data; secondly, the data acquisition from the two sensors may not be strictly synchronized, resulting in temporal differences, especially during platform movement, which further exacerbates spatial mismatches. Therefore, this spatiotemporal inconsistency makes it impossible to directly establish a correspondence between 3D points in point cloud data and pixels in images, making it difficult to effectively utilize visual features in images to assist in identifying aquatic plant reflection points in point cloud data.

[0054] Even with two data sources, effectively fusing these heterogeneous data to achieve accurate aquatic plant area identification remains a challenge. Point cloud data provides discrete three-dimensional geometric information, while image data provides continuous two-dimensional visual features; the two differ fundamentally in data structure, information content, and expression. Without a precise spatiotemporal correspondence, the system cannot determine which point cloud data corresponds to which areas in the image, thus failing to accurately apply image analysis results to point cloud processing, ultimately affecting the accuracy of the mapping results. Furthermore, the dynamic changes in densely aquatic plant areas also increase the difficulty of mapping. In actual mapping, the position and shape of aquatic plants can change rapidly with variations in wind and water flow, requiring the data acquisition and processing system to maintain high temporal consistency to ensure the effectiveness of the fusion analysis. Without establishing an accurate temporal correspondence, it is difficult to ensure that the point cloud data and image data capture the scene state at the same moment, thus affecting the accuracy of aquatic plant identification.

[0055] To solve the above technical problems, based on the above embodiments, as an optional embodiment, after performing step S101, the following steps may also be performed:

[0056] Step S201: Obtain first pose information and second pose information. The first pose information and second pose information represent the spatial position and attitude angle of the target device when acquiring point cloud data and target images of the target area. Specifically, the first pose information of the point cloud data represents the spatial position and attitude angle information of the lidar during the acquisition of point cloud data of the target area. In this embodiment of the invention, it can be understood as the three-dimensional coordinate position and attitude angle, and other spatial positioning parameters, of the lidar sensor mounted on the mapping platform at each scan or data acquisition moment. It also includes corresponding timestamp information to record the precise time of data acquisition. Through the first pose information, the acquisition conditions and spatial reference system of each point cloud dataset can be traced.

[0057] Specifically, in the dynamic mapping of densely aquatic plant lake wetland environments, the first pose information helps the system accurately locate the acquisition position and time of each frame of point cloud data. This allows for the identification of the most spatiotemporally matching data pairs during image data fusion, reducing registration errors caused by platform movement or aquatic plant swaying. Correspondingly, the second pose information of the target image represents the spatial position and attitude angle information of the visual camera during the acquisition of target images of the target area. In this embodiment, it can be understood as the three-dimensional coordinate position (including longitude, latitude, elevation, or X, Y, Z coordinate values) and attitude angle (including pitch angle, yaw angle, roll angle) of the visible light camera or multispectral camera mounted on the mapping platform at each image acquisition moment, along with corresponding timestamp information to record the precise time of image capture. Through the second pose information, the shooting conditions and spatial reference frame of each image can be traced. In particular, in the dynamic environment of densely planted lakes and wetlands, the second pose information can help the system accurately locate the shooting angle and time of each image. This allows the system to calculate the relative positional relationship and temporal correspondence between the two sensors when fusing with point cloud data. This eliminates errors caused by differences in the installation position of high-definition cameras and asynchronous acquisition times, ensuring that when identifying aquatic plant areas, the three-dimensional points in the point cloud data can accurately correspond to the correct pixel positions in the image.

[0058] Step S202: Based on the first pose information and the second pose information, determine the temporal correspondence and spatial transformation relationship between the point cloud data and the target image. The temporal correspondence refers to the matching or mapping relationship between the point cloud data and the target image at the time of acquisition. In this embodiment, it can be understood as determining which point cloud datasets are most temporally close to or correspond to which images by comparing the timestamp in the first pose information of the point cloud data and the timestamp in the second pose information of the target image. In practical applications, the temporal correspondence may take various forms. In one feasible implementation, the temporal correspondence may be a one-to-one correspondence, meaning each frame of point cloud data precisely matches one frame of image data; this typically occurs when two sensors acquire data synchronously. In another feasible implementation, due to differences in sensor acquisition frequencies, the temporal correspondence may also be a one-to-many or many-to-one approximate correspondence; for example, one frame of image data may correspond to multiple frames of point cloud data, or multiple frames of image data may correspond to one frame of point cloud data. Furthermore, in another feasible implementation, when processing the situation between two acquisition time points, the time correspondence needs to establish a continuous mapping relationship through time interpolation. By interpolating the pose information at adjacent time points, the sensor state at a specific time is estimated. The time correspondence is used to solve the problem of asynchronous acquisition time between point cloud data and image data. By establishing an accurate time correspondence, it can be ensured that the scene state captured at the same time or a sufficiently close time is being processed, effectively reducing the impact of scene changes (such as swaying aquatic plants, water surface ripples, etc.) caused by time differences on data fusion. Correspondingly, the spatial transformation relationship refers to the three-dimensional spatial coordinate system transformation rules between two different data sources: point cloud data and target image. In the embodiments of this invention, it can be understood as the mapping relationship from the LiDAR coordinate system to the camera coordinate system, usually represented as a set of mathematical transformation parameters, including rotation matrices and translation vectors, describing how to transform three-dimensional points in the LiDAR coordinate system to the camera coordinate system, or how to back-project two-dimensional image pixels captured by the camera into three-dimensional space. The spatial transformation relationship comprehensively considers the relative installation positions and orientation differences of the two sensors on the mapping platform, as well as their respective internal parameters (such as camera focal length and LiDAR scanning angle). Through accurate spatial transformation, the 3D points in the point cloud data can be correctly projected onto the corresponding pixel positions in the image, enabling precise spatial alignment of the two data sets.

[0059] Step S203: Based on the temporal correspondence and the spatial transformation relationship, perform spatiotemporal registration of the point cloud data and the target image. After determining the temporal correspondence and spatial transformation relationship, data processing is required based on these relationships to achieve precise alignment between the point cloud data and the target image. Spatiotemporal registration refers to the process of unifying and aligning heterogeneous data collected by different sensors at different times and locations in both time and space dimensions. In this embodiment of the invention, it can be understood as a technical method that uses established temporal correspondence and spatial transformation relationships to match point cloud data and target images temporally and align them spatially, enabling the two types of data to be fused and analyzed within a unified spatiotemporal reference framework.

[0060] Specifically, firstly, based on the established temporal correspondence, time-corresponding point cloud data and target image pairs are selected from all collected data. Then, for each time-matched data pair, spatial transformation is applied for spatial registration. During spatial registration, each 3D point in the point cloud data is transformed from the LiDAR coordinate system to the camera coordinate system using the calculated rotation matrix and translation vector. Then, the 3D point is projected onto the image plane using the camera's internal parameters (such as focal length and optical center position), obtaining the corresponding pixel coordinates of the 3D point in the image. Through this method, each 3D point in the point cloud data can find a precise corresponding position in the target image, achieving spatial alignment between the point cloud data and the image data.

[0061] However, for dynamically changing aquatic plant environments, simply considering basic spatiotemporal correspondence may not be accurate enough. Even if the closest data pair is found in time, a slight time difference between the two can still lead to minor spatial deviations due to the swaying of aquatic plants with water flow or the movement of the mapping platform. To address this issue, this invention introduces a time interpolation mechanism during spatiotemporal registration. Based on pose information at consecutive time points, the sensor pose at a specific moment is estimated using interpolation or extrapolation methods, thereby deriving a more accurate spatial transformation relationship. Furthermore, for data collected during the movement of the mapping platform, attitude change information provided by the inertial measurement unit can be used to perform motion compensation on point cloud data and image data, eliminating blur and distortion caused by platform movement and ensuring registration accuracy.

[0062] The final result of spatiotemporal registration is a series of precisely aligned point cloud and image data pairs. The 3D points in the point cloud data can be accurately projected onto the corresponding pixel positions in the target image, while each pixel in the target image can be mapped back to a point position or region in 3D space. Spatiotemporal alignment enables the system to effectively combine the rich visual features provided by the image (such as color, texture, and edges) with the geometric information provided by the point cloud data (such as height, density, and reflection intensity), thereby more accurately distinguishing between aquatic plant reflection points and actual water surface points during subsequent aquatic plant area recognition.

[0063] For example, let the first pose information of the lidar at time t be:

[0064] P L (t)=[X L (t), Y L (t), Z L (t), roll L (t), pitch L (t), yaw L (t)]

[0065] Correspondingly, let the second pose information of the visual camera at time t be:

[0066] P I (t)=[X I (t), Y I (t), Z I (t), roll I (t), pitch I (t), yaw I (t)];

[0067] To achieve time synchronization between the LiDAR and the vision camera, a time synchronization threshold Δt is set. When the LiDAR point P... i Collection time With image frame acquisition time The two are considered to correspond in time if the following relationship is satisfied: in, This represents the acquisition time of the i-th point cloud point. This represents the acquisition time of the j-th image;

[0068] Transform points in the lidar point cloud coordinate system to the image coordinate system or a unified world coordinate system.

[0069] LiDAR point P i The original coordinates are:

[0070] Based on the attitude angles (roll, pitch, yaw) of the lidar at time t, construct the rotation matrix:

[0071] R L (t)=R z (yaw L )·R y (pitch L )·Rx(roll L ); where R z R y R x These are the rotation matrices about the x, y, and z axes, respectively;

[0072] The translation vector is:

[0073] Point cloud After transformation, it is mapped to the image coordinate system (or world coordinate system):

[0074] Among them, R L→I (t) represents the rotation matrix for transforming the lidar into the image coordinate system, T L→I (t) represents the translation vector between coordinate systems; if a unified world coordinate system is used, both can be transformed to the world coordinate system before registration.

[0075] If the registered 3D points need to be projected onto the image plane, the vision camera intrinsic parameter matrix must be used: 3D points Projection to image pixel coordinates:

[0076] Among them, (u i v i ) represents the pixel position of the point cloud in the image; f x f y c represents the camera's focal length. x c y This represents the coordinates of the principal point in the image.

[0077] By employing the embodiments of the present invention, spatiotemporal registration can fully leverage the complementary advantages of the two types of data: on the one hand, the unique visual features of aquatic plants in the image are used to assist in identifying three-dimensional points in the point cloud data that may belong to aquatic plants; on the other hand, the height distribution and density characteristics in the point cloud data are used to help determine the spatial range and specific location of the aquatic plant area in the image.

[0078] Step S102: Based on the target image, determine the first and second anomalous objects in the point cloud data. The first anomalous object represents interference points caused by reflections from foreign objects in the target area, and the second anomalous object represents the interference area formed by the foreign objects. The first density of the first anomalous object is less than the second density of the second anomalous object. The first anomalous object refers to interference points caused by reflections from foreign objects in the target area. In this embodiment, it can be understood as a low-density, dispersed noise point cloud formed by discrete reflections from the surface of aquatic plants when the lidar scans a lake wetland environment with dense aquatic plants. The first anomalous object typically appears as spatially independent, discontinuously distributed anomalous elevation points, characterized by low local point density and significant elevation differences from the surrounding real water surface point cloud. These are used to identify and directly remove anomalous objects during the mapping data processing process to improve the purity and accuracy of the point cloud data, ensuring that the subsequently generated water surface topography model can more accurately reflect the real water surface characteristics.

[0079] For example, in a shallow lake surveying project within an ecological reserve, lidar and high-definition cameras were mounted on an unmanned surface vessel for surveying operations. The lake area is home to different types and densities of aquatic plants: sparse reeds line the shore, large areas of lotus leaves cover parts of the lake, and there are also scattered floating duckweed and fragments of aquatic vegetation.

[0080] When a lidar scanner passes through a sparse reed area, some laser beams are reflected by the reed stems and leaves protruding from the water surface, forming a small number of discrete abnormal elevation points, which are the first anomalous objects. These first anomalous objects are sparsely distributed in space and have no obvious spatial continuity with the surrounding water surface points. In data processing, the system calculates the local point density, identifies these points whose density is significantly lower than a preset threshold, classifies them as first anomalous objects, and directly removes them in subsequent processing. Similarly, the second anomalous object refers to the interference area formed by foreign objects. In this embodiment of the invention, it can be understood as a set of high-density abnormal reflection points caused by a large area of ​​concentrated aquatic plants when the lidar scans a lake wetland environment with dense aquatic plants. The second anomalous object forms a continuous, high-density point cloud in space, characterized by a local point density significantly higher than the normal water surface reflection area, and the elevation values ​​within the area show a certain continuity or similarity, typically corresponding to clearly visible aquatic plant-covered areas in the image. The second type of anomaly is used to identify and perform special regional processing during data processing, rather than simply removing it, in order to preserve the topographic features of the area, ensure the integrity and continuity of the surveying results, and eliminate the interference of aquatic plants on the actual water surface measurement.

[0081] For example, when a lidar scanner passes through an area densely covered with lotus leaves, a large number of laser beams are reflected by the lotus leaf surface, forming a continuous high-density anomalous point cloud. This anomalous point cloud forms a continuously distributed cluster of points in space, with significant spatial correlations between the points within it, collectively describing the surface morphology of the lotus leaf-covered area. Calculations revealed that the point density in these areas was significantly higher than a preset threshold, classifying them as a second type of anomalous object.

[0082] The first and second anomalous objects exhibit clear distinctions and intrinsic connections in the mapping of densely aquatic plant-covered lake wetlands. Their core difference lies in their spatial distribution density: the first density of the first anomalous object is significantly lower than the second density of the second anomalous object. This density difference not only serves as the basis for identifying the two types of anomalous objects but also determines the different strategies to be adopted in actual mapping data processing. From a causal perspective, both usually originate from the same type of foreign matter, but reflect different distribution patterns of aquatic plants in lake wetlands. The first type of anomalous object is mainly caused by sparse distribution areas or edges of aquatic plants, such as scattered stems and leaves protruding from the water surface, the edge areas of aquatic plant communities, or fragments of aquatic plants temporarily floating on the water surface under the action of wind and waves; while the second type of anomalous object is formed by large-area concentrated aquatic plant communities, such as patches of emergent plants, floating-leaved plant-covered areas, or densely submerged plant-covered areas. From a spatial perspective, the first type of anomalous object appears as discrete and independent anomalous points in the point cloud, with no obvious spatial correlation between the points; while the second type of anomalous object forms contiguous point cloud clusters with a certain internal structure, with obvious spatial correlation between the points.

[0083] Based on the above embodiments, as an optional embodiment, step S102 may further include the following steps:

[0084] Step S301: Perform color space conversion on the target image to obtain a color image.

[0085] Target images are typically acquired and stored using the standard RGB (red, green, blue) color space because it matches the native acquisition mode of most camera sensors, facilitating data transmission and storage. However, the RGB color space has a significant limitation: the three color channels are highly correlated and greatly affected by lighting conditions. This makes it easy to be disturbed by changes in external lighting conditions when directly identifying aquatic plants based on the RGB color space in complex lighting environments (such as water reflections and shadow changes), leading to unstable recognition accuracy.

[0086] To overcome the limitations of the RGB color space in aquatic plant recognition, this embodiment converts the original RGB color space target image into a color space more suitable for aquatic plant feature extraction, resulting in a color image. In this embodiment, the color space conversion mainly considers converting the RGB color space to the HSV (Hue, Saturation, Lightness) color space or the Lab color space.

[0087] Because the HSV color space separates color information (hue H and saturation S) from brightness information (brightness value V), color feature extraction is less affected by changes in lighting conditions. In the HSV space, the green characteristics of aquatic plants are mainly reflected within a specific hue range and typically exhibit relatively stable saturation characteristics. Even under different lighting conditions, the hue and saturation characteristics of aquatic plants remain relatively stable, enabling the system to more robustly identify aquatic plant areas and reducing interference from changes in ambient lighting.

[0088] In another feasible implementation, the Lab color space is also a viable alternative. The Lab color space is a perceptually uniform color space, where the L channel represents brightness, the a channel represents the transition from green to red, and the b channel represents the transition from blue to yellow. In the Lab color space, the green characteristics of aquatic plants are mainly reflected in the negative values ​​of the a channel. This characteristic makes aquatic plant areas significantly distinguishable in the a channel, which helps improve the recognition accuracy of aquatic plant areas.

[0089] Specifically, the original RGB target image is first loaded, and then the RGB color values ​​are mapped to their corresponding values ​​in the target color space (such as HSV or Lab) through mathematical transformations. For HSV color space conversion, the hue angle (0-360 degrees), saturation (0-100%), and lightness value (0-100%) of each pixel are calculated; for Lab color space conversion, the lightness (0-100) and chromaticity parameters a (green-red) and b (blue-yellow) of each pixel are calculated. The conversion process involves non-linear mathematical transformations, but modern image processing libraries (such as OpenCV) provide efficient color space conversion functions.

[0090] Step S302: Determine candidate interference regions in the target image based on preset color threshold ranges and texture thresholds. After obtaining the color image, it is necessary to further identify regions in the image that may contain foreign objects such as aquatic plants. To this end, this embodiment adopts a method based on dual feature analysis of color and texture, and accurately filters out regions that may interfere with the point cloud data through predefined threshold conditions.

[0091] First, the preset color threshold range refers to the set of numerical ranges used to identify aquatic plant features in a specific color space. In this embodiment of the invention, it can be understood as the value range of each channel in the HSV color space predefined by the system, which is used to determine whether an image pixel belongs to an aquatic plant region.

[0092] Specifically, color threshold ranges typically include hue threshold range, saturation threshold range, and brightness threshold range. Each pixel is evaluated, and only when the values ​​of that pixel in all three channels simultaneously meet the corresponding range requirements is it preliminarily identified as a pixel in the aquatic plant region. This multi-channel comprehensive judgment effectively improves the accuracy of color screening. However, relying solely on color features is insufficient to completely distinguish aquatic plants from other green objects; therefore, texture feature analysis is also introduced. Texture threshold refers to the judgment criteria of statistical quantities describing the structural features of local regions of an image. In this embodiment of the invention, it can be understood as the judgment criteria based on feature values ​​calculated using texture analysis methods such as gray-level co-occurrence matrix or local binary pattern, including multiple dimensions of judgment conditions such as contrast threshold, uniformity threshold, entropy threshold, and energy threshold. These thresholds collectively describe the unique texture features of the aquatic plant region, such as the edge sharpness of the aquatic plant leaves, texture complexity, and directionality. Through texture analysis, the regions after initial color screening can be further verified, effectively eliminating non-aquatic plant regions that are similar in color but do not match the texture features, thereby significantly improving the accuracy and reliability of aquatic plant region identification.

[0093] It is worth noting that the preset color threshold range and texture threshold are not arbitrarily set, but determined through scientific methods. The method for determining these thresholds combines three strategies: prior knowledge, data-driven approaches, and adaptive adjustment, ensuring the scientific validity and effectiveness of the threshold settings.

[0094] First, based on prior knowledge of aquatic botany, a database of color and texture features of different types of aquatic plants under various lighting conditions was established. By sampling and analyzing the spectral characteristics of common aquatic plant species (such as *Potamogeton crispus*, *Hydrilla verticillata*, *Ceratophyllum demersum*, and *Water lily*) at different growth stages, seasons, and lighting conditions, a preliminary color threshold range was established. Simultaneously, through texture feature extraction and statistical analysis of these aquatic plant sample images, the characteristic distribution range of typical aquatic plant textures was determined. This initial setting based on professional knowledge provided a scientific foundation for subsequent optimization. Furthermore, a data-driven approach was used to optimize the initial thresholds. Before the commencement of a specific lake mapping project, sample images of aquatic plants and non-aquatic plant areas within that lake environment were collected to construct a local training dataset. By statistically analyzing the color and texture features of these sample images, the optimal threshold parameters that best distinguish between aquatic plants and non-aquatic plant areas under the current environment could be calculated. This data-driven approach allows the threshold setting to adapt to the characteristics of a specific lake environment, significantly improving the targeting and accuracy of identification.

[0095] In addition, in practical applications, preset color threshold ranges are often customized for different aquatic plant types and aquatic environments. For example, for lakes dominated by floating-leaved plants, a narrower hue range and a higher lower saturation limit may be set; while for lakes dominated by submerged plants, considering the absorption and scattering effects of light by the water, a wider hue range and a lower saturation limit may be set. Similarly, texture thresholds are also adjusted according to the morphological characteristics of the aquatic plants. For example, for water lilies with large leaves and regular textures, uniformity and energy characteristics may be emphasized; while for slender leaves and complex textures of *Hydrilla verticillata*, contrast and entropy characteristics may be emphasized. This targeted parameter adjustment further improves the ability to adapt to different environments.

[0096] Through the aforementioned dual feature analysis of color and texture, candidate interference regions in the target image are finally identified. Candidate interference regions refer to image areas in the target image that are preliminarily determined to potentially contain foreign objects such as aquatic plants based on color and texture feature analysis. In this embodiment of the invention, this can be understood as the set of pixels marked as potential aquatic plant-covered areas in the image space after color threshold screening and texture feature verification. These areas visually exhibit color and texture features that are significantly different from the surrounding water surface, typically displaying a specific range of green hues and a texture pattern unique to aquatic plants.

[0097] Step S303 involves mapping the point cloud data to the color space corresponding to the target image to obtain the mapped point cloud. As shown in steps S201 to S203 above, the system has established the temporal correspondence and spatial transformation relationship between the point cloud data and the target image, which provides the necessary support for the mapping operation in step S303. The mapping process requires utilizing the known spatial transformation relationship to transform the point cloud data in the LiDAR coordinate system to the camera coordinate system, and then projecting the three-dimensional points onto the image plane through the camera intrinsic parameters to achieve accurate mapping from three-dimensional space to a two-dimensional image.

[0098] Specifically, based on the sensor pose information obtained in step S201 and the time correspondence determined in step S202, the image frame that is temporally closest for each frame of point cloud data is found. For cases with large time differences, interpolation methods are needed to estimate the camera pose at a specific moment to ensure the accuracy of the mapping. After determining the matching data pairs, a spatial transformation relationship is further applied to transform each point cloud point from the LiDAR coordinate system to the camera coordinate system, and then the 3D points are projected onto the image plane using the camera's projection model.

[0099] It is worth noting that in complex, densely vegetationed lake environments, the mapping between point cloud data and images not only needs to consider geometric transformations but also needs to handle issues such as occlusion and reflection. For example, when multiple point cloud points from a densely vegetationed area are projected onto the same location in the image, depth comparison is needed to determine which point is closer to the camera and whose color information should be obtained. Simultaneously, point cloud points that may fall outside the image boundaries can be marked as having no color information to avoid introducing erroneous data. Furthermore, this embodiment of the invention also considers potential distortion problems during projection, correcting the projection results using camera calibration parameters to ensure mapping accuracy.

[0100] The above processing yields a mapped point cloud, where each point in the cloud is associated with the color information of a corresponding image pixel. In this embodiment of the invention, the mapped point cloud not only includes attributes such as the three-dimensional coordinates and reflection intensity of the original point cloud, but also adds color information obtained from the target image, such as the three channel values ​​of the HSV color space. This enhanced point cloud data, which integrates geometric and visual information, provides support for subsequent anomaly object recognition based on multi-source information.

[0101] Step S304: Based on the density space of the mapped point cloud within the candidate interference region, the mapped point cloud is divided into the first anomalous object and the second anomalous object. The density space of the mapped point cloud within the candidate interference region refers to a mathematical representation describing the degree of spatial concentration of the 3D point cloud data within the identified candidate interference region. In this embodiment, it can be understood as a 3D density distribution feature field formed after performing local density calculations on the point cloud data projected onto the candidate interference region. This density space reflects the ratio of the number of points in a specific radius neighborhood around each point cloud point to the spatial volume, quantifying the degree of aggregation and distribution pattern of the point cloud within the aquatic plant area. The density space serves as the characteristic basis for distinguishing between the first and second anomalous objects. By setting an appropriate density threshold, points with low density distribution can be marked as the first anomalous object (discrete interference points), and points with high density aggregation can be marked as the second anomalous object (dense interference region), thereby achieving the classification of different types of aquatic plant reflection anomalies.

[0102] For example, the target image is converted from RGB color space to HSV color space; where, let the target image be: I RGB (x, y) = [R(x, y), G(x, y), B(x, y)]; Convert to HSV space: I HSV (x, y) = [H(x, y), S(x, y), V(x, y)]; The threshold range for aquatic plant color in the HSV space is defined as: H ∈ [H... min H max ], S∈[S min S max ], V∈[Vmin V max The candidate interference region is defined as: R i ={(x,y)|Hmin≤H(x,y)≤Hmax,Smin≤S(x,y)≤Smax,Vmin≤V(x,y)≤Vmax}; Use the Gray-Level Co-occurrence Matrix (GLCM) to extract texture contrast features: C(x,y)=∑ i，j (ij) 2 P(i, j); sets the texture threshold τ tex The final interference region is: R′ i ={(x, y)∈R} i |C(x,y)≥τ tex Let the point cloud dataset be: Using the known spatial transformation relationship F between sensors map Mapping point clouds from 3D space to the image plane: (u i v i ) = F map (X i Y i Z i ).

[0103] Based on the above embodiments, as an optional embodiment, step S304 may further include the following steps:

[0104] Step S401: Using any point in the mapped point cloud as the center, determine a spherical neighborhood with a preset radius. Specifically, using each point in the mapped point cloud as the center, determine a spherical local neighborhood. The radius of the spherical neighborhood is a preset value, and the selection of the spherical neighborhood radius needs to consider several factors: the radius should match the physical scale of the aquatic plants, usually set to 1-2 times the average width of the aquatic plant stems and leaves, to ensure that the neighborhood can contain enough relevant points without over-expansion; the radius also needs to consider the sampling density of the point cloud, and may need to be appropriately increased in sparse point cloud regions; the radius should also consider computational efficiency, as an excessively large radius will lead to a sharp increase in computational load.

[0105] Step S402: Calculate the spatial density of the center within the spherical neighborhood. Specifically, spatial density refers to the number of point cloud points contained within a unit volume of space, directly reflecting the concentration of the point cloud in a local region. A density value is obtained by counting the number of point cloud points contained within the spherical neighborhood and dividing by the volume of the spherical neighborhood. To improve computational efficiency, spatial index structures such as kd-trees or octrees are typically used for quick lookup of points within the neighborhood. Furthermore, to reduce the influence of boundary effects, a weighted density calculation method may be used, assigning different weights to neighborhood points at different distances from the center point, making the density calculation more accurately reflect the local distribution characteristics.

[0106] Step S403: If the spatial density is greater than or equal to the first preset density threshold, the mapped point cloud is identified as the first abnormal object.

[0107] Step S404: If the spatial density is less than or equal to a second preset density threshold, the mapped point cloud is identified as a second anomalous object. The first preset density threshold refers to a spatial density threshold used to determine whether the mapped point cloud belongs to the first anomalous object. In this embodiment, it can be understood as an upper limit standard for the local aggregation degree of the point cloud predetermined based on the distribution characteristics of aquatic plants. When the local density of the point cloud is greater than or equal to this threshold, it indicates that the point cloud in that area exhibits a highly aggregated state, conforming to the characteristics of a densely aquatic plant area. The second preset density threshold refers to a spatial density threshold used to determine whether the mapped point cloud belongs to the second anomalous object. In this embodiment, it can be understood as a lower limit standard for the local aggregation degree of the point cloud predetermined based on the distribution characteristics of aquatic plants. When the local density of the point cloud is less than or equal to this threshold, it indicates that the point cloud in that area exhibits a relatively dispersed state, conforming to the characteristics of scattered aquatic plant distribution or single aquatic plant.

[0108] Specifically, when the spatial density of the center point is greater than or equal to a first preset density threshold, the point is classified as a first anomalous object; when the spatial density is less than or equal to a second preset density threshold, the point is classified as a second anomalous object. The setting of these two density thresholds is also based on a data-driven approach, and the optimal threshold parameters can be determined by analyzing the density distribution characteristics of multiple sample regions or by combining manually labeled training data.

[0109] It is important to note that the relationship and specific values ​​of the first and second preset density thresholds need to be flexibly adjusted according to the actual application scenario. In the standard configuration, the first preset density threshold is usually greater than the second preset density threshold, forming a decision interval. When the density value of a point falls between these two thresholds, it may be necessary to consider more features (such as the point's color information, the classification results of surrounding points, etc.) for a comprehensive judgment, or to use a fuzzy classification method to assign a probability distribution to the point belonging to each category, rather than a simple binary classification result.

[0110] For example, for each candidate point p i Perform the following operations: Define a spherical neighborhood N(p) with radius r centered at this point. i ,r):N(p i ,r)={p j ∈P candidate |||p j -p i ||≤r};where ||p j -p i || represents point p j Point Pi Euclidean distance;

[0111] Spatial density, defined at point p i Local spatial density ρ(p) i The number of points in the neighborhood is divided by the volume of the sphere.

[0112] Where, |N(p i r)| is the point p i The number of points in the neighborhood; V spher Let be the volume of the spherical neighborhood.

[0113] Two density thresholds are set: the first density threshold is ρ1; the second density threshold is ρ2, and ρ1 < ρ2 is satisfied.

[0114] Based on the local density of a point, the following determination is performed: if ρ(p i If )≤ρ1, then this point is determined as the first abnormal object: p i ∈O1; if ρ(p i If )≥ρ2, then this point is determined as the second abnormal object: p i ∈O2; To further verify the rationality of the classification, the average density of the two objects can be calculated and verified:

[0115]

[0116] Where |O| represents the number of points in the object; V total Indicates the volume of space occupied by an object.

[0117] Step S103: Generate measurement data based on the first abnormal object and the second abnormal object.

[0118] Based on the above embodiments, as an optional embodiment, step S103 may further include the following steps:

[0119] In the point cloud data, the point cloud data corresponding to the first anomalous object is removed, and the point cloud data corresponding to the second anomalous object is reconstructed to obtain the measurement data. Specifically, the first anomalous object usually corresponds to a densely planted area with a high point cloud distribution density, occupying a large continuous spatial area. Simply deleting these points will leave large data holes in the measurement data, affecting the integrity and accuracy of the terrain model. The second anomalous object corresponds to a sparsely planted area with a relatively dispersed point cloud distribution and a limited impact range, which can be handled by a relatively simple method. Differentiated processing strategies are adopted for the different characteristics of these two types of anomalous objects. For the first anomalous object, a regional reconstruction method can be used. First, non-anomalous points around the first anomalous object are identified; these points usually represent the true water surface elevation. Then, based on the spatial distribution and elevation values ​​of these non-anomalous points, an appropriate interpolation algorithm is used to reconstruct the water surface elevation within the area occupied by the first anomalous object. This reconstruction process considers the continuity and smoothness of the water surface, and can generate a water surface model that conforms to the laws of hydraulics. In certain special cases, such as when the area of ​​the first anomalous object is too large or the surrounding non-anomaly points are sparse, historical data or hydrological models may be combined to assist in the reconstruction of water surface elevation, ensuring the physical rationality of the generated measurement data. For the second anomalous object, due to its relatively scattered distribution and limited range, local filtering and replacement methods are used. Specifically, the point cloud points identified as the second anomalous object are first deleted. Then, at the location of each deleted point, local interpolation is performed based on the elevation values ​​of the surrounding non-anomaly points to obtain the estimated water surface elevation at that location. This method is relatively simple and efficient, suitable for handling scattered anomalous points, while maintaining the spatial continuity of the measurement data. In addition, multi-source features such as the reflection intensity and color information of the point cloud may be combined to optimize the effect of local filtering, especially in areas where the boundary between aquatic plants and the water surface is blurred. Because the local data generated by different processing methods may have slight discontinuities or inconsistencies, in addition to processing the first and second anomalous objects, overall data smoothing and consistency processing is also required. Weighted averaging or smoothing filtering methods are usually used to eliminate potential boundary effects and discontinuities, generating the final measurement data.

[0120] The following section will provide a detailed explanation of how to handle the first and second abnormal objects, using practical application scenarios as examples.

[0121] In practical applications, when lidar scans water surfaces densely covered with aquatic plants, the laser beam generates numerous redundant reflection signals from the plant surface. This results in a large number of noise points unrelated to the actual water surface topography being mixed into the point cloud data, severely degrading the quality of the original data. Simultaneously, due to the combined effects of wind and water flow, aquatic plants exhibit irregular dynamic swaying characteristics, causing significant differences in data acquired by lidar at different times when scanning the same area. This time-varying characteristic leads to a lack of consistency in continuously acquired data, increasing the complexity of subsequent processing. Related technologies lack effective prediction mechanisms for height anomalies caused by the dynamic swaying of aquatic plants, making real-time compensation for such anomalies impossible. This results in numerous height anomalies caused by aquatic plant swaying in the final mapping data. Regarding anomaly area restoration, traditional methods often employ single interpolation or fitting algorithms, lacking the ability to adaptively select the optimal restoration strategy based on the characteristics of the anomaly area. When facing large areas with particularly dense aquatic plants, the scarcity and uneven distribution of effective data points make it difficult for existing restoration methods to generate smooth and continuous water surface topography, and the restoration quality cannot meet the requirements of high-precision mapping.

[0122] To address the above problems, the measurement method provided in this embodiment of the invention may further include the following steps:

[0123] Step S501: Obtain the motion characteristics of the foreign object. These characteristics characterize the dynamic changes of the foreign object under the influence of wind and water flow. Specifically, the motion characteristics refer to the dynamic changes of disturbances such as aquatic plants under the influence of wind and water flow. In this embodiment, this can be understood as the time-varying characteristics parameters of aquatic plants swaying on or below the water surface with the influence of water flow and wind, such as periodicity, amplitude, and direction. These parameters are used to construct a predictive model for the height changes of aquatic plants, effectively compensating for dynamic anomalies in lidar point cloud data. In densely aquatic plant lake wetland environments, aquatic plants are not static but exhibit complex dynamic swaying under the combined influence of wind and water flow. This dynamic characteristic directly leads to significant differences in elevation data obtained by lidar when scanning the same area at different times, making it difficult for traditional static filtering methods to effectively handle such time-varying anomalies. To solve this problem, it is necessary to first obtain the motion characteristics of the aquatic plants and analyze their dynamic changes.

[0124] Specifically, a multi-time observation and analysis method was employed to obtain the motion characteristics of aquatic plants. First, point cloud data and image data of a specific area were continuously collected at fixed locations to form a time-series observation dataset. Analysis of this time-series data allowed for the extraction of key characteristic parameters of aquatic plant swaying. Regarding the point cloud data, the focus was on analyzing the elevation changes of the same area at different times, calculating the period, amplitude, and phase characteristics of these elevation changes. Through methods such as Fourier analysis or time-series decomposition, regular periodic patterns of elevation fluctuations could be extracted from seemingly chaotic elevation fluctuations. Regarding the image data, computer vision technology was used to analyze the displacement changes of aquatic plant areas in consecutive image frames, aiding in the verification and refinement of the motion characteristics extracted from the point cloud data. Through the fusion analysis of multi-source data, a more comprehensive description of the motion characteristics of aquatic plants could be obtained, including key parameters such as swaying frequency, amplitude, initial phase, and correlation with environmental factors. Significant differences in motion characteristics exist for different types of aquatic plants and different hydrological environments. In one feasible implementation, a database mapping aquatic plant types to their movement characteristics can be established. Typical movement characteristic parameters can be extracted and stored for different types of aquatic plants (such as emergent plants, floating-leaved plants, and submerged plants), plants at different growth stages, and plants in different water depths. In practical applications, the most suitable initial parameters can be selected from the feature database based on the aquatic plant type and environmental conditions of the target area, and dynamically adjusted in conjunction with real-time observation data to ensure the accuracy and applicability of the movement characteristics.

[0125] Step S502: Construct a height change prediction model for foreign objects based on motion characteristics. This model characterizes the change in height of foreign objects over time under the influence of wind and water flow. Specifically, the height change prediction model refers to a mathematical expression describing the change in height of foreign objects such as aquatic plants over time under the influence of wind and water flow. In this embodiment, it can be understood as a simple harmonic motion model based on the dynamic swaying motion characteristics of aquatic plants, capable of predicting height changes of aquatic plants at different times. This model is used to predict and compensate for height anomalies caused by the swaying of aquatic plants during lidar data acquisition, improving the temporal consistency and spatial accuracy of the mapping data.

[0126] Specifically, statistical analysis of aquatic plant height changes over short time series revealed that their oscillation typically exhibits periodic characteristics, approximating a simple harmonic motion pattern. Based on this, a simple harmonic motion prediction model was constructed, using the water surface base height, oscillation amplitude, oscillation frequency, and initial phase as key parameters. This model represents the aquatic plant oscillation height as the superposition of the base height and a periodic variation term, where the periodic variation term is described by a sine function, accurately capturing the periodic characteristics of the aquatic plant oscillation. To determine the model parameters, a nonlinear least squares algorithm was used to fit historical observation data within a short time window. For example, a sequence of height observations for the same aquatic plant area over a period of time can be collected, and then the model parameter values ​​that best fit the observation data can be found. During the optimization process, the objective function is to minimize the sum of squares of the height prediction error, and the parameter values ​​are continuously adjusted through iterative calculations until the convergence condition is met. This data-driven parameter estimation method can determine the most suitable prediction model parameters for the aquatic plant oscillation characteristics of different regions.

[0127] Step S503: Compensate for height anomalies caused by the dynamic swaying of foreign objects in the point cloud data based on the height change prediction model. In densely vegetated lake wetland environments, the dynamic swaying of aquatic plants causes significant differences in the data collected by lidar at different times. Even if abnormal objects have been identified, accurate mapping data reflecting the true topography of the water surface cannot be obtained without compensating for this dynamic change. Therefore, after constructing the height change prediction model, it is necessary to apply the model to actual point cloud data processing to compensate for height anomalies caused by the dynamic swaying of foreign objects such as aquatic plants. For example, within a certain local area, the height of the top of a foreign object at the same spatial location at different time points is observed to obtain an observation dataset: In the formula, t k h represents the time of the kth observation. k Let represent the height value at the k-th observation, and N represent the number of observation time points. Based on actual observations, aquatic plants and other foreign objects typically exhibit periodic oscillations, which can be fitted into a simple harmonic motion model. Let the mathematical model of the change in the height of the aquatic plant top over time be: In the formula, h(t) represents the predicted height at time t; h0 represents the height of the aquatic plant base; A represents the oscillation amplitude; ω represents the oscillation angular frequency; Indicates the initial phase; uses a nonlinear optimization algorithm on the observation data. Perform fitting to find the optimal parameters: The above optimization problem can be solved using common numerical algorithms.

[0128] For point cloud data point p at a certain time t i If its location is in an anomalous area, attempt to compensate using a predictive model. Calculate the swaying height of the aquatic plants at that point using the predictive model: Compensate for the original point cloud height: Update the height value of point pi to the height after removing the effect of aquatic plant swaying.

[0129] Based on the above embodiments, step S503 may further include the following steps:

[0130] Step S601: If abnormal areas still exist after compensating the point cloud data, obtain the spatial scale of the abnormal areas and the valid point cloud data within the abnormal areas. The spatial scale of the abnormal areas refers to the size and extent of the data gaps or abnormal areas that still exist after compensation by the height change prediction model. In this embodiment, it can be understood as the spatial extent characteristics of data anomalies or missing areas that still exist in aquatic plant-dense lake wetland environment after dynamic compensation processing of the point cloud data, including geometric parameters such as area size, width, and length. In the high-precision mapping process of aquatic plant-dense lake wetland environment, although the height change prediction model has effectively compensated for height anomalies caused by the dynamic movement of aquatic plants, due to the extremely uneven distribution of aquatic plants, complex water depth changes, and possible sensor blind spots, data gaps or abnormal areas may still exist in the compensated point cloud data. Specifically, identify the remaining abnormal areas. The identification of abnormal areas is based on multiple indicators, including features such as point cloud density significantly lower than the surrounding area, unreasonable abrupt changes in elevation values ​​compared to the surrounding area, or obvious missing measurement points. Region growing algorithms or density-based spatial clustering methods can be used to aggregate points or regions with similar anomalous characteristics into continuous anomalous regions. Each identified anomalous region is assigned a unique identifier for subsequent individual processing. After identifying the anomalous regions, the spatial scale parameters of each region are calculated. A boundary tracking algorithm is used to determine the perimeter contour of the anomalous region, and then geometric features such as area, perimeter, maximum width, and length are calculated based on the contour. For anomalous regions with complex shapes, advanced geometric features such as the minimum bounding rectangle, convex hull area, and principal axis direction are also calculated to more comprehensively describe their spatial distribution characteristics. Spatial scale parameters directly affect the selection of subsequent restoration strategies: for anomalous regions with small spatial scales, simple local interpolation methods usually achieve good results; while for anomalous regions with large spatial scales, more complex surface fitting or model constraint methods are required to ensure restoration effectiveness. Simultaneously, it is also necessary to identify valid point cloud data within and around the anomalous regions. Valid point cloud data refers to point cloud points that, after the aforementioned steps, are confirmed to be reliable and able to reflect the true characteristics of the water surface; these points will serve as valid support points for subsequent data restoration. By analyzing the consistency of elevation values, spatial relationship with surrounding points, and other quality indicators, highly reliable and effective points are selected.

[0131] Step S602: If the spatial scale of the abnormal region is greater than a preset scale threshold, a surface fitting model is generated based on the effective point cloud data. The surface fitting model is solved to obtain a first height, and the abnormal region is compensated based on the first height. Here, the surface fitting model refers to a functional expression that describes the continuous surface morphology in three-dimensional space using mathematical methods. In this embodiment, it can be understood as a mathematical model established based on the effective point cloud data around the abnormal region, capable of describing the continuous changes in water surface topography, used to predict and reconstruct missing water surface elevation data within the abnormal region, ensuring that the repaired water surface morphology conforms to natural hydrological laws and topographic continuity requirements. Similarly, the first height refers to the predicted elevation value at each location within the abnormal region calculated by the surface fitting model. In this embodiment, it can be understood as the water surface elevation estimate derived from the surface fitting model at a specific coordinate location within the abnormal region, representing the best prediction of the actual water surface height at that location. Specifically, firstly, it is assessed whether the spatial scale of the abnormal region exceeds a preset scale threshold. The preset scale threshold is usually determined based on practical experience and test results, representing the applicable boundary between local interpolation methods and global fitting methods. For large-scale anomaly areas exceeding this threshold, the surface fitting model construction process begins. During this process, a set of points spatially relevant to the anomaly area to be repaired is selected from the acquired valid point cloud data. These points primarily include valid points surrounding the anomaly area, as well as a small number of reliable points potentially existing within the anomaly area. To ensure model stability and accuracy, not only is the spatial distribution of points considered, but also their reliability is evaluated, assigning different weights to points with varying degrees of reliability, thus increasing the influence of highly reliable points on the model. Next, based on the selected valid point set, an appropriate mathematical model is selected for surface fitting. Different model structures are chosen for different types of water surface topography. For relatively gently changing lake surfaces, quadratic or cubic polynomial models may be chosen, as these models effectively capture the slow changing trends of the water surface. For areas with more complex water flow or significant topographic relief, more flexible nonparametric models, such as radial basis functions or thin-plate spline functions, may be selected, as these models can adapt to more complex topographic variations. The model selection process also considers the balance between computational efficiency and fitting accuracy, ensuring that the repair quality is maintained without excessively increasing the computational burden.

[0132] After the model structure is determined, the model parameters are solved using the least squares method or other optimization methods to minimize the difference between the predicted values ​​and actual observations at known effective points. For complex models, an optimization objective function with a regularization term is typically used to avoid overfitting and ensure the model's generalization ability. After solving the parameters of the surface fitting model, the model is applied to each grid point within the anomaly region to calculate the predicted elevation values, i.e., the first height, at these locations. Specifically, a regular sampling grid is generated within the anomaly region, and the planar coordinates of each grid point are substituted into the solved surface fitting model to obtain the predicted elevation value of that point. Finally, the anomaly region is compensated based on the calculated first height. For locations within the anomaly region where data is completely missing, the first height is directly used as the elevation value for that location; for data points within the anomaly region that exist but may be unreliable, a decision is made based on the degree of difference between the original data points and the first height to determine whether to completely replace them or perform weighted fusion.

[0133] For example, suppose the set of points that still exhibit high anomalies or are missing after dynamic compensation is: A = {p i =(x i y i NaN, t i ), i = 1, 2, ..., N a Where NaN indicates that the point is highly missing or abnormal, and N... a Indicate the number of outliers; calculate the two-dimensional boundary envelope B of the outlier set A, and define the spatial scale S as its diagonal length:

[0134]

[0135] Where, x max x min Represents the maximum / minimum X-coordinate in the set of outliers; y max y min This represents the maximum / minimum Y-coordinate in the set of outliers.

[0136] Set a preset scale threshold. If S is greater than the preset scale threshold, perform surface fitting; otherwise, perform interpolation. Obtain the effective point set in the neighborhood. Extract the effective point cloud set near the boundary of the anomaly region A. The quadratic surface fitting model is constructed as follows:

[0137] h(x, y) = a + bx + cy + dx 2 +exy+fy 2 ;

[0138] The objective function to fit is to minimize the sum of squared residuals:

[0139]

[0140] Solving for the parameter vector using the least squares method:

[0141] in, Each action:

[0142] Calculate the compensation height for each outlier point:

[0143] By applying surface fitting models, the problem of restoring large-scale anomaly areas can be effectively solved, generating water surface models that conform to hydrological patterns and topographic continuity. Compared with simple local interpolation, this global morphology-based restoration method can better maintain the overall shape and trend of the water surface, significantly improving the geometric accuracy and visual naturalness of the restored area.

[0144] Step S603: If the spatial scale of the abnormal region is less than or equal to a preset scale threshold, then the valid point cloud data is interpolated to obtain a second height, and the abnormal region is compensated based on the second height. Here, the second height refers to the estimated elevation value obtained by interpolating the valid point cloud data around the abnormal region. In this embodiment of the invention, this can be understood as using local interpolation algorithms such as inverse distance weighted interpolation to calculate the estimated water surface elevation value of each abnormal point or missing point location based on the valid point cloud data around the abnormal region.

[0145] During the interpolation process, the first step is to select a set of valid points from the effective point cloud data that are spatially closest to the current anomaly region to be processed. This selection process considers not only the Euclidean distance between the points and the anomaly region but also the uniformity of point distribution, ensuring that the selected points provide effective information support for the anomaly region from all directions. Simultaneously, the reliability and representativeness of these valid points need to be evaluated, such as checking for abrupt changes in elevation values ​​and whether they form a reasonable water surface trend with surrounding points, eliminating potentially problematic points to ensure the quality of the underlying data used for interpolation. After determining the valid points for interpolation, an inverse distance weighted interpolation algorithm is used to calculate the second elevation at each location within the anomaly region. The elevation value at a location within the anomaly region should be correlated with the elevation values ​​of surrounding known valid points, and this correlation weakens with increasing distance.

[0146] Specifically, a weight can be assigned to each valid point, which is inversely proportional to the distance from that point to the interpolation location, typically a negative power function of the distance. Thus, points closer to the interpolation location have a greater impact on the interpolation result, while points farther away have a smaller impact. The second height estimate of the interpolation location is calculated by weighted averaging of all valid points. In the actual implementation of inverse distance weighted interpolation, various optimization strategies are considered. For example, to avoid excessive influence from a single point, a search radius might be set, considering only valid points within that radius; or directional balance might be introduced to ensure that points in different directions can influence the interpolation result, preventing points in one direction from excessively dominating the result. After calculating the second height, compensation is applied to outlier regions based on these height values. This involves filling in locations with completely missing data in the outlier regions by creating new point cloud points with their elevations set to the calculated second height; and correcting existing points in the outlier regions with abnormal elevations by adjusting their elevations to the calculated second height. Through compensation processing, data holes and elevation anomalies in small-scale abnormal areas can be effectively eliminated, restoring the continuity and smoothness of the water surface topography.

[0147] For example, for each outlier point belonging to the outlier region, find the set of valid points within a radius r:

[0148] N(x i y i )={(x k y k , z k )|(x k y k )-(x i y i )≤r};

[0149] Using inverse distance weighted interpolation, let The interpolation height is then: Where m represents the distance weight index, which is usually taken as 2, and n represents the number of valid points in the neighborhood.

[0150] The aforementioned local interpolation-based restoration method is particularly effective for small-scale anomaly areas because the water surface topography within small areas typically changes relatively gently, and local interpolation can accurately capture this trend. Furthermore, compared to complex surface fitting, local interpolation significantly reduces computational cost and offers faster processing speed, which is especially important for surveying applications requiring real-time or near-real-time processing.

[0151] Figure 2The schematic diagram illustrates a structural block diagram of a measuring device provided by the present invention, which may include: a data acquisition module for acquiring point cloud data of a target area and a target image; an anomaly determination module for determining a first anomaly object and a second anomaly object in the point cloud data based on the target image; and a data generation module for generating measurement data based on the first anomaly object and the second anomaly object: wherein the first anomaly object represents an interference point caused by reflection of a foreign object in the target area, the second anomaly object represents an interference area formed by the foreign object, and the first density of the first anomaly object is less than the second density of the second anomaly object.

[0152] Based on the above embodiments, as an optional embodiment, the anomaly determination module is further configured to perform color space conversion on the target image to obtain a color image; determine candidate interference regions in the target image according to a preset color threshold range and texture threshold; map the point cloud data to the color space corresponding to the target image to obtain a mapped point cloud; and divide the mapped point cloud into the first anomaly object and the second anomaly object according to the density space of the mapped point cloud in the candidate interference region.

[0153] Based on the above embodiments, as an optional embodiment, the anomaly determination module is further configured to determine a spherical neighborhood with a preset radius centered on any point in the mapped point cloud; calculate the spatial density of the center in the spherical neighborhood; determine the mapped point cloud as the first abnormal object if the spatial density is greater than or equal to a first preset density threshold; and determine the mapped point cloud as the second abnormal object if the spatial density is less than or equal to a second preset density threshold.

[0154] Based on the above embodiments, as an optional embodiment, the measuring device further includes a spatiotemporal registration module, used to obtain first pose information and second pose information, wherein the first pose information and the second pose information characterize the spatial position and attitude angle of the target device when acquiring point cloud data and target image of the target area; based on the first pose information and the second pose information, determine the temporal correspondence and spatial transformation relationship between the point cloud data and the target image; and based on the temporal correspondence and the spatial transformation relationship, perform spatiotemporal registration on the point cloud data and the target image.

[0155] Based on the above embodiments, as an optional embodiment, the measuring device further includes an anomaly compensation module, used to obtain the motion characteristics of the foreign object, the motion characteristics representing the dynamic change pattern of the foreign object under the action of wind and water flow; construct a height change prediction model of the foreign object based on the motion characteristics; and compensate for the height anomalies caused by the dynamic swinging of the foreign object in the point cloud data based on the height change prediction model.

[0156] Based on the above embodiments, as an optional embodiment, the anomaly compensation module is further configured to, when anomaly regions still exist after compensation of the point cloud data, obtain the spatial scale of the anomaly region and the valid point cloud data within the anomaly region; if the spatial scale of the anomaly region is greater than a preset scale threshold, then a surface fitting model is generated based on the valid point cloud data, the surface fitting model is solved to obtain a first height, and the anomaly region is compensated based on the first height; if the spatial scale of the anomaly region is less than or equal to the preset scale threshold, then the valid point cloud data is interpolated to obtain a second height, and the anomaly region is compensated based on the second height. Based on the above embodiments, as an optional embodiment, the data generation module is further configured to, from the point cloud data, remove the point cloud data corresponding to the first anomaly object and reconstruct the point cloud data corresponding to the second anomaly object to obtain measurement data.

[0157] Figure 3 The diagram illustrates a structural block diagram of an electronic device suitable for implementing the methods described above, according to an embodiment of the present invention. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality or scope of the embodiments of the present invention. Figure 3As shown, the electronic device 300 includes a processor 301, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 302 or a program loaded from storage portion 308 into random access memory (RAM) 303. The processor 301 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 301 may also include onboard memory for caching purposes. The processor 301 may include a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of the present invention. Various programs and data required for the operation of the electronic device 300 are stored in RAM 303. The processor 301, ROM 302, and RAM 303 are interconnected via bus 304. The processor 301 performs various operations of the method flow according to embodiments of the present invention by executing programs in ROM 302 and / or RAM 303. It should be noted that the programs may also be stored in one or more memories other than ROM 302 and RAM 303. The processor 301 can also perform various operations of the method flow according to embodiments of the present invention by executing programs stored in the one or more memories.

[0158] According to an embodiment of the present invention, the electronic device 300 may further include an input / output (I / O) interface 305, which is also connected to a bus 304. The system 300 may further include one or more of the following components connected to the input / output (I / O) interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the input / output (I / O) interface 305 as needed. A removable medium 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 310 as needed so that computer programs read from it can be installed into the storage section 308 as needed.

[0159] According to embodiments of the present invention, the method flow according to embodiments of the present invention can be implemented as a computer software program. For example, embodiments of the present invention 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 method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by processor 301, it performs the functions defined in the system of the embodiments of the present invention. According to embodiments of the present invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0160] The present invention 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, which, when executed, implement the method according to the embodiments of the present invention.

[0161] According to embodiments of the present invention, a 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 invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include one or more memories other than ROM 302 and / or RAM 303 described above.

[0162] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of the present invention. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the methods provided in the embodiments of the present invention.

[0163] When the computer program is executed by processor 301, it performs the functions defined in the system / device of this embodiment of the invention. According to embodiments of the invention, the systems, devices, modules, units, etc., described above can be implemented by computer program modules. In one embodiment, the computer program can rely on tangible storage media such as optical storage devices or magnetic storage devices. In another embodiment, the computer program can also be transmitted and distributed in the form of signals over a network medium, and can be downloaded and installed via communication section 309, and / or installed from removable medium 311. 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. According to embodiments of the invention, program code for executing the computer program provided in the embodiments of the invention 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 the user's computing device, partially on the 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 devices 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 they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

Claims

1. A measurement method, characterized in that, include: Obtain point cloud data and target image of the target area; Based on the target image, determine the first and second anomalous objects in the point cloud data; The step of determining the first and second anomalous objects in the point cloud data based on the target image includes: performing color space conversion on the target image to obtain a color image; determining candidate interference regions in the target image according to a preset color threshold range and texture threshold; mapping the point cloud data to the color space corresponding to the target image to obtain a mapped point cloud; and dividing the mapped point cloud into the first anomalous object and the second anomalous object according to the density space of the mapped point cloud in the candidate interference region. Measurement data is generated based on the first abnormal object and the second abnormal object; Wherein, the first abnormal object represents an interference point caused by reflection of a foreign object in the target area, the second abnormal object represents an interference area formed by the foreign object, and the first density of the first abnormal object is less than the second density of the second abnormal object. In addition, the motion characteristics of the foreign object are obtained, which characterize the dynamic change pattern of the foreign object under the action of wind and water flow; a height change prediction model of the foreign object is constructed based on the motion characteristics, including: constructing a prediction model with water surface foundation height, swing amplitude, swing frequency and initial phase as key parameters; and compensating for height anomalies caused by the dynamic swing of the foreign object in the point cloud data based on the height change prediction model. After compensating for height anomalies caused by the dynamic swaying of foreign objects in the point cloud data based on the height change prediction model, the method further includes: If abnormal regions still exist after compensating the point cloud data, the spatial scale of the abnormal regions and the valid point cloud data within the abnormal regions are obtained. If the spatial scale of the abnormal regions is greater than a preset scale threshold, a surface fitting model is generated based on the valid point cloud data, and the surface fitting model is solved to obtain a first height. The abnormal regions are then compensated based on the first height. If the spatial scale of the abnormal regions is less than or equal to the preset scale threshold, the valid point cloud data is interpolated to obtain a second height, and the abnormal regions are then compensated based on the second height. The identification of abnormal regions is based on multiple indicators, including point cloud density significantly lower than the surrounding area and unreasonable abrupt changes in elevation values ​​compared to the surrounding area.

2. The measurement method according to claim 1, characterized in that, After obtaining the point cloud data and target image of the target area, the process further includes: First pose information and second pose information are obtained, and the first pose information and the second pose information represent the spatial position and attitude angle of the target device when collecting point cloud data and target image of the target area; Based on the first pose information and the second pose information, the temporal correspondence and spatial transformation relationship between the point cloud data and the target image are determined; Based on the time correspondence and the spatial transformation relationship, spatiotemporal registration is performed on the point cloud data and the target image.

3. The measurement method according to claim 1, characterized in that, The generation of measurement data based on the first abnormal object and the second abnormal object includes: In the point cloud data, the point cloud data corresponding to the first abnormal object is removed and the point cloud data corresponding to the second abnormal object is reconstructed to obtain the measurement data.

4. A measuring device, wherein the measuring device applies the measuring method according to any one of claims 1-3, characterized in that, The measuring device includes: The data acquisition module is used to acquire point cloud data and target images of the target area. An anomaly determination module is used to determine a first anomaly object and a second anomaly object in the point cloud data based on the target image; The data generation module is used to generate measurement data based on the first abnormal object and the second abnormal object; Wherein, the first abnormal object represents an interference point caused by reflection of a foreign object in the target area, the second abnormal object represents an interference area formed by the foreign object, and the first density of the first abnormal object is less than the second density of the second abnormal object.

5. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the measurement method according to any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores executable instructions that, when executed by a processor, cause the processor to implement the measurement method of any one of claims 1-3.

Images