A 3D modeling method for the granary environment based on 3D point cloud data

Abstract

An embodiment of the present invention discloses a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data. The method includes: performing non-uniform weighted filtering on the point cloud data in the granary to obtain first point cloud data; removing dust noise from the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data; identifying the position of the robot through a dynamic occlusion detection method to remove the local point cloud data at the position of the robot in the second point cloud data to obtain third point cloud data; classifying the hole regions in the third point cloud data and repairing different types of hole regions; performing downsampling on the repaired point cloud data by using a voxel grid method to generate fourth point cloud data with a consistent point density; calculating an elevation map and performing flatness analysis on the fourth point cloud data to obtain a granary analysis result; and a high-precision three-dimensional model of the granary environment can be constructed and key environmental information can be extracted.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional point cloud data processing, and particularly to a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data. Background Art

[0002] With the development of intelligent and digital technologies, the demand for granary environment monitoring and management is increasing day by day.

[0003] Traditional granary monitoring and management methods rely on manual operations, which have problems such as low efficiency and poor accuracy. The modeling method based on three-dimensional point cloud data can achieve an accurate description of the internal environment of the granary, provide comprehensive data support for managers, and also provide a map for robot operations.

[0004] However, in the granary application scenario, the length and width of the granary are usually much larger than the distance from the grain surface to the lidar installation position, forming a flat acquisition environment. The point cloud data usually has problems such as noise, multiple holes, and large differences in point density. In addition, when there are robots operating in the granary, the robots and the operations generate dust, which directly affects the accuracy of subsequent modeling and calculation. Summary of the Invention

[0005] The main purpose of the present invention is to provide a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data, solve the data quality problem, construct a high-precision three-dimensional model of the granary environment, and accurately extract key environmental information.

[0006] To achieve the above object, in the first aspect of the present application, a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data is provided, and the method includes:

[0007] Performing non-uniform weighted filtering on the point cloud data in the granary to obtain first point cloud data;

[0008] Removing dust noise from the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data;

[0009] Identifying the position of the robot through a dynamic occlusion detection method, and removing the local point cloud data at the position of the robot in the second point cloud data to obtain third point cloud data;

[0010] Classifying the hole regions in the third point cloud data, and repairing different types of hole regions;

[0011] Performing downsampling on the repaired point cloud data by using a voxel grid method to generate fourth point cloud data with a consistent point density;

[0012] Calculating an elevation map and analyzing the flatness of the fourth point cloud data to obtain a granary analysis result.

[0013] The second aspect of the present application provides a three-dimensional modeling device for a granary environment based on three-dimensional point cloud data, including:

[0014] A filtering processing module for performing non-uniform weighted filtering on the point cloud data in the granary to obtain first point cloud data;

[0015] The filtering processing module is further configured to perform dust noise removal on the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data;

[0016] The filtering processing module is further configured to identify the position of the robot by a dynamic occlusion detection method, so as to remove the local point cloud data at the position of the robot in the second point cloud data to obtain third point cloud data;

[0017] A complement and repair module for classifying the hole regions in the third point cloud data and repairing different types of hole regions;

[0018] A voxelization module for downsampling the repaired point cloud data by using a voxel grid method to generate fourth point cloud data with a consistent point density;

[0019] A calculation and analysis module for performing elevation map calculation and flatness analysis on the fourth point cloud data to obtain a granary analysis result.

[0020] The third aspect of the present application provides an electronic device, including a memory and a processor. When the computer program stored in the memory is executed by the processor, the processor is caused to execute the steps of the first aspect and any possible implementation manner thereof.

[0021] To achieve the above object, the fourth aspect of the present application provides a computer-readable storage medium storing a computer program, which when executed by a processor, causes the processor to execute each step in the method described in the first aspect.

[0022] The present application provides a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data. By performing non-uniform weighted filtering on the point cloud data in the granary, the first point cloud data is obtained; based on the reflection intensity and dynamic filtering, dust noise is removed from the first point cloud data to obtain the second point cloud data; the position of the robot is identified through a dynamic occlusion detection method, and the local point cloud data at the position of the robot in the second point cloud data is removed to obtain the third point cloud data; the hole regions in the third point cloud data are classified, and different types of hole regions are repaired; the voxel grid method is used to downsample the repaired point cloud data to generate the fourth point cloud data with a consistent point density; the elevation map is calculated and the flatness is analyzed for the fourth point cloud data to obtain the granary analysis result; a high-precision three-dimensional model of the granary environment can be constructed, and the key environmental information can be extracted. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Among them:

[0025] Figure 1 It is a schematic flowchart of a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data provided by an embodiment of the present application;

[0026] Figure 2 It is a flowchart block diagram of a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data provided by an embodiment of the present application;

[0027] Figure 3 It is a schematic structural diagram of a three-dimensional modeling device for a granary environment based on three-dimensional point cloud data provided by an embodiment of the present application;

[0028] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed Embodiments

[0029] In order to enable those skilled in the art to better understand the solutions of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0030] In the description, claims, and above-mentioned drawings of this application, terms such as "first" and "second" are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include steps or units not listed, or may optionally further include other steps or units inherent to these processes, methods, products, or devices.

[0031] Referring to "embodiments" herein means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of this application. The phrase appearing at various positions in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0032] Please refer to Figure 1 , which is a schematic flow chart of a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data provided by an embodiment of this application. Figure 1 It includes the following steps:

[0033] 101. Perform non-uniform weighted filtering on the point cloud data in the granary to obtain the first point cloud data.

[0034] In the embodiment of this application, the execution subject of the method can be a three-dimensional modeling of a granary environment based on three-dimensional point cloud data, and in specific applications, it can be a terminal device, such as a computer.

[0035] A three-dimensional modeling method for a granary environment based on three-dimensional point cloud data provided in the embodiment of this application mainly includes steps such as point cloud filtering processing, point cloud completion, point cloud rasterization, elevation map calculation, and flatness analysis. A high-precision three-dimensional model is constructed through a series of data processing and analysis techniques, and key environmental information is extracted.

[0036] In view of the uneven distribution density of the point cloud in the granary, non-uniform weighted filtering is adopted. First, a local density estimation algorithm is used to obtain the sparsity of the point cloud.

[0037] In an alternative embodiment, the above step 101 includes:

[0038] Calculate the local density of the point cloud data in the granary;

[0039] Weight the points according to the local density of the point cloud data in the granary;

[0040] The statistical filtering algorithm is adopted, and the mean and standard deviation are used to evaluate the outlier degree of each point in the point cloud to filter out the outlier points.

[0041] Specifically, the above local density can be calculated in the following way:

[0042] Let the point cloud data set be , for each point , calculate its local density:

[0043]

[0044] Where:

[0045] represents 's neighborhood (such as nearest neighbor points);

[0046] is the kernel function:

[0047]

[0048] Among them, is the scale parameter.

[0049] In the embodiment of the present application, points are weighted according to the sparsity of the point cloud, and points in the lower density area obtain lower filtering weights to avoid feature loss. The weighting factor can be calculated according to the local density :

[0050]

[0051] So that , and points with lower density have smaller weights.

[0052] Subsequently, the statistical filtering algorithm is applied. The mean and standard deviation are used to perform statistical filtering on the point cloud, and the outlier degree of point is defined as:

[0053]

[0054] Where:

[0055] is the value of point in a certain dimension (such as axis height).

[0056] and are the mean and standard deviation of this dimension respectively.

[0057] According to the set threshold , filter out those that satisfy And Larger points are used to retain the feature points in the low-density area.

[0058] 102. Based on the reflection intensity and dynamic filtering, the above first point cloud data is subjected to dust noise removal to obtain second point cloud data.

[0059] In the embodiment of the present application, for the possible dust interference in the grain bin, the reflection intensity caused by the dust is relatively high but the duration is short. Dust noise removal is carried out based on the reflection intensity and dynamic filtering.

[0060] In an alternative embodiment, the above step 102 includes:

[0061] Calculate the global mean and standard deviation of the reflection intensities of all points in the above first point cloud data;

[0062] Determine the outlier points according to the above global mean standard deviation and the set threshold;

[0063] Perform dynamic filtering based on the scanning period;

[0064] Calculate the occurrence frequency of each point within the time window, and remove the dust points according to the occurrence frequency of each point within the time window and the stability threshold.

[0065] Reflection intensity anomaly detection. Let the point cloud data set be , where is the reflection intensity of point . Calculate the global mean and standard deviation:

[0066]

[0067] Set the threshold , and determine the dust points:

[0068]

[0069] where can be an empirical parameter. For example, the value can be between 2 and 3.

[0070] Dynamic filtering based on the scanning period. Let the multi-frame point cloud data be represents the th frame of point cloud. Define a time window containing the most recent frames:

[0071]

[0072] For each point , calculate its occurrence frequency within the time window:

[0073]

[0074] wherein is an indicator function, indicating whether the point appears in a certain frame of point cloud.

[0075] Set the stability threshold , if:

[0076]

[0077] then it is considered that the point is caused by dust and should be removed.

[0078] 103. Identify the position of the robot through the dynamic occlusion detection method, and remove the local point cloud data at the position of the robot in the second point cloud data above to obtain the third point cloud data.

[0079] In the embodiment of the present application, aiming at the problem that the robot operating in the granary affects the accuracy of the point cloud data, a dynamic occlusion detection method is adopted. The position of the robot in the granary point cloud is identified through the object detection algorithm of deep learning, and the local point cloud data at the position of the robot is removed.

[0080] Specifically, the granary point cloud data can be set as , where each point is represented as . The position of the robot in the granary is identified by the object detection algorithm of deep learning, denoted as , wherein is the point set defined by the detection frame , that is:

[0081]

[0082] Remove the point cloud data in the area where the robot is located, that is, calculate the point cloud after removing the influence of the robot :

[0083]

[0084] By combining statistical methods and time series analysis, abnormal points caused by dust are effectively identified and removed, thereby improving the quality and reliability of the point cloud data.

[0085] 104. Classify the hole areas in the above third point cloud data, and repair different types of hole areas.

[0086] The larger holes in the granary point cloud data are mainly the scanning shadow surfaces formed by the occlusion of the laser beam by the grain pile, or the areas generated by the removal of the point cloud where the robot is located. The smaller holes are caused by the excessive sparsity of the local point cloud.

[0087] In the embodiments of the present application, for the hole regions in the point cloud, they can be classified according to the size of their area or diameter:

[0088] Small hole regions: Those with an area smaller than the set area threshold are defined as small holes.

[0089] Large hole regions: Holes with an area exceeding the set area threshold are defined as large holes. The specific area threshold can be dynamically adjusted according to the resolution and accuracy requirements of the grain bin point cloud data, such as being set to 40 square centimeters.

[0090] In an alternative embodiment, the above step 104 includes:

[0091] Classify the hole regions in the above third point cloud data according to the size of the area or diameter of the hole regions to obtain the first type of holes and the second type of holes;

[0092] For the above first type of holes, use an interpolation-based reconstruction method to perform bilinear interpolation filling based on the surrounding data;

[0093] For the above second type of holes, use the Poisson surface reconstruction algorithm to complete the repair.

[0094] The area (or diameter) of the above first type of holes is smaller than the area (or diameter) of the first type of holes.

[0095] Specifically, the steps for calculating the hole area may include:

[0096] Use the Delaunay triangulation algorithm to triangulate the point cloud, divide the point cloud into a series of triangles, and the unfilled edges during the triangulation process are the hole boundaries. Project the boundary points onto the XY plane to construct a closed polygon. Given the boundary points :

[0097] Calculate the area

[0098]

[0099] where , which is applicable to the closed polygon.

[0100] For the smaller hole regions, use an interpolation-based reconstruction method to perform bilinear interpolation filling based on the surrounding data.

[0101] Specifically, interpolation can be performed based on the height values of the nearest known points (four neighboring points) around it:

[0102]

[0103] They are the height values of the four surrounding adjacent points respectively.

[0104] and is the normalized distance from the interpolation point to the adjacent point.

[0105] For a relatively large hole area, the Poisson surface reconstruction algorithm is used to complete the repair. By setting a reasonable neighborhood range for surface fitting, such as three times the point spacing of the boundary points, the missing area is filled. The specific steps may include:

[0106] Gradient field generation

[0107]

[0108] Establish the Poisson equation

[0109]

[0110] Among them, represents the Laplace operator.

[0111] Establish the regularization objective and minimize the energy

[0112]

[0113] Zero isosurface extraction, that is, extract the surface

[0114]

[0115] 105. The voxel grid method is used to downsample the repaired point cloud data to generate the fourth point cloud data with a consistent point density.

[0116] After completion and repair, the point cloud still has a large point density distribution and a large overall data volume, which is not conducive to subsequent calculations. In the embodiments of the present application, the voxel grid method can be used to downsample the point cloud to generate point cloud data with a consistent point density and reduce the calculation amount.

[0117] In an alternative embodiment, the above step 105 includes:

[0118] Construct a 3D voxel grid and partition the above repaired point cloud data according to the voxel size;

[0119] For each voxel, calculate the center point or the average point as the downsampled point to generate the downsampled point cloud.

[0120] Specifically, the above steps may include:

[0121] Voxel division: Construct a 3D voxel grid and divide the point cloud according to the voxel size for partitioning, and each voxel Containing a point set

[0122]

[0123] Representative point selection: For each voxel, calculate the center point or average point as the downsampled point :

[0124]

[0125] Generate the downsampled point cloud: Obtain a new uniform point cloud:

[0126]

[0127] where is the number of points in the point cloud after voxel grid downsampling.

[0128] 106. Perform elevation map calculation and flatness analysis on the above fourth point cloud data to obtain the granary analysis result.

[0129] Specifically, map the extracted grain surface point cloud data to a regular grid, such as a spacing of , and calculate the average elevation of each grid through interpolation. Use the nearest neighbor interpolation method to interpolate the point cloud data in each grid to the grid center to generate a relatively smooth elevation map. Perform Gaussian filtering on the elevation map to remove isolated noise points and smooth the elevation map. For example, the elevation map resolution can be 10 cm, the convolution kernel size is , and the standard deviation is 15.

[0130] In an alternative embodiment, perform flatness analysis on the above fourth point cloud data, including:

[0131] For the above fourth point cloud data, use the neighborhood gradient region averaging method to calculate the gradient of the neighborhood and calculate the average gradient according to the set size of the rectangular region.

[0132] In the embodiments of the present application, to characterize the overall flatness index of the grain surface and guide the operation of the grain surface leveling robot, the neighborhood gradient region averaging method is used. First, calculate the gradient of the neighborhood, and calculate the flatness through regional averaging. Different granularities can be achieved by adjusting the size of different calculation regions according to requirements.

[0133] Specifically, the above steps can be expressed as follows:

[0134] Calculate the gradient value of each point in the elevation map to reflect the change in the shape of the grain pile in the local area.

[0135] For each point in the elevation map , calculate the gradient

[0136]

[0137]

[0138] The modulus matrix of the gradient, where each point represents the intensity of local terrain change at that position.

[0139]

[0140] Set a rectangular area, the size of which is determined according to requirements, such as , calculate its center point The average gradient of:

[0141]

[0142] Where N represents The neighborhood of, is the number of points in the neighborhood.

[0143] Figure 2 FIG. is a flowchart of a three-dimensional modeling method for a granary environment based on three-dimensional point cloud data provided by an embodiment of the present application, to more clearly show the method flow in the embodiment of the present application; Figure 2 The steps involved in can be referred to Figure 1 The relevant specific descriptions in the shown embodiments will not be repeated here.

[0144] In the three-dimensional modeling method for a granary environment based on three-dimensional point cloud data in the embodiment of the present application, non-uniform weighted filtering is performed on the point cloud data in the granary to obtain first point cloud data; dust noise is removed from the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data; the position where the robot is located is identified through a dynamic occlusion detection method, so as to remove the local point cloud data at the position where the robot is located in the second point cloud data to obtain third point cloud data; the hole regions in the third point cloud data are classified, and different types of hole regions are repaired; the voxel grid method is used to downsample the repaired point cloud data to generate fourth point cloud data with consistent point density; elevation map calculation and flatness analysis are performed on the fourth point cloud data to obtain a granary analysis result; a high-precision three-dimensional model of the granary environment can be constructed, and key environmental information can be extracted. This method is applicable to application scenarios such as global map establishment of granary robots, internal space analysis of granaries, and evaluation of grain surface flatness.

[0145] Based on the description of the foregoing method embodiments, an embodiment of the present application further provides a three-dimensional modeling device for a granary environment based on three-dimensional point cloud data.

[0146] Figure 3 FIG. is a schematic structural diagram of a three-dimensional modeling device for a granary environment based on three-dimensional point cloud data provided by an embodiment of the present application. Figure 3The three-dimensional modeling device 300 for the granary environment based on three-dimensional point cloud data includes:

[0147] A filtering processing module 310, configured to perform non-uniform weighted filtering on the point cloud data in the granary to obtain first point cloud data;

[0148] The filtering processing module 310 is further configured to perform dust noise removal on the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data;

[0149] The filtering processing module 310 is further configured to identify the position where the robot is located through a dynamic occlusion detection method, so as to remove the local point cloud data at the position where the robot is located in the second point cloud data to obtain third point cloud data;

[0150] A complement and repair module 320, configured to classify the hole regions in the third point cloud data and repair different types of hole regions;

[0151] A rasterization module 330, configured to perform downsampling on the repaired point cloud data by using a voxel grid method to generate fourth point cloud data with a consistent point density;

[0152] A calculation and analysis module 340, configured to perform elevation map calculation and flatness analysis on the fourth point cloud data to obtain a granary analysis result.

[0153] Wherein, Figure 1 The steps in the illustrated embodiments can be executed by the respective modules of the three-dimensional modeling device 300 for the granary environment based on three-dimensional point cloud data, which will not be elaborated here.

[0154] In an embodiment of the present application, an electronic device is further proposed. Please refer to Figure 4 , Figure 4 which is a schematic structural diagram of an electronic device provided by an embodiment of the present application. Figure 4 In this, the electronic device 400 includes a processor 401 and a memory 402. The memory 402 stores a computer program. When the computer program is executed by the processor 401, it will execute any step in the method embodiment as Figure 1 shown; the electronic device 400 can be used for model training and / or application of a data enhancement network and a point cloud classification network. The electronic device 400 may further include input / output devices, etc. In a specific implementation manner, the electronic device may be a terminal device, etc.

[0155] In an embodiment, a computer-readable storage medium is further proposed. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor 401, it causes the processor 401 to execute any step in the above method embodiment.

[0156] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0157] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0158] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A three-dimensional modeling method for a granary environment based on three-dimensional point cloud data, characterized in that: The method comprises: Perform non-uniform weighted filtering on the point cloud data in the granary to obtain the first point cloud data; Dust noise is removed from the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data; the dust noise removal from the first point cloud data based on reflection intensity and dynamic filtering includes: calculating the global mean and standard deviation of the reflection intensity of all points in the first point cloud data; determining outliers according to the global mean, the standard deviation and a set threshold; performing dynamic filtering based on a scanning cycle; calculating the occurrence frequency of each point in a time window, and removing dust points according to the occurrence frequency of each point in the time window and a stability threshold; Identify the location of the robot by a dynamic occlusion detection method, so as to remove the local point cloud data of the location of the robot in the second point cloud data, and obtain third point cloud data; Classifying the hole areas in the third point cloud data and repairing the hole areas of different types; the classifying the hole areas in the third point cloud data and repairing the hole areas of different types include: classifying the hole areas in the third point cloud data according to the size of the area or diameter of the hole areas to obtain the first type of holes and the second type of holes; for the first type of holes, using an interpolation-based reconstruction method to perform bilinear interpolation filling based on surrounding data; for the second type of holes, using a Poisson surface reconstruction algorithm to complete the repair; The restored point cloud data is downsampled using the voxel grid method to generate the fourth point cloud data with consistent point density; Elevation map calculation and flatness analysis are performed on the fourth point cloud data to obtain a granary analysis result.

2. The method for three-dimensional modeling of a granary environment based on three-dimensional point cloud data according to claim 1, characterized in that: The non-uniform weighted filtering of the point cloud data in the granary includes: Calculate and obtain the local density of the point cloud data in the granary; weighting the points according to the local density of the point cloud data in the granary; A statistical filtering algorithm is used to evaluate the outlier degree of each point in the point cloud using the mean and standard deviation to filter out outliers.

3. The method for three-dimensional modeling of a granary environment based on three-dimensional point cloud data according to claim 1 is characterized in that: The method of downsampling the repaired point cloud data using the voxel grid method to generate fourth point cloud data with consistent point density includes: Constructing a 3D voxel grid, partitioning the repaired point cloud data according to voxel size; For each voxel, the center point or average point is calculated as the down-sampling point to generate a down-sampling point cloud.

4. The method for three-dimensional modeling of a granary environment based on three-dimensional point cloud data according to any one of claims 1 to 3, characterized in that: The step of calculating the elevation map of the fourth point cloud data includes: Mapping the fourth point cloud data to a regular grid, and calculating the average elevation of each grid by interpolation; interpolating the point cloud data in each grid to the center of the grid using a nearest neighbor interpolation method to generate a first elevation map; Gaussian filtering is performed on the first elevation map, and the elevation map is smoothed to obtain a target elevation map.

5. The method for three-dimensional modeling of a granary environment based on three-dimensional point cloud data according to any one of claims 1 to 3, characterized in that: Performing a flatness analysis on the fourth point cloud data includes: A neighborhood gradient regional averaging method is used for the fourth point cloud data to obtain the gradient of the neighborhood, and the average gradient is calculated according to the set rectangular area size.

6. A three-dimensional modeling device for a granary environment based on three-dimensional point cloud data, characterized in that: include: A filtering processing module is used to perform non-uniform weighted filtering on the point cloud data in the granary to obtain the first point cloud data; The filtering processing module is further used to remove dust noise from the first point cloud data based on reflection intensity and dynamic filtering to obtain second point cloud data; The filtering processing module is specifically used to: calculate the global mean and standard deviation of the reflection intensity of all points in the first point cloud data; determine outliers according to the global mean, the standard deviation and the set threshold; perform dynamic filtering based on the scanning cycle; calculate the frequency of occurrence of each point in the time window, and remove dust points according to the frequency of occurrence of each point in the time window and the stability threshold; The filtering processing module is further used to identify the position of the robot through a dynamic occlusion detection method, so as to remove the local point cloud data of the position of the robot in the second point cloud data to obtain third point cloud data; A completion and repair module, used to classify the hole areas in the third point cloud data and repair different types of hole areas; The completion and repair module is specifically used to: classify the hole areas in the third point cloud data according to the area or diameter of the hole areas to obtain the first type of holes and the second type of holes; for the first type of holes, adopt an interpolation-based reconstruction method to perform bilinear interpolation filling based on surrounding data; for the second type of holes, adopt a Poisson surface reconstruction algorithm to complete the repair; A rasterization module, used to downsample the repaired point cloud data using a voxel grid method to generate fourth point cloud data with consistent point density; The calculation and analysis module is used to perform elevation map calculation and flatness analysis on the fourth point cloud data to obtain a granary analysis result.

7. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 5.

Images