A method and system for calculating the volume of a cavity three-dimensional point cloud with adaptive incomplete deformation

Abstract

The present invention discloses a method and system for calculating the volume of a three-dimensional point cloud of a cavity with self-adaptive incomplete deformation, which relates to the technical field of lidar point cloud processing, and includes: obtaining a reference point cloud; the reference point cloud is an empty bin point cloud directly measured or a virtual empty bin point cloud jointly generated by historical cavity point clouds and a granary design drawing; dividing the reference point cloud into a bin wall point cloud above the material surface and a bin wall point cloud below the material surface through the material surface point cloud; fusing the bin wall point cloud below the material surface corresponding to the cavity point cloud in the divided reference point cloud and the material surface point cloud of the real-time cavity point cloud to obtain the material point cloud to be calculated; calculating the volume of the material point cloud to be calculated; the present invention fuses historical cavity point clouds and granary design parameters to obtain a reference point cloud, effectively compensates for the deformation of the bin body and the missing point cloud in actual measurement through this virtual empty bin modeling technology, and combines the material surface point cloud segmentation and hierarchical fusion algorithm, significantly improving the calculation accuracy of the material volume in the granary.

Description

Technical Field

[0001] The present invention relates to the technical field of laser radar point cloud processing, and in particular to a method and system for calculating the volume of a cavity three-dimensional point cloud with adaptive incomplete deformation. Background Art

[0002] Radar level meters are widely used in industry, especially in the measurement of solid materials. They determine the height of the material by emitting electromagnetic waves and measuring the time it takes for the reflected waves to be reflected. However, the disadvantage is that radar level meters usually have a low resolution, which can result in inaccurate measurements of inventory levels, especially when measuring small changes.

[0003] Structured light scanning technology acquires the 3D shape of an object by projecting a grating or stripe pattern, and is commonly used in industrial inspection and reverse engineering. This technology does not work well in environments with strong light or reflective surfaces, and is affected by the texture and color of the object's surface, which may result in incomplete or inaccurate data collection.

[0004] Image processing and computer vision use cameras and image processing algorithms, combined with deep learning technology, to extract three-dimensional information of objects from images. This method is sensitive to ambient lighting conditions and may have difficulty accurately identifying objects in complex backgrounds.

[0005] In the calculation of the granary point cloud volume, because the granary itself is closed and it is impossible to directly obtain the complete model of the granary raw materials, the granary inventory volume is generally obtained by subtracting the volume of the blank area model above the material surface from the volume of the empty granary model. Ideally, the empty granary model can be directly measured or obtained based on the granary drawings when the granary is initially installed. This is an ideal empty granary model, but in the subsequent practical process, the interior of the granary will deform and differ from the ideal empty granary model. This difference will lead to inaccurate final volume measurement. In addition, dust will be generated inside the granary, and most laser radars cannot effectively penetrate the dust. Directly using the point cloud measured by the laser radar in a dusty granary will result in a large volume error in calculating the blank area model above the material surface. Summary of the invention

[0006] The purpose of the present invention is to solve the problem of low accuracy of the existing granary point cloud volume calculation method. The present invention provides a cavity three-dimensional point cloud volume calculation method with adaptive incomplete deformation. The technical solution includes the following steps:

[0007] Obtain a reference point cloud; the reference point cloud is a directly measured empty warehouse point cloud or a virtual empty warehouse point cloud generated by combining historical cavity point clouds and grain warehouse design drawings;

[0008] The real-time cavity point cloud is segmented to obtain the material surface point cloud, and the reference point cloud is segmented into the warehouse wall point cloud above the material surface and the warehouse wall point cloud below the material surface through the material surface point cloud;

[0009] Fuse the bin wall point cloud below the material surface and the material point cloud of the real-time cavity point cloud in the segmented reference point cloud to obtain the material point cloud to be calculated;

[0010] Calculate the volume of the material point cloud to be calculated.

[0011] Specifically, the obtaining of the reference point cloud includes:

[0012] If the empty bin point cloud can be directly measured, use the measured empty bin point cloud as the reference point cloud, or directly select the historical cavity point cloud that meets the requirements of the first volume confidence interval as the reference point cloud; otherwise, construct a virtual empty bin point cloud through the historical cavity point cloud and the geometric parameters of the grain bin as the reference point cloud.

[0013] Specifically, the constructing of the virtual empty bin point cloud through the historical cavity point cloud and the geometric parameters of the grain bin as the reference point cloud includes:

[0014] Select the historical cavity point cloud that meets the requirements of the second volume confidence interval;

[0015] Align the axis of the historical cavity point cloud with the central axis of the grain bin design drawing;

[0016] Perform data fusion on the axially aligned historical cavity point cloud and the design parameters of the grain bin to obtain a virtual empty bin point cloud;

[0017] Perform error compensation on the virtual empty bin point cloud to obtain the reference point cloud;

[0018] Preferably, the first volume confidence interval is ; the second volume confidence interval is: , is the volume confidence level, defined as:

[0019] ;

[0020] Wherein, is the volume of a certain historical cavity point cloud, is the theoretical volume in the empty bin state.

[0021] Specifically, the performing of data fusion on the axially aligned historical cavity point cloud and the design parameters of the grain bin to obtain a virtual empty bin point cloud is specifically:

[0022] Couple the measured data with the design parameters through axial translation transformation and parametric stretching coefficient to achieve rapid geometric reconstruction of the virtual empty bin. Among them, the virtual empty bin point cloud to be reconstructed includes a measured section and an extended section. The measured section directly uses the bin wall points in the historical cavity point cloud; the extended section is divided into the side wall and the bin bottom. The side wall is based on the side wall height in the grain bin design drawing Formed by axial stretching:

[0023] ;

[0024] ;

[0025] Among them, is the side wall height of the historical cavity point cloud, is the difference between the side wall height of the granary design drawing and the side wall height of the historical cavity point cloud, is the stretching coefficient, which is used to control the degree of stretching so that the extended section is smoothly connected to the measured section, are the virtual bin wall point coordinates of the side wall of the extended section, is the coordinate of the highest point of the measured section, is the axial direction of the granary; the measured section and the extended section are spliced to obtain the virtual empty bin point cloud. The essence of the above method is the parametric deformation method in geometric modeling. While ensuring the calculation efficiency, the practicality is enhanced by introducing adjustable parameters. Using the height of the design drawing ensures that the reconstructed virtual granary meets the design specifications, and the historical height guarantees the consistency with the measured data.

[0026] Independently reconstruct the bin bottom according to the design drawing parameters:

[0027] The flat bin bottom generates a uniformly gridded point cloud based on the height of the design drawing and the projected side wall bottom boundary;

[0028] The conical bin bottom is driven by the design drawing parameters (drawing cone top , drawing half angle , drawing radius ) and generates a point cloud based on the discretization of the parametric analytical surface, which is expressed as:

[0029] ;

[0030] Among them, is the conical bin bottom, is the drawing cone top, is the drawing half angle, is the drawing radius, is the radial distance, is the azimuth angle, is the plane orthogonal basis vector perpendicular to the axial direction ;

[0031] Finally, through the normal vector consistency constraint and the KD-Tree based nearest neighbor interpolation algorithm, the topological closed fusion of the side wall-bottom point cloud is realized, eliminating the geometric seam, ensuring that the virtual empty bin point cloud strictly conforms to the design specifications and structural continuity in the global form while inheriting the local details of the measured data.

[0032] The measured section and the extended section are spliced to obtain the virtual empty bin point cloud.

[0033] Specifically, the steps of segmenting the real-time cavity point cloud to obtain the material surface point cloud and then segmenting the reference point cloud into the bin wall point cloud above the material surface and the bin wall point cloud below the material surface through the material surface point cloud are as follows:

[0034] Perform spatial registration on the real-time cavity point cloud and the reference point cloud;

[0035] Perform differential distance field calculation and spatial mapping on the real-time cavity point cloud;

[0036] Perform adaptive threshold segmentation on the material surface and the bin wall of the real-time cavity point cloud;

[0037] Perform morphological optimization and topological verification on the material surface point cloud;

[0038] Based on the material surface point cloud of the cavity point cloud, segment the reference point cloud into the bin wall point cloud above the material surface and the bin wall point cloud below the material surface.

[0039] Specifically, the steps of segmenting the reference point cloud into the bin wall point cloud above the material surface and the bin wall point cloud below the material surface based on the material surface point cloud of the cavity point cloud include:

[0040] Modeling the height field of the material surface; calculating its height distribution field from the material surface point cloud :

[0041] ;

[0042] Among them, is the neighborhood of the horizontal projection of the material surface, is the point on the material surface is the vertical coordinate.

[0043] Based on the height distribution field , perform regional segmentation on the reference point cloud:

[0044] (a) Bin wall point cloud below the material surface: ;

[0045] (b) Bin wall point cloud above the material surface: ;

[0046] Among them, is the reference point cloud, is the point on the reference point cloud The vertical coordinate, is the standard deviation of the noise;

[0047] Specifically, the step of fusing the bin wall point cloud below the material surface of the corresponding cavity point cloud in the segmented reference point cloud and the material point cloud of the real-time cavity point cloud to obtain the material point cloud to be calculated includes the following steps:

[0048] Extract the boundary of the material point cloud through the α-shape algorithm and the boundary of the bin wall point cloud below the material surface , where is the point on the boundary of the material point cloud, is the membership function of α-shape;

[0049] Construct a KD-Tree acceleration structure and perform a radius r search on the boundary , , is the threshold of the radius search, mark the overlapping area with the boundary , is the point on the boundary of the reference point cloud in the overlapping area, is the nth point on the material surface within the neighborhood whose distance is less than , and use the non-maximum suppression strategy to eliminate redundant points;

[0050] Establish a buffer at the junction of the boundary and the boundary , and realize the continuous fusion of the material point cloud and the bin wall point cloud below the material surface by fitting the transition surface with the moving least squares method. The formula is as follows:

[0051] ;

[0052] Among them, and are two independent variables for parameterizing the surface, is the distance weight function, defined as , represents the point in the three-dimensional space mapped from the parameter space , is the point to the shortest Euclidean distance of the material surface boundary, is the buffer width, controlling the transition range of the fusion area, and respectively represent the local fitting surfaces of the material point cloud and the bin wall point cloud, represents the transition surface of the fusion area;

[0053] Adopt a multi - resolution strategy based on Poisson disk sampling to adaptively adjust the point spacing in the fusion area , where is the resampled point spacing, and adaptively take the minimum spacing between the material point cloud and the bin wall point cloud below the material surface, is the average point spacing of the material point cloud, is the average point spacing of the bin wall point cloud below the material surface;

[0054] Apply anisotropic curvature diffusion filtering to eliminate local distortions caused by noise:

[0055] ;

[0056] where is the Gaussian curvature of point and its neighboring points , quantifying the local surface curvature, is the normal vector of the normal neighboring point , indicating the local direction of the surface, is the set of neighboring points of point (based on KD - Tree search), is the th three - dimensional coordinate of the point.

[0057] Furthermore, after the historical cavity point cloud and the real - time cavity point cloud are obtained, pre - processing operations need to be performed on them, including the following steps:

[0058] Downsample the grain bin measurement point cloud and remove outliers; for the convenience of description here, the historical point cloud and the real - time cavity point cloud are called the grain bin measurement point cloud;

[0059] Evaluate and repair the integrity of the grain bin measurement point cloud.

[0060] Meanwhile, the present invention also provides a system for calculating the volume of a three - dimensional point cloud of a cavity with adaptive incomplete deformation. The system is used to calculate the volume of the three - dimensional point cloud of the cavity, including:

[0061] A data acquisition module for automatically collecting or manually importing point cloud data;

[0062] A data processing module for calculating the volume according to the point cloud data in the data acquisition module;

[0063] A data transmission module for transmitting the volume of the three - dimensional point cloud of the cavity calculated by the data processing module to the terminal;

[0064] The volume calculation method adopted in the data processing module is the above - mentioned method for calculating the volume of a three - dimensional point cloud of a cavity with adaptive incomplete deformation.

[0065] After adopting the above solution, the beneficial effects of the present invention are as follows: The virtual empty bin modeling technology effectively compensates for the deformation of the bin body and the lack of point cloud in actual measurement. Combining the material point cloud segmentation and hierarchical fusion algorithm, it significantly improves the calculation accuracy of the volume of the grain in the bin. It innovatively fuses the historical point cloud and design parameters to reconstruct the reference point cloud, breaking through the dependence on on-site empty bin scanning and providing a highly robust solution for dynamic warehouse management. Brief Description of the Drawings

[0066] Figure 1 It is the main flow chart of the present invention;

[0067] Figure 2 It is a schematic diagram of generating a virtual cavity point cloud through the historical cavity point cloud and design drawings in the specific implementation of the present invention;

[0068] Figure 3 It is a schematic diagram of the principle of dividing the reference point cloud into the wall point cloud above the material surface and the wall point cloud below the material surface in the specific implementation of the present invention;

[0069] Figure 4 It is a schematic diagram of the fusion of the material point cloud and the wall point cloud below the material surface in the specific implementation of the present invention;

[0070] Figure 5 It is a usage scenario diagram of the method for calculating the volume of the three-dimensional point cloud of the cavity with adaptive incomplete deformation provided by the present invention. Detailed Description of the Embodiments

[0071] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0072] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0073] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to implement and use the present invention. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but rather to be in line with the broadest scope consistent with the principles and features disclosed in the present invention.

[0074] For the convenience of understanding, the specific terms in the present invention are explained here first:

[0075] 1. Unoccupied Silo Point Cloud (USC): The point cloud of the silo in the unoccupied state obtained by a single measurement at any time.

[0076] 2. Open Space Point Cloud (OSC): The point cloud above the material surface in the point cloud obtained by a single measurement of the silo at any time, including the complete point cloud of the material surface and the cavity silo wall above the material surface. When there is no material in the silo, it is the unoccupied silo point cloud. Note that the silo wall here also includes the silo wall at the top of the silo, and both the silo top and the silo bottom can be part of the silo wall, so the silo wall is collectively referred to in the description of the present invention.

[0077] 3. Material Point Cloud: The point cloud below the material surface in the point cloud obtained by a single measurement of the silo at any time, including the material point cloud and the silo wall point cloud below the material surface. Note that the silo wall here also includes the silo wall at the bottom of the material.

[0078] 4. Reference Point Cloud: The unoccupied silo point cloud used as a template, which can be the directly measured unoccupied silo point cloud or the point cloud of the silo in the unoccupied state obtained by virtual reconstruction.

[0079] The core idea of the present invention is: calculating the volume of the single-time silo point cloud is achieved by combining the reference point cloud and the material point cloud in the cavity point cloud of the current measurement to obtain the material point cloud composed of the complete material surface and the silo wall below the material surface, and three-dimensionally meshing the material point cloud to obtain the volume of the material point cloud. In the application scenario of the silo, the silo is basically in the state of feeding and discharging every day, and it is difficult to directly measure the empty silo model. Therefore, the present invention obtains a more accurate unoccupied silo point cloud closer to the current measurement time according to the historical cavity point cloud during use and the silo drawing parameters, rather than directly using the initial ideal empty silo model (i.e., the unoccupied silo point cloud obtained only according to the drawing).

[0080] Please refer to Figure 5 , Figure 5This is a usage scenario diagram of the method for calculating the volume of a three-dimensional point cloud of a cavity with adaptive incomplete deformation provided by the present invention. As Figure 5 shown, the terminal and the server are connected through a network, for example, through a wired or wireless network connection, etc. Among them, the terminal may include, but is not limited to, portable terminals such as mobile phones and tablets installed with various network platform applications, as well as fixed terminals such as computers, inquiry machines, and advertising machines. Among them, the server provides various business services for users, including service push servers, user recommendation servers, etc.

[0081] It should be noted that Figure 5 the scenario diagram of the method for calculating the volume of a three-dimensional point cloud of a cavity with adaptive incomplete deformation shown is only an example. The terminal, server, and application scenario described in the embodiments of the present invention are for more clearly explaining the technical solutions of the embodiments of the present invention, and do not generate limitations on the technical solutions provided by the embodiments of the present invention. Those of ordinary skill in the art can know that with the evolution of the system and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0082] A method for calculating the volume of a point cloud of a granary with adaptive incomplete deformation in a specific implementation includes the following steps:

[0083] S1. Preprocessing of the measured point cloud of the granary; this step is the preprocessing of the measured point cloud of the granary, and subsequent historical cavity point clouds and real-time cavity point clouds need to be preprocessed. For the convenience of description, they are uniformly referred to as the measured point cloud of the granary here.

[0084] S1.1. Downsampling the measured point cloud of the granary and removing outliers; the specific steps are as follows: Considering the particularity of the large dust in the granary scenario and the complexity of data processing, it is necessary to preprocess the point cloud data first. Among them, downsampling and outlier removal are key preprocessing steps, aiming to improve data quality and reduce the complexity of subsequent calculations.

[0085] S1.1.1. Introduce a voxel grid downsampling algorithm, establish a three-dimensional voxel rasterization model, and weight-average the point cloud coordinates within each voxel unit into a single centroid point, realizing the spatial uniformity of point cloud density while maintaining the significance of geometric features, effectively reducing data redundancy and computational complexity.

[0086] S1.1.2. Based on the statistical outlier removal mechanism, construct a k-nearest neighbor search domain and calculate the distance distribution between each point and the neighborhood point set, and remove the outlier noise points deviating from the main distribution through a preset standard deviation multiplier threshold. The specific steps are as follows:

[0087] (1) Construct a k-nearest neighbor search domain. For each point in the point cloud, calculate its Euclidean distance , ) from all other points in the point cloud, where , choose the one with the smallest distance Points, constitute of Nearest Neighbor Search Domain ),in , the neighborhood distance is With Euclidean distance between neighboring points;

[0088] (2) Traverse the points outside the k nearest neighbor search domain in the cavity point cloud. If the point satisfy , then determine The noise points are removed, among which, , are the mean and standard deviation of neighborhood distances, is the empirical coefficient;

[0089] (3) Identify the largest point cloud subset with topological continuity through Euclidean distance clustering and eliminate the discrete noise point set;

[0090] Through the above steps, the signal-to-noise ratio and spatial consistency of the point cloud are significantly improved while maintaining the integrity of the main structure of the point cloud, laying a solid data foundation for subsequent three-dimensional reconstruction and morphological analysis.

[0091] S1.2. Integrity assessment and repair of the granary measurement point cloud; After the granary measurement point cloud is preprocessed, the point cloud volume calculation method of the subsequent grid generation needs to ensure the integrity of the point cloud, so the denoised point cloud needs to be evaluated for integrity. Due to the non-uniform distribution of dust interference in the granary scene, the measured granary point cloud will have a certain degree of incompleteness and holes. Therefore, before subsequent processing, the granary point cloud needs to be repaired to obtain a valid granary point cloud to avoid the influence of the residual point cloud edge on the subsequent feature extraction and grid volume calculation. The specific steps are as follows:

[0092] S1.2.1. Construct the topological relationship of the point cloud surface based on Delaunay triangulation, extract the boundary contour B through the α-shape algorithm, and identify the hole area The geometric parameters (perimeter , curvature etc.), providing geometric basis for subsequent restoration;

[0093] S1.2.2. Calculate the standard deviation of the point cloud density per unit volume , when the local area density is lower than the overall density mean of 3 It is determined as a potential incomplete area when [conditions are met]; non - continuous edge segments are detected by the Curvature Variation Rate (CVR), and when CVR > 0.25, it is determined as an abnormal fracture; after constructing the Delaunay triangulation network, the hole areas that cannot form a closed surface are detected. When any of the above defect types (incompleteness, fracture, and holes) are detected, the point cloud repair process is initiated:

[0094] (1)Local geometric feature modeling; for each identified hole, a local reference coordinate system is established along its boundary, and geometric modeling is carried out in two steps:

[0095] (a) Principal curvature analysis: The maximum / minimum curvature of the boundary points is calculated using the covariance matrix method ( ), and the curvature tensor field is constructed, where are the eigenvalues, and are the principal directions of curvature;

[0096] (b) Feature propagation: In the boundary neighborhood ), the local surface is fitted by Moving Least Squares (MLS), and its energy function is:

[0097] ;

[0098] where are the point cloud data points in the neighborhood ), is the normal vector of for constraining the direction of surface fitting, is the position of the current fitting point, is the radial basis function for weighting the fitting error according to the distance between the point and to ensure the smoothness and continuity of the local surface; the goal of this energy function is to minimize the geometric deviation between the fitted surface and the neighborhood point cloud while maintaining the smooth characteristics of the surface;

[0099] (2)Global structure constraint extraction; to enhance the global consistency of the above repair results, a geometric reasoning model based on graph neural network is introduced:

[0100] (a) Symmetry detection: The symmetry axis of the point cloud is matched by the non - rigid ICP algorithm, and a mirror transformation is established to generate a symmetric candidate point set ;

[0101] (b) Topological consistency constraint: A Markov random field model is constructed, and the energy term is defined as:

[0102] ;

[0103] where, and For matching point pairs, is the single-point geometric likelihood (such as curvature continuity), It is a topological smoothness constraint between adjacent points to ensure the geometric consistency of the repaired area;

[0104] (3) Adaptive repair strategy: A divide-and-conquer strategy is used to handle holes of different scales according to the scale of the holes:

[0105] (a) For holes whose perimeter is less than the threshold, : Based on radial basis function interpolation, solve the equation:

[0106] ;

[0107] in, is the point to be estimated, for , is a cubic kernel function that controls the local influence range of the interpolation function. is a low-order polynomial that compensates for the global deviation of the radial basis function. is the order, taking a low order can avoid overfitting, through the boundary conditions Determine the weight , generate a smooth transition surface;

[0108] (b) For holes whose perimeter is greater than or equal to the threshold, : Combining Generative Adversarial Networks with Physical Constraints to Train Conditional Generators , where the latent variable z follows a Gaussian distribution, condition Contains geometric descriptors such as boundary curvature and normal vector field, and generates point clouds by jointly optimizing adversarial loss and geometric regularization terms;

[0109] Through the above steps, the curvature continuity and physical rationality of manufacturing constraints (such as flatness and roundness) are maintained, and the incomplete and hole parts of the granary point cloud are effectively repaired. Subsequent granary point clouds are processed based on the repaired ones in S1.2, which effectively avoids the problem of feature extraction and meshing anomalies caused by incomplete holes;

[0110] S2. Obtain a reference point cloud; the reference point cloud is a directly measured empty warehouse point cloud or a virtual empty warehouse point cloud generated by combining historical cavity point clouds and grain warehouse design drawings;

[0111] In the existing scenario of grain silo usage, since the grain silo is in a state of adding materials for a long time, there will be a period of 1 - 3 months when a single grain silo is not empty. It is difficult to directly measure and obtain the empty silo point cloud. Moreover, due to the deformation of the grain silo itself and the error of the measuring equipment over time, there will be small - angle rotation and offset between the two empty silo point clouds measured 1 - 3 months apart. Therefore, a standard empty silo point cloud is needed as a template. If the actual empty silo point cloud can be directly measured, it can be used as the template, that is, the reference point cloud. However, it is difficult to obtain in the actual usage scenario. Therefore, a virtual empty silo point cloud needs to be constructed as the reference point cloud. It should be understood that due to the actual process error, the grain silo will increase over time. Therefore, the empty silo point cloud selects the most recently measured empty silo point cloud, and the historical empty cavity point cloud also uses the most recently measured empty silo point cloud. That is to say, the historical empty cavity point cloud and the empty silo point cloud cannot be used indefinitely after being measured once, but will be updated according to needs. After each measurement, the historical empty cavity point cloud and the empty silo point cloud measured last time will be overwritten.

[0112] Generally, grain silo manufacturers will have design drawings of the grain silo. Taking a square grain silo as an example, according to the geometric parameters of the length, width, and height of the grain silo obtained from the drawings, a virtual empty silo point cloud in a custom coordinate system can be generated. However, because some grain silos are not built strictly according to the design drawings, there will be a large error in directly generating a virtual empty silo point cloud based on the geometric parameters. Therefore, it is necessary to generate a virtual empty silo point cloud by combining the geometric parameters of the design drawings and the measured historical empty cavity point cloud to be used as the reference point cloud.

[0113] S2.1. If the empty silo point cloud can be directly measured, use the measured empty silo point cloud as the reference point cloud, or directly select the historical empty cavity point cloud that meets the requirements of the first volume confidence interval as the reference point cloud; specifically, define the volume confidence as:

[0114] ;

[0115] where is the volume of a certain historical empty cavity point cloud, is the theoretical volume in the empty silo state; when the volume confidence , it is determined that the historical empty cavity point cloud meets the requirements of the first volume confidence interval, that is, the empty cavity point cloud is close to the empty silo state, and directly use this historical empty cavity point cloud as the reference point cloud; otherwise, enter S2.2;

[0116] S2.2. Construct a virtual empty silo point cloud as the reference point cloud through the historical empty cavity point cloud and the geometric parameters of the grain silo; the steps are as follows:

[0117] Since the grain bin is rarely in an empty state, and the cavity point cloud measured once is also part of the point cloud in the empty state of the grain bin. For example, when the volume of the material in the bin is 1 / 4, the volume of the 3 / 4 cavity above the material is the same as or approximately the same as the empty bin point cloud (because there may be a small amount of material on the bin wall). Therefore, the present invention can also generate a virtual empty bin point cloud through the cavity point cloud measured once and the geometric parameters of the grain bin, and use this as the reference point cloud. The principle is as Figure 2 shown.

[0118] S2.2.1. Select the historical cavity point cloud that meets the requirements of the second volume confidence interval; to ensure the accuracy and reliability of the virtual empty bin point cloud, the selected historical cavity point cloud needs to meet geometric integrity. Therefore, the volume confidence of the historical cavity point cloud needs to be within the second volume confidence interval: , that is, the historical cavity point cloud should be able to cover the main structural features of the grain bin to ensure that it can accurately reflect the local geometric shape of the grain bin;

[0119] S2.2.2. Align the axis of the historical cavity point cloud with the central axis of the grain bin design drawing; based on the reference axis (grain bin central axis) of the grain bin design drawing, use the principal component analysis method to establish a design coordinate system with the grain bin central axis as the main axis; calculate the covariance matrix of the historical cavity point cloud: , where is the covariance matrix, used to describe the spatial distribution of data points, is the total number of points in the historical cavity point cloud, is the th point in the historical cavity point cloud in three-dimensional coordinates, is the geometric center of the historical cavity point cloud; perform eigenvalue decomposition on the covariance matrix , and extract the eigenvector corresponding to the largest eigenvalue (reflecting the main direction of the grain bin) as the axis of the historical cavity point cloud, and achieve alignment by rotating ;

[0120] S2.2.3. Perform data fusion on the axially aligned historical cavity point cloud and the design parameters of the grain bin to obtain the virtual empty bin point cloud; since the historical cavity point cloud meets the second confidence level, the height of the historical cavity point cloud will be less than the design height of the grain bin drawing. Therefore, the virtual empty bin point cloud to be reconstructed includes the measured section and the extended section. The measured section directly uses the bin wall points in the historical cavity point cloud; the extended section is divided into the side wall and the bottom of the bin. The side wall is axially stretched based on the side wall height in the grain bin design drawing:

[0121] ;

[0122] ;

[0123] The above solution utilizes the parametric deformation method in geometric modeling to generate a new geometric shape by stretching the point cloud along a specific direction. Among them, and are respectively the height of the historical cavity point cloud (the vertical distance between the highest point and the lowest point in the point cloud) and the height in the design drawing (the vertical distance between the highest point and the lowest point of the granary in the design drawing), is the stretching coefficient, which is used to control the degree of stretching so that the extended section is smoothly connected to the measured section, are the virtual bin wall point coordinates of the side wall of the extended section, is the coordinate of the highest point of the measured section, is the axial direction of the granary; the measured section and the extended section are spliced to obtain the virtual empty bin point cloud; the essence of the above method is the parametric deformation method in geometric modeling. While ensuring the calculation efficiency, it enhances the practicability by introducing adjustable parameters. Using the height of the design drawing ensures that the reconstructed virtual granary conforms to the design specifications, while the historical height

[0124] Independently reconstruct the bin bottom according to the design drawing parameters:

[0125] The flat bin bottom generates a uniformly rasterized point cloud based on the height of the design drawing and projects the bottom boundary of the side wall;

[0126] The conical bin bottom is driven by the design drawing parameters (the drawing cone top , the drawing half angle , the drawing radius ) and generates a point cloud based on the discretization of the parametric analytical surface, which is expressed as:

[0127] ;

[0128] Among them, is the conical bin bottom, is the drawing cone top, is the drawing half angle, is the drawing radius, is the radial distance, is the azimuth angle, is the plane orthogonal basis vector perpendicular to the axial direction ;

[0129] Finally, through the normal vector consistency constraint and the KD-Tree based nearest neighbor interpolation algorithm, the topological closure fusion of the side wall-bin bottom point cloud is realized, eliminating the geometric seam, and ensuring that the virtual empty bin point cloud strictly conforms to the design specifications and structural continuity in the global morphology while inheriting the local details of the measured data.

[0130] Splice the measured section and the extended section to obtain the virtual empty bin point cloud.

[0131] S2.2.4. Perform error compensation on the virtual empty bin point cloud to obtain the reference point cloud; specifically: taking a cylindrical grain bin as an example, perform cylindrical fitting through the random sample consensus algorithm to extract the bin radius The objective function of cylindrical fitting is: where is the radius of the cylinder to be fitted, is the reference point (three-dimensional coordinates) on the axis of the cylinder, is the th point in the historical cavity point cloud, and iteratively optimize to obtain the optimal The confidence level of the fitted cylinder radius is: where is the number of inliers extracted by the random sample consensus algorithm, is the total number of points in the point cloud; when (insufficient confidence level), introduce the designed cylinder radius from the design drawing for weighted fusion to obtain the fused cylinder radius : ;

[0132] Through the above steps, when the grain bin is cylindrical, by adding the methods of confidence level and weighted fusion, it is possible to effectively combine the measured cavity point cloud and the design parameters to generate a more stable virtual empty bin point cloud, and use this virtual empty bin point cloud as the reference point cloud. For grain bin point clouds of other shapes, virtual empty bin point clouds with high confidence levels can be established through the same data fusion scheme and error compensation mechanism; but in fact, most existing grain bins are cylindrical.

[0133] S3. Segment the real-time cavity point cloud to obtain the material surface point cloud, and segment the reference point cloud into the bin wall point cloud above the material surface and the bin wall point cloud below the material surface through the material surface point cloud, and the principle is as Figure 3 shown;

[0134] After obtaining the reference point cloud through S2, the material surface of the grain bin to be measured at this moment can be extracted by measuring the real-time cavity point cloud (the cavity point cloud at a certain moment), and at the same time, the bin wall parts above and below the material surface in the reference point cloud relative to the real-time cavity point cloud can be extracted, providing a geometric basis for subsequent data fusion. In view of the geometric consistency of the bin wall structures of the reference point cloud and the real-time cavity point cloud, the bin walls of the reference point cloud and the real-time cavity point cloud are defined as the consistency regions, and the material surface of the real-time cavity point cloud is defined as the non-consistency region.

[0135] The reference point cloud obtained from S2 is in the custom-designed coordinate system, while the real-time cavity point cloud is in the measured local coordinate system. Therefore, it is necessary to unify the two coordinate systems first and then use the differential distance field to segment the material surface and the silo wall.

[0136] S3.1. Perform spatial registration on the real-time cavity point cloud and the reference point cloud;

[0137] Use the improved ICP algorithm to perform rigid registration on the two point clouds: screen the matching point pairs through the RANSAC mechanism, and solve the optimal rotation matrix and translation vector of the real-time cavity point cloud and the reference point cloud respectively, so that the objective function is minimized, eliminate the systematic offset caused by the pose error of the measuring device, establish a unified spatial reference system, and perform registration on the two through the optimal rotation matrix and translation vector of the real-time cavity point cloud and the reference point cloud;

[0138] S3.2. Calculate the differential distance field and perform spatial mapping on the real-time cavity point cloud;

[0139] Based on the KD-Tree data structure, construct an accelerated spatial index of the reference point cloud, and perform nearest neighbor search ( ) on each point in the real-time cavity point cloud to obtain its nearest Euclidean distance to the reference point cloud, and generate the scalar distance field on the surface of the real-time cavity point cloud, where represents the point in the real-time cavity point cloud; this distance field quantifies the degree of local deformation caused by material accumulation in the cavity point cloud, and the material surface area will show high distance value characteristics due to significant protrusions;

[0140] S3.3. Perform adaptive threshold segmentation on the material surface and the silo wall of the real-time cavity point cloud;

[0141] To distinguish between silo wall noise and the real material surface, design a dynamic threshold function , where , are the mean and standard deviation of the scalar distance field D respectively, and β is the sensitivity coefficient (usually taken as 2.5 - 3.5). Through binary operation, segment the real-time cavity point cloud into the material point cloud and the silo wall point cloud, is the distance threshold; this threshold strategy can effectively suppress the pseudo-signals caused by point cloud registration residuals and measuring device jitter;

[0142] S3.4. Perform morphological optimization and topological verification on the material point cloud;

[0143] To improve the topological rationality of the segmentation results, the following post-processing steps are adopted for optimization: (a) Perform Euclidean clustering analysis on the material surface point cloud and remove discrete noise blocks with an area smaller than the threshold ; (b) Construct a surface mesh of the material surface based on Delaunay triangulation and use the region growing algorithm to fill local holes caused by occlusion; (c) Extract the largest connected domain as the final material surface point cloud to ensure a smooth transition with the bin wall boundary;

[0144] S3.5. Material surface point cloud based on cavity point cloud , the reference point cloud is segmented into the bin wall point cloud above the material surface and the bin wall point cloud below the material surface ; the specific steps are as follows:

[0145] S3.5.1. Modeling the height field of the material surface; From the material surface point cloud extracted in S3.4 , calculate its height distribution field :

[0146] ;

[0147] where is the neighborhood of the horizontal projection of the material surface, and is the vertical coordinate of the point on the material surface.

[0148] S3.5.2. Based on the height distribution field , perform region segmentation on the reference point cloud cloud:

[0149] (a) Bin wall point cloud below the material surface: ;

[0150] (b) Bin wall point cloud above the material surface: ;

[0151] where is the vertical coordinate of the point on the reference point cloud, is the standard deviation of the noise, estimated by statistical filtering residuals, and used to offset the influence of point cloud noise on segmentation;

[0152] Through the above process, this solution can effectively segment the bin wall area and the material surface area in the cavity point cloud, as well as the bin wall areas above and below the material surface of the corresponding cavity point cloud in the empty bin point cloud. While retaining the fine-grained geometric features of the material surface, it realizes high-precision extraction of the material surface, providing a reliable three-dimensional data basis for the combination of the material surface of the subsequent measurement point cloud and the bin wall of the reference point cloud, as well as the volume calculation of the material point cloud.

[0153] S4. Combine the wall point cloud below the material surface of the cavity point cloud corresponding to the segmented reference point cloud and the material point cloud of the real-time cavity point cloud to obtain the material point cloud to be calculated. The principle is as Figure 4 shown;

[0154] After the point cloud segmentation in S3, the material surface part in the cavity point cloud is obtained , as well as the wall point cloud above the material surface and the wall point cloud below the material surface in the reference point cloud relative to the material surface in the cavity point cloud and . Combining the material point cloud and the wall point cloud below the material surface of the reference point cloud can obtain the material point cloud. The specific steps are as follows:

[0155] S4.1. Extract the boundary of the material point cloud and the boundary of the wall point cloud below the material surface , where is the point on the boundary of the material point cloud, is the membership function of α-shape;

[0156] S4.2. Construct a KD-Tree acceleration structure and perform a radius r search on the boundary , , is the threshold of the radius search. Mark the overlapping area with the boundary , is the point on the boundary of the reference point cloud in the overlapping area, is the th point on the material surface within the neighborhood with a distance less than . Use the non-maximum suppression strategy to eliminate redundant points;

[0157] S4.3. Establish a buffer at the junction of the boundary and the boundary . Fit the transition surface by the moving least squares method to achieve the continuous fusion of the material point cloud and the wall point cloud below the material surface. The formula is as follows:

[0158] ;

[0159] where and are two independent variables for parameterizing the surface, is the distance weight function, defined as , Represents the parameter space Mapped to a point in three-dimensional space Is the point The shortest Euclidean distance to the boundary of the material surface Is the buffer width, controlling the transition range of the fusion area And Respectively represent the local fitting surfaces of the material point cloud and the silo wall point cloud Represents the transition surface of the fusion area

[0160] S4.4. Adopt a multi-resolution strategy based on Poisson disk sampling to adaptively adjust the point spacing in the fusion area , where Is the resampled point spacing, adaptively taking the minimum spacing between the material point cloud and the silo wall point cloud below the material surface Is the average point spacing of the material point cloud Is the average point spacing of the silo wall point cloud below the material surface, maintaining the uniformity of point distribution and avoiding point aggregation or sparsity

[0161] S4.5. Apply anisotropic curvature diffusion filtering to eliminate local distortions caused by noise

[0162] ;

[0163] Among them Is the point And its neighborhood points Of the Gaussian curvature, quantifying the local surface curvature Is the normal vector of the normal neighborhood point , indicating the local direction of the surface Is the point Of the neighborhood point set (based on KD-Tree search) Is the Three-dimensional coordinates of the th point

[0164] Through the above steps, the material surface part of the real-time cavity point cloud and the silo wall part in the reference point cloud are fused together, and the edge position at the fusion is optimized through distance weight and geometric consistency enhancement, maintaining the geometric integrity at the edge

[0165] S5. Calculate the volume of the material points to be calculated; The volume calculation uses the method of meshing the point cloud, converting the point cloud into a three-dimensional mesh model to obtain the model volume of the grain silo. The specific steps are as follows

[0166] S5.1. Perform surface reconstruction on the material point cloud to be calculated and construct a tetrahedral mesh ; Using the Delaunay triangulation method, all points in the point cloud satisfy the empty sphere criterion, that is, the circumsphere of any tetrahedron does not contain other vertices, and a tetrahedral mesh is constructed (formed by the tetrahedral point set ), ensuring the integrity of the geometric structure of the point cloud; extracting the surface manifold through convex hull cutting and the α - shape algorithm (geometric constraint conditions for screening surface manifolds and removing internal tetrahedrons):

[0167] ;

[0168] Among them, is a single triangular patch in the surface manifold, formed by points ,, is the surface manifold;

[0169] S5.2. Optimize and repair the tetrahedral mesh ; Detect non - manifold edges (shared by more than two faces) and isolated vertices between boundary patches, and apply edge collapse and vertex split operations. The optimization goal is to minimize the mesh geometric error and topological complexity:

[0170] ;

[0171] Among them, is the regularization coefficient, balancing the geometric error and topological complexity, is the original mesh, is the optimized mesh, () is the topological complexity of the mesh (such as the number of holes);

[0172] S5.3. Calculate the volume of the material using the tetrahedral mesh ; Select an internal base point , and form a tetrahedron with each surface triangle , through , where, Among them, is the number of tetrahedrons, is the edge vector of the n - th tetrahedron.

[0173] Through the above steps, a surface topological structure close to the real material surface can be generated in the case of a complex material surface, and the problems of surface mesh repetition and incompleteness are avoided.

[0174] For calculating the volume of the material point cloud to be calculated, the slicing method can also be used. The slicing method slices the material point cloud to be calculated along a certain direction, calculates the area of each slice, and then adds up the areas of all slices and multiplies by the thickness of the slice to obtain the total volume. This is the prior art and will not be elaborated here.

[0175] It should be understood that the present invention can be used not only for measuring the volume of a granary, but also for measuring the volume of other materials. It can be used not only for cylindrical granaries, but also for granaries of other shapes. Currently, almost all granaries are cylindrical granaries.

[0176] The adaptive incomplete deformation cavity three-dimensional point cloud volume calculation system proposed in the embodiment of the present invention is used to calculate the volume of the cavity three-dimensional point cloud, and includes:

[0177] A data acquisition module for automatically collecting or manually importing point cloud data;

[0178] A data processing module for calculating the volume according to the point cloud data in the data acquisition module;

[0179] A data transmission module for transmitting the volume of the cavity three-dimensional point cloud calculated by the data processing module to the terminal;

[0180] The volume calculation method adopted in the data processing module is the above-mentioned adaptive incomplete deformation cavity three-dimensional point cloud volume calculation method.

[0181] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0182] Obtain a reference point cloud; the reference point cloud is a directly measured empty warehouse point cloud or a virtual empty warehouse point cloud jointly generated by historical cavity point clouds and granary design drawings;

[0183] Segment the real-time cavity point cloud to obtain the material surface point cloud, and divide the reference point cloud into the warehouse wall point cloud above the material surface and the warehouse wall point cloud below the material surface through the material surface point cloud;

[0184] Fuse the warehouse wall point cloud below the material surface in the segmented reference point cloud and the material surface point cloud of the real-time cavity point cloud to obtain the material point cloud to be calculated;

[0185] Calculate the volume of the material point cloud to be calculated.

[0186] It should be noted that in the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0187] Those skilled in the art will appreciate that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0188] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0189] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0190] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0191] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0192] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. An adaptive method for calculating the volume of a cavity three-dimensional point cloud with incomplete deformation, characterized in that, Including the following steps: Obtain a reference point cloud; the reference point cloud is an empty bin point cloud directly measured or a virtual empty bin point cloud jointly generated by historical cavity point clouds and a granary design drawing; Segment the real-time cavity point cloud to obtain a material surface point cloud, and segment the reference point cloud into a bin wall point cloud above the material surface and a bin wall point cloud below the material surface through the material surface point cloud; Fuse the bin wall point cloud below the material surface in the segmented reference point cloud and the material surface point cloud of the real-time cavity point cloud to obtain a material point cloud to be calculated; Calculate the volume of the material point cloud to be calculated; The step of segmenting the real-time cavity point cloud to obtain a material surface point cloud and segmenting the reference point cloud into a bin wall point cloud above the material surface and a bin wall point cloud below the material surface through the material surface point cloud includes the following steps: Perform spatial registration on the real-time cavity point cloud and the reference point cloud; Perform differential distance field calculation and spatial mapping on the real-time cavity point cloud; Perform adaptive threshold segmentation on the material surface and the bin wall of the real-time cavity point cloud; Perform morphological optimization and topological verification on the material surface point cloud; Based on the material surface point cloud of the cavity point cloud, segment the reference point cloud into a bin wall point cloud above the material surface and a bin wall point cloud below the material surface; The step of segmenting the reference point cloud into a bin wall point cloud above the material surface and a bin wall point cloud below the material surface based on the material surface point cloud of the cavity point cloud includes: Material surface height field modeling; calculating the height distribution field from the point cloud of the material : ； Among them, is the horizontal projection neighborhood of the material surface; is the point on the material surface and is the vertical coordinate; Based on the height distribution field , perform regional segmentation on the reference point cloud: (a)Point cloud of the bin wall below the material surface: ; (b) Point cloud of the bin wall above the material surface: ; Among them, is the reference point cloud, is the point on the reference point cloud 's vertical coordinate, is the noise standard deviation; The step of fusing the bin wall point cloud below the material surface in the segmented reference point cloud and the material surface point cloud of the real-time cavity point cloud to obtain a material point cloud to be calculated includes the following steps: Extract the boundary of the material surface point cloud through the α-shape algorithm and the boundary of the bin wall point cloud below the material surface , where is the point on the boundary of the material surface point cloud, is the member function of α-shape; Construct a KD-Tree acceleration structure for the boundary Perform a radius r search , is the threshold for the radius search, mark the overlapping area with the boundary Overlapping area , is the point on the boundary of the reference point cloud within the overlapping area is the point on the material surface within the neighborhood with a distance less than the th point, and use the non-maximum suppression strategy to eliminate redundant points At the boundary With the boundary Establish a buffer zone at the junction , and realize the continuous fusion of the material point cloud and the bin wall point cloud below the material surface by fitting the transition surface with the moving least squares method. The formula is as follows: ； Among them, and are two independent variables for parameterizing the surface, is the distance weight function, defined as , represents the parameter space mapped to a point in three-dimensional space, is the point to the shortest Euclidean distance of the material surface boundary, is the buffer width, controlling the transition range of the fusion area, and respectively represent the local fitting surfaces of the material point cloud and the silo wall point cloud, represents the transition surface of the fusion area; Adopt a multi-resolution strategy based on Poisson disk sampling to adaptively adjust the point spacing in the fusion area , where is the resampled point spacing, and adaptively take the minimum spacing between the material surface point cloud and the bin wall point cloud below the material surface, is the average point spacing of the material surface point cloud, is the average point spacing of the bin wall point cloud below the material surface; Apply anisotropic curvature diffusion filtering to eliminate local distortions caused by noise: ； Among them, is the point and its neighboring points of the Gaussian curvature, which quantifies the local surface curvature, is the normal vector of the neighboring point indicating the local direction of the surface, is the point of the set of neighboring points, is the three-dimensional coordinates of the nth point.

2. The three-dimensional point cloud volume calculation method for an adaptive deformed cavity with defects according to claim 1, characterized in that The step of obtaining the reference point cloud includes: If an empty bin point cloud can be directly measured, use the measured empty bin point cloud as the reference point cloud, or directly select a historical cavity point cloud that meets the requirements of the first volume confidence interval as the reference point cloud; otherwise, construct a virtual empty bin point cloud through the historical cavity point cloud and the geometric parameters of the granary as the reference point cloud.

3. The three-dimensional point cloud volume calculation method for a cavity with adaptive incomplete deformation according to claim 2, wherein, The step of constructing a virtual empty bin point cloud through the historical cavity point cloud and the geometric parameters of the granary as the reference point cloud includes: Select a historical cavity point cloud that meets the requirements of the second volume confidence interval; Align the axis of the historical cavity point cloud with the central axis of the granary design drawing; Perform data fusion on the axially aligned historical cavity point cloud and the design parameters of the granary to obtain a virtual empty bin point cloud; Perform error compensation on the virtual empty bin point cloud to obtain a reference point cloud; 4. The method for calculating the volume of a three-dimensional point cloud of a cavity with self-adaptive incomplete deformation according to claim 3, wherein, The first volume confidence interval is ; the second volume confidence interval is , is the volume confidence level, defined as: ； Among them, is the volume of a certain historical cavity point cloud, is the theoretical volume in the empty bin state.

5. The three-dimensional point cloud volume calculation method for an adaptive deformed cavity with defects according to claim 3, characterized in that The step of performing data fusion on the axially aligned historical cavity point cloud and the design parameters of the granary to obtain a virtual empty bin point cloud is specifically: The virtual empty bin point cloud to be reconstructed includes a measured section and an extended section. The measured section directly uses the bin points in the historical cavity point cloud. Wall The extended section is divided into a side wall and a bin bottom. The side wall is axially stretched based on the side wall height in the grain bin design drawing. It is formed by axial stretching: ； ； Among them, is the side wall height of the historical cavity point cloud, is the difference between the side wall height of the granary design drawing and the side wall height of the historical cavity point cloud, is the stretching coefficient, are the virtual bin wall point coordinates of the side wall of the extension section, is the coordinate of the highest point of the measured section, is the granary axis; The bottom of the bin is a flat bottom or a conical bottom. The flat bottom is based on the height of the design drawing. The projection sidewall bottom boundary generates a uniformly rasterized point cloud; the conical bottom is driven by the design drawing parameters and generates a point cloud based on the discretization of the parametric analytical surface, which is expressed as: ； Among them, is the conical silo bottom, is the drawing cone top, is the drawing half angle, is the drawing radius, is the radial distance, is the azimuth angle, is perpendicular to the axial direction and is the plane orthogonal basis vector; Realize the topological closure fusion of the side wall-bottom point cloud through normal vector consistency constraint and the nearest neighbor interpolation algorithm based on KD-Tree; Stitch the measured section and the extended section to obtain a virtual empty bin point cloud.

6. The three-dimensional point cloud volume calculation method for an adaptive deformed cavity with defects according to claim 1, wherein After the historical cavity point cloud and the real-time cavity point cloud are obtained, preprocessing operations need to be performed on them, including the following steps: Downsample the granary measurement point cloud and remove outliers; the historical point cloud and the real-time cavity point cloud are called the granary measurement point cloud; Evaluate and repair the integrity of the granary measurement point cloud.

7. A three-dimensional point cloud volume calculation system for an adaptive deformed cavity with defects, characterized in that, The system is used to calculate the volume of the three-dimensional point cloud of the cavity, including: A data acquisition module for automatically collecting or manually importing point cloud data; A data processing module, configured to calculate the volume based on the point cloud data in the data acquisition module; A data transmission module, configured to transmit the volume of the three-dimensional point cloud of the cavity calculated by the data processing module to the terminal; The volume calculation method adopted in the data processing module is an adaptive deformed cavity three-dimensional point cloud volume calculation method described in any one of claims 1-6.

Images