Three-dimensional model reconstruction method and system

Abstract

The invention provides a three-dimensional model reconstruction method and system, and relates to the technical field of three-dimensional imaging. The method comprises the following steps: performing one-dimensional interval division on multiple curve sample data to obtain curve probability distribution and construct a curve probability density function; performing two-dimensional interval division on the plurality of curved surface sample data to obtain curved surface probability distribution and construct a curved surface probability density function; performing weight distribution on the acquired curve target data through a curve probability density function to obtain reconstructed curve target data and construct a three-dimensional model of the target object; and performing weight distribution on the curved surface target data through the curved surface probability density function to obtain reconstructed curved surface target data and construct a three-dimensional model of the target object. The method comprises the following steps: preparing enough sample data for a sample object to approach actual data, estimating accurate distribution of a probability function model of the sample object in a three-dimensional space, performing data fitting on the sample data, and efficiently reconstructing a three-dimensional model by using a constructed probability density function in an actual measurement stage.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of three-dimensional imaging, and particularly relates to a three-dimensional model reconstruction method and system. Background Art

[0002] 3D imaging technology is a technology that can acquire the three-dimensional spatial information of an object and perform imaging, that is, construct a three-dimensional model of a three-dimensional structure and display it on a terminal, and plays an extremely important role in many fields such as industry, military, and medicine. The key steps of three-dimensional modeling and imaging include data acquisition and data processing. Traditional data acquisition usually has problems such as low accuracy of the acquired data, incomplete data, and complex acquisition algorithms; in the traditional data processing process, due to the improvement of the accuracy of the acquisition device and the expansion of the range, the data volume has increased exponentially, and there is a problem that massive data cannot be efficiently processed and useful information cannot be extracted. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a three-dimensional model reconstruction method and system in view of the deficiencies of the prior art.

[0004] The technical solution of the present invention to solve the above technical problem is as follows:

[0005] A three-dimensional model reconstruction method includes the following steps:

[0006] Import a plurality of acquired one-dimensional curve sample data, perform one-dimensional interval division on the plurality of curve sample data to obtain a curve probability distribution, and construct a curve probability density function based on the curve probability distribution;

[0007] Import a plurality of acquired two-dimensional surface sample data, perform two-dimensional interval division on the plurality of surface sample data to obtain a surface probability distribution, and construct a surface probability density function based on the surface probability distribution;

[0008] Collect data of a target object to obtain a plurality of target data;

[0009] If the plurality of target data are a plurality of one-dimensional curve target data, weight distribution is respectively performed on the plurality of curve target data through the curve probability density function to obtain a plurality of reconstructed curve target data, and a three-dimensional model of the target object is constructed from the plurality of reconstructed curve target data;

[0010] If the plurality of target data are a plurality of two-dimensional surface target data, weight distribution is respectively performed on the plurality of surface target data through the surface probability density function to obtain a plurality of reconstructed surface target data, and a three-dimensional model of the target object is constructed from the plurality of reconstructed surface target data.

[0011] Another technical solution of the present invention to solve the above technical problem is as follows:

[0012] A three-dimensional model reconstruction system, comprising:

[0013] A curve calibration module, configured to import a plurality of collected one-dimensional curve sample data, perform one-dimensional interval division on the plurality of curve sample data to obtain a curve probability distribution, and construct a curve probability density function based on the curve probability distribution;

[0014] A surface calibration module, configured to import a plurality of collected two-dimensional surface sample data, perform two-dimensional interval division on the plurality of surface sample data to obtain a surface probability distribution, and construct a surface probability density function based on the surface probability distribution;

[0015] A data acquisition module, configured to acquire data of a target object to obtain a plurality of target data;

[0016] A curve reconstruction module, configured to, if the plurality of target data are a plurality of one-dimensional curve target data, respectively perform weight assignment on the plurality of curve target data through the curve probability density function to obtain a plurality of reconstructed curve target data, and construct a three-dimensional model of the target object from the plurality of reconstructed curve target data;

[0017] A surface reconstruction module, configured to, if the plurality of target data are a plurality of two-dimensional surface target data, respectively perform weight assignment on the plurality of surface target data through the surface probability density function to obtain a plurality of reconstructed surface target data, and construct a three-dimensional model of the target object from the plurality of reconstructed surface target data.

[0018] The beneficial effects of the present invention are: by importing a large amount of sample data, fitting the sample data in regions, estimating the exact distribution of the probability model based on the sample space, constructing a corresponding probability density function, approximating the actual data of the three-dimensional object as accurately as possible, thereby selecting the corresponding probability density function to perform three-dimensional model reconstruction on the target data of the target object, making the obtained three-dimensional image more accurate, improving the reconstruction accuracy and efficiency, reducing the calculation amount, surpassing the traditional method in terms of accuracy, efficiency and generality, and being particularly suitable for processing non-uniform and high-noise real data. Description of the Drawings

[0019] Figure 1 It is a flowchart of the three-dimensional model reconstruction method provided by an embodiment of the present invention;

[0020] Figure 2 It is a structural diagram of the three-dimensional model reconstruction method provided by an embodiment of the present invention;

[0021] Figure 3 It is a schematic diagram of the curve model data probability density function in a three-dimensional space provided by an embodiment of the present invention;

[0022] Figure 4 Schematic diagram of the probability density function of the surface model data in three-dimensional space provided by the embodiment of the present invention;

[0023] Figure 5 XOY plane view of the probability density function of the surface model data in three-dimensional space provided by the embodiment of the present invention;

[0024] Figure 6 Block diagram of the modules of the three-dimensional model reconstruction system provided by the embodiment of the present invention. Detailed implementation manners

[0025] The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0026] The data acquisition methods for three-dimensional modeling and imaging are structured light method and laser scanning method. The laser scanning method has higher accuracy and is widely used in lidar. In the traditional method of three-dimensional imaging based on lidar point shooting technology, the lidar point shooting technology cannot traverse all data points of the target object and can only measure a countable number of points. However, the actual data is distributed in a continuous interval, which is infinitely unmeasurable. Moreover, due to objective factors such as mechanical structure, motor accuracy, circuit noise, and power supply, the measured data is also contaminated by noise; when constructing a three-dimensional model according to the data points, there are disadvantages such as large amount of data, large amount of calculation, and low accuracy.

[0027] As Figure 1 and Figure 2 shown, a three-dimensional model reconstruction method provided by the embodiment of the present invention includes the following steps:

[0028] Import a plurality of collected one-dimensional curve sample data, perform one-dimensional interval division on the plurality of curve sample data to obtain a curve probability distribution, and construct a curve probability density function based on the curve probability distribution;

[0029] Import a plurality of collected two-dimensional surface sample data, perform two-dimensional interval division on the plurality of surface sample data to obtain a surface probability distribution, and construct a surface probability density function based on the surface probability distribution;

[0030] Collect data of the target object to obtain a plurality of target data;

[0031] If the plurality of target data are a plurality of one-dimensional curve target data, then weight distribution is performed on the plurality of curve target data respectively through the curve probability density function to obtain a plurality of reconstructed curve target data, and a three-dimensional model of the target object is constructed from the plurality of reconstructed curve target data;

[0032] If multiple pieces of the target data are multiple two-dimensional curved surface target data, weight distribution is performed on the multiple pieces of curved surface target data respectively through the curved surface probability density function to obtain multiple reconstructed curved surface target data, and a three-dimensional model of the target object is constructed from the multiple pieces of reconstructed curved surface target data.

[0033] Among them, the steps of constructing the curve probability density function and the curved surface probability density function are classified as the calibration stage, and the steps of collecting data of the target object and constructing a three-dimensional model of the target object through the curve probability density function or the curved surface probability density function are classified as the actual measurement stage.

[0034] Starting from limited test data, approximating the actual data as accurately as possible. Through the previous initialization data preparation stage (i.e., the calibration stage), prepare enough samples to estimate the exact distribution of the probability function model based on this sample space, and perform data fitting accordingly, so as to obtain fast and accurate 3D imaging data (i.e., reconstructed curve target data or reconstructed curved surface target data) in the actual measurement stage. That is, approximate the actual data through limited sample data, prepare enough sample data based on the sample object to estimate the exact distribution of the probability function model of the sample object in the three-dimensional space, perform data fitting on the sample data, and use the constructed probability density function to efficiently reconstruct the three-dimensional model in the actual measurement stage.

[0035] In the embodiments of the present invention, a radar point scanning technology is used to obtain a three-dimensional image. By importing a large amount of sample data, fitting the sample data in regions, estimating the exact distribution of the probability model based on the sample space, constructing the corresponding probability density function, approximating the actual data of the three-dimensional object as accurately as possible, and then selecting the corresponding probability density function to reconstruct the three-dimensional model of the target data of the target object, and displaying the basic shape of the image on the PC side. The obtained three-dimensional image is more accurate, can improve the reconstruction accuracy and efficiency, and reduce the amount of calculation, surpassing the traditional method in terms of accuracy, efficiency and versatility, and is especially suitable for processing non-uniform and high-noise real data.

[0036] Preferably, collecting multiple one-dimensional curve sample data specifically includes:

[0037] Using the radar point scanning technology to scan a known standard target object, and collecting enough curve sample data {x1′, x2′,..., x′ M}, a total of M data, and the data length M is set according to needs.

[0038] Preferably, the one-dimensional interval division of the multiple pieces of curve sample data to obtain the curve probability distribution includes:

[0039] Select the maximum curve value and the minimum curve value from multiple one-dimensional curve sample data, calculate the difference between the maximum curve value and the minimum curve value to obtain the total curve interval length;

[0040] Calculate the ratio of the total curve interval length to the set number of curve intervals to obtain the curve interval interval. Divide the total curve interval length into multiple curve intervals according to the curve interval interval, and each of the curve intervals includes multiple corresponding curve sample data;

[0041] Take the midpoint value of any one of the curve intervals as the curve interval sequence, count the number of curve sample data in the curve interval to obtain the curve distribution number, calculate the ratio of the curve distribution number to the number of multiple curve sample data to obtain the curve distribution probability; in this process, calculate for each of the curve intervals to obtain the curve distribution probability corresponding to each of the curve intervals, and the curve interval sequence and the corresponding curve distribution probability of each of the curve intervals form the curve probability distribution.

[0042] Specifically, calculate the minimum value x M from the curve sample data {x1′, x2′,..., x′ min} = min{x′1, x′2,..., x′ M} and the maximum value x max = max{x′1, x′2,..., x′ M}, and obtain the total curve interval length L = x max - x min by taking the difference between the maximum value and the minimum value. Preset the number of curve intervals N x , calculate the ratio of the total curve interval length to the number of curve intervals to obtain the interval interval length Δ, that is:

[0043]

[0044] Divide the total curve interval length L by the interval interval length Δ. Let a = x min , b = x max , then the i-th curve interval is [a + (i - 1)Δ, a + iΔ], and the curve sample data at the midpoint of the i-th curve interval is denoted as x i , then x j ′ ∈ {x1′, x′2,..., x′ M}, j = 1, 2,..., M, traverse the curve sample data {x′1, x′2,..., x′ M}, and if a + (i - 1)Δ ≤ x′ j < a + iΔ, it means that x′ jFalling within the i-th curve interval to obtain the number of samples N falling within the i-th curve interval i (i.e., the number of curve distributions). The curve distribution probability p of the sample data falling within the i-th curve interval i is as follows:

[0045]

[0046] The probability distribution is as follows:

[0047]

[0048] where X is the curve interval and p i is the curve distribution probability.

[0049] In the embodiments of the present invention, the one-dimensional data is divided into multiple one-dimensional intervals according to the interval spacing, which is convenient for analyzing the distribution of data in the model. Each collected one-dimensional curve sample data is traversed to obtain the probability values falling within each interval, so as to construct a probability space.

[0050] Preferably, constructing the curve probability density function based on the curve probability distribution includes:

[0051] Deriving that the distribution of the curve probability approximately follows a normal distribution according to the central limit theorem, and screening out the starting point, the first curve inflection point, the curve peak point, the second curve inflection point and the end point from the curve probability distribution;

[0052] Respectively taking the starting point, the first curve inflection point, the curve peak point, the second curve inflection point and the end point as regional nodes, and dividing the curve probability distribution into multiple curve regions;

[0053] Constructing an initial parabola equation, and respectively calculating the initial parabola equation through the curve interval sequences of multiple curve regions and the corresponding curve distribution probabilities to obtain the curve probability density function.

[0054] It should be understood that since there are many random factors in the noise source of the acquisition system, for the sample space (i.e., the one-dimensional curve data in the three-dimensional space), according to the central limit theorem, the requirement for the curve probability density function in the three-dimensional space to converge to the normal distribution is that the random factors are independent and the result of taking the limit as the number tends to infinity. However, the actual system random factors are not completely independent, the number of random factors cannot tend to infinity, and the symmetry of the probability density function can only exist in theoretically ideal data. Therefore, for the characteristics of the distribution of the actually collected curve sample data, that is, the actual curve sample data is concentrated in a finite interval near the peak, and its probability density function tends to the normal distribution but shows asymmetry. At the same time, considering the limited computing resources, a data processing technology based on piecewise parabola fitting is used to construct the curve probability density function. The normal distribution function to which the curve probability density function converges is:

[0055]

[0056] Among them, f(x) is the normal distribution function to which the curve probability density function converges, x is the curve sample data, σ is the standard deviation of the curve sample data, and μ is the mean value of the curve sample data.

[0057] Preferably, screening out the starting point, the first curve inflection point, the curve peak point, the second curve inflection point and the end point from the curve probability distribution includes:

[0058] Screen out the maximum curve distribution probability from the curve probability distribution, take the curve interval sequence corresponding to the maximum curve distribution probability as the curve peak point, take the first curve interval sequence as the starting point, and take the last curve interval sequence as the end point;

[0059] Calculate multiple curve sample data within the range from the starting point to the curve peak point through a pre-constructed curve central difference formula to obtain multiple first curve difference results, screen out the minimum value of the multiple first curve difference results to obtain the first difference minimum value, and take the curve interval sequence corresponding to the curve sample data corresponding to the first difference minimum value as the first curve inflection point;

[0060] Calculate multiple curve sample data within the range from the curve peak point to the end point through a pre-constructed curve central difference formula to obtain multiple second curve difference results, screen out the minimum value of the multiple second curve difference results to obtain the second difference minimum value, and take the curve interval sequence corresponding to the curve sample data corresponding to the second curve difference minimum value as the second curve inflection point.

[0061] Specifically, estimate the inflection point data of the curve sample data within a given range through a curve central difference formula, and the curve central difference formula is:

[0062]

[0063] Among them, S is the curve difference result, p is the curve distribution probability, and x i is the curve interval;

[0064] The minimum value corresponding point x that satisfies the curve central difference formula is determined as the inflection point. i

[0065] Within the range from the starting point to the curve peak point, that is, x1 ≤ x j < x peak Within the range, the corresponding curve distribution probability and curve interval are solved through the curve central difference formula, and the inflection point x inflexion1 is calculated; within the range from the curve peak point to the end point, that is, x peak ≤ x o < x N0 Within the range, the corresponding curve distribution probability and curve interval are solved through the curve central difference formula, and the inflection point x inflexion2 is calculated.

[0066] In the embodiments of the present invention, the inflection point where the probability distribution changes is found in the curve interval of the curve sample data through the curve central difference formula, and the curve interval of the data distribution is divided into multiple regions according to the two end points, the inflection point, and the peak point, so as to improve the fitting accuracy for constructing the curve probability density function.

[0067] Preferably, as Figure 3 shown, respectively taking the starting point, the first curve inflection point, the curve peak point, the second curve inflection point, and the end point as regional nodes, and dividing the curve probability distribution into multiple curve regions, including:

[0068] Dividing the range from the starting point x1 to the first curve inflection point x inflexion1 into the first curve region D1, dividing the range from the first curve inflection point x inflexion1 to the curve peak point x peak into the second curve region D2, dividing the range from the curve peak point x peak to the second curve inflection point, x inflexion2 into the third curve region D3, and dividing the range from the second curve inflection point, x inflexion2 to the end point x N0 into the fourth curve region D4.

[0069] In the embodiments of the present invention, the curve probability density function tends to be normally distributed but shows asymmetry, and it is reasonable that most of the actually collected sample data is concentrated in the D1, D2, D3, and D4 regions near the peak.

[0070] ​Preferably, to construct the initial parabola equation, the initial parabola equation is calculated respectively through the curve interval sequences of the plurality of curve regions and the corresponding curve distribution probabilities to obtain a curve probability density function, including:

[0071] Count the number of the curve sample data in each of the curve regions respectively to obtain the regional sample numbers corresponding to the respective curve regions;

[0072] For any one of the curve regions, construct an initial parabola equation, substitute the regional sample number in the curve region and the curve distribution probability corresponding to the curve interval sequence into the initial parabola equation for calculation respectively to obtain a first design matrix, a first response vector and a first coefficient matrix, calculate the first design matrix, the first response vector and the first coefficient matrix through the least square method to obtain a curve coefficient matrix, and update the initial parabola equation according to the curve coefficient matrix to obtain a parabola equation; in this process, process all the curve regions to obtain a plurality of parabola equations, and the curve probability density function is composed of the plurality of parabola equations.

[0073] Specifically, counting the number of the curve sample data in each of the curve regions respectively to obtain the regional sample numbers corresponding to the respective curve regions includes:

[0074] Count the number of the curve sample data in the first curve region to obtain a first regional sample number; count the number of the curve sample data in the second curve region to obtain a second regional sample number; count the number of the curve sample data in the third curve region to obtain a third regional sample number; count the number of the curve sample data in the fourth curve region to obtain a fourth regional sample number. It can be understood that the region K where the curve sample data is distributed is divided into four parts, including the first curve region D1 ∈ [x 1, x inflexion1 ), and N1 curve sample data are counted; the second curve region D2 ∈ [x inflexion1, x peak ), and N2 curve sample data are counted; the third curve region D3 ∈ [x oeak , x inflexion2 ), and N3 curve sample data are counted; the fourth curve region D4 ∈ [x inflexion2, x N0 , and N4 curve sample data are counted.

[0075] Counting the number of the curve sample data in each curve region facilitates subsequent probability distribution fitting according to the corresponding curve sample data with the corresponding quantity.

[0076] Specifically, due to limited computing resources and the inability of linear functions to fully cover all curve sample data, while higher-degree functions of degree three and above have problems such as high computational complexity and large error accumulation in the calculation process, a parabola of a quadratic function is selected to match and approximate the property that the probability density function is a bell curve (i.e., an approximately normal distribution curve).

[0077] In the subspace within the first curve region D1, the parabola equation is set as:

[0078] f(x) = a1x 2 + b1x + c1,

[0079] where a1, b1, and c1 are coefficients to be determined;

[0080] Construct the design matrix and response vector. Within the first curve region D1, use N1 curve sample data to construct the design matrix X and construct the response vector P according to the corresponding curve distribution probability, that is, substitute the curve interval sequence of the number of region samples within the curve region and the corresponding curve distribution probability into the initial parabola equation for calculation, and obtain the fitting expression for the first curve region as:

[0081]

[0082] Briefly recorded as: Xβ1 = P.

[0083] Solve the coefficient matrix β1 of the fitting expression for the first curve region by the least squares method to obtain the curve coefficient matrix for the first curve region, and update the parabola equation according to the curve coefficient matrix to obtain the parabola equation for the first curve region; the calculation expression of the least squares method is:

[0084] β1 = (X T P) -1 X T P.

[0085] Similarly, calculate the parabola equations for the second curve region D2, the third curve region D3, and the fourth curve region D4, specifically:

[0086] In the subspace within the second curve region D2, the parabola equation is set as:

[0087] f(x) = a2x 2 + b2x + c2,

[0088] where a2, b2, and c2 are coefficients to be determined.

[0089] Construct a design matrix X using N2 curve sample data and construct a response vector P according to the corresponding curve distribution probability, that is, substitute the curve interval sequence of the number of regional samples in the curve region and the corresponding curve distribution probability into the initial parabola equation for calculation, and obtain the fitting expression for the second curve region as follows:

[0090]

[0091] Briefly recorded as: Xβ2 = P.

[0092] Solve the coefficient matrix β2 of the fitting expression for the second curve region by the least squares method to obtain the curve coefficient matrix for the second curve region, and update the parabola equation according to the curve coefficient matrix to obtain the parabola equation for the second curve region; the calculation expression of the least squares method is:

[0093] β2 = (X T P) -1 X T P.

[0094] In the subspace within the third curve region D3, set the parabola equation as:

[0095] f(x) = a3x 2 + b3x + c3,

[0096] where a3, b3, and c3 are coefficients to be determined.

[0097] Construct a design matrix X using N3 curve sample data and construct a response vector P according to the corresponding curve distribution probability, that is, substitute the curve interval sequence of the number of regional samples in the curve region and the corresponding curve distribution probability into the initial parabola equation for calculation, and obtain the fitting expression for the third curve region as follows:

[0098]

[0099] Briefly recorded as: Xβ3 = P.

[0100] Solve the coefficient matrix β3 of the fitting expression for the third curve region by the least squares method to obtain the curve coefficient matrix for the third curve region, and update the parabola equation according to the curve coefficient matrix to obtain the parabola equation for the third curve region; the calculation expression of the least squares method is:

[0101] β3 = (X T P) -1 X T P.

[0102] In the subspace within the fourth curve region D4, set the parabola equation as:

[0103] f(x) = a4x 2 + b4x + c4,

[0104] where a4, b4, and c4 are coefficients to be determined.

[0105] Use N4 curve sample data to construct the design matrix X and construct the response vector P according to the corresponding curve distribution probability, that is, substitute the curve interval sequence of the number of region samples in the curve region and the corresponding curve distribution probability into the initial parabola equation for calculation, and the fitting expression for the fourth curve region is obtained as:

[0106]

[0107] Briefly recorded as: Xβ4 = P.

[0108] Solve the coefficient matrix β3 of the fitting expression for the fourth curve region by the least squares method to obtain the curve coefficient matrix of the fourth curve region, and update the parabola equation according to the curve coefficient matrix to obtain the parabola equation of the fourth curve region; the calculation expression of the least squares method is:

[0109] β4 = (X T P) -1 X T P.

[0110] In the above calculation expression of the least squares method, X T P is the product of the transpose of the design matrix and the response vector, and (X T P) -1 is the inverse matrix of X T P.

[0111] The curve probability density function is composed of multiple parabola equations (that is, the parabola equations of the first curve region, the second curve region, the third curve region, and the fourth curve region), and the expression of the curve probability density function is:

[0112]

[0113] In the embodiment of the present invention, in the calibration stage for the curve sample data, a four-segment quadratic function is used to fit the collected data. This segmented fitting method not only significantly reduces the amount of calculation, but also effectively reduces the overall error by optimizing the fitting accuracy of each segment, thereby achieving higher modeling accuracy.

[0114] Preferably, collect multiple two-dimensional surface sample data, specifically:

[0115] Use the radar point-shot scanning technology to scan a known standard target object to collect enough surface sample data (x′ i , y′ i), a total of M data, where M is set as needed.

[0116] Preferably, the two-dimensional interval division of the multiple surface sample data to obtain the surface probability distribution includes:

[0117] Select the first maximum surface value, the first minimum surface value, the second maximum surface value, and the second minimum surface value from the multiple two-dimensional surface sample data, calculate the difference between the first maximum surface value and the first minimum surface value to obtain the first total surface interval length, and calculate the difference between the second maximum surface value and the second minimum surface value to obtain the second total surface interval length;

[0118] Calculate the ratio of the first total surface interval length to the set number of the first surface intervals to obtain the first surface interval interval, calculate the ratio of the second total surface interval length to the set number of the second surface intervals to obtain the second surface interval interval, divide the first total surface interval length into multiple first surface intervals according to the first surface interval interval, divide the second total surface interval length into multiple second surface intervals according to the second surface interval interval, and each first surface interval and each second surface interval include multiple corresponding surface sample data;

[0119] Take the midpoint value of any one of the first surface intervals as the first surface interval sequence, take the midpoint value of the second surface interval as the second surface interval sequence, count the number of surface sample data in the first surface interval and the second surface interval to obtain the surface distribution number, calculate the ratio of the surface distribution number to the number of the multiple surface sample data to obtain the surface distribution probability; in this process, calculate for each of the first surface intervals and the second surface intervals to obtain the surface distribution probability corresponding to each surface interval, and the surface interval sequence and the corresponding surface distribution probability of each surface interval form the surface probability distribution.

[0120] Specifically, calculate the minimum value and the maximum value from the surface sample data (x′ i , y′ i ), x min = min{x′ i}, x max = max{x′ i}, y min = min{y′ i}, y max = max{y′ i}. Calculate the difference between the maximum value and the minimum value to obtain the first total surface interval length L x = x max - x min and the second total surface interval length Ly = y max -y min , preset the number N of the first surface intervals x and the number N of the first surface intervals y , calculate the ratio of the total interval length of the first surface to the number of the first surface intervals to obtain the first surface interval spacing Δ x , calculate the ratio of the total interval length of the second surface to the number of the second surface intervals to obtain the second surface interval spacing Δ y , that is:

[0121]

[0122] Through the first surface interval spacing Δ x divide the total interval length L of the first surface x , let a x = x min , b x = x max , then the i-th surface interval is [a x +(i - 1)Δ x , a x +iΔ x , and the surface sample data at the midpoint of the i-th surface interval is denoted as x i , then Through the second surface interval spacing Δ y divide the total interval length L of the second surface y , let a y = y min , b y = y max , then the j-th surface interval is [a y +(j - 1)Δ y , a y +jΔ y , and the surface sample data at the midpoint of the j-th surface interval is denoted as y j , then Traverse all surface sample data (x i ′, y i ′), and count that the number of samples falling into each two-dimensional interval [a x +(i - 1)Δ x , a x +iΔ x ×[ay + (j - 1)Δ y , a y +jΔ y is N ij (i.e., the number of surface distributions). The surface distribution probability that the sample data falls into the (i, j)-th surface interval is:

[0123]

[0124] The probability distribution of the surface sample data in each two-dimensional interval is as follows:

[0125]

[0126] Wherein, X is the surface interval, and p ij is the surface distribution probability.

[0127] In the embodiment of the present invention, the two-dimensional data is divided into multiple two-dimensional intervals according to the interval spacing, which is convenient for analyzing the distribution of the data in the model. Each collected two-dimensional surface sample data is traversed to obtain the probability values falling in each interval, so as to construct a probability space.

[0128] Preferably, constructing the surface probability density function based on the surface probability distribution includes:

[0129] Screening out the surface peak points from the surface probability distribution, establishing a space rectangular coordinate system based on the surface peak points, and dividing the surface probability distribution into multiple initial surface regions according to the coordinate axes of the space rectangular coordinate system;

[0130] Specifically, screening out the surface peak points from the surface probability distribution, that is, p max = max{p ij}, taking p max as the surface peak point p peak , first constructing a plane rectangular coordinate system xy plane with the surface peak point p peak as the origin, projecting the peak point onto the xy plane to determine the projection point p' peak , and taking p' peak as the origin and the line where p peak p' peak is located as the z axis to establish a space rectangular coordinate system (X, Y, Z), where the z axis is perpendicular to the xy plane. Divide four quadrants in the xy plane, the first quadrant is X>0, Y>0, the second quadrant is X<0, Y>0, the third quadrant is X<0, Y<0, and the fourth quadrant is X>0, Y<0, that is, using the x-axis vertical plane and the y-axis vertical plane to cut the surface sample data, and dividing the surface probability distribution into initial surface regions corresponding to the four quadrants.

[0131] According to the central limit theorem, it is deduced that the distribution of the surface probability in multiple initial surface regions approximately follows a normal distribution, and the surface inflection points corresponding to each initial surface region are screened out from the surface probability distribution.

[0132] Divide the corresponding initial surface region into multiple surface regions according to any one of the surface inflection points, and in this process, process the initial surface region to obtain multiple surface regions corresponding to each initial surface region.

[0133] Construct an initial parabolic equation, and calculate the initial parabolic equation respectively through the surface probability distributions (i.e., the surface interval sequences and the corresponding surface distribution probabilities) of the multiple surface regions to obtain a surface probability density function.

[0134] It should be understood that the requirement for the probability density function of the surface in the three-dimensional space deduced from the two-dimensional sample space (i.e., the two-dimensional surface data in the three-dimensional space) to converge to the normal distribution according to the central limit theorem is that the random factors are independent and the result of taking the limit as the quantity tends to infinity. However, the actual random factors are not completely independent, the quantity of random factors cannot tend to infinity, and the symmetry of the probability density function can only exist in theoretically ideal data. Therefore, for the characteristics of the actual sample data distribution collected by the actual system, and its surface probability density function tends to the normal distribution but shows asymmetry. The normal distribution function to which the surface probability density function converges is:

[0135]

[0136] where f(x, y) is the normal distribution function to which the surface probability density function converges, x is the two-dimensional surface data in one dimension, y is the two-dimensional surface data in another dimension, is the standard deviation of x, is the standard deviation of y, is the correlation coefficient between x and y, u1 is the mean of x, and u2 is the mean of y.

[0137] Preferably, the screening of the surface inflection points corresponding to each initial surface region from the surface probability distribution includes:

[0138] Calculate multiple curve sample data in any one of the initial surface regions through a pre-constructed surface central difference formula to obtain multiple surface difference results, screen out the minimum surface difference from the multiple surface difference results, and use the surface interval sequence corresponding to the surface sample data corresponding to the minimum surface difference as the surface inflection point; in this process, process all the initial surface regions to obtain the surface inflection points corresponding to each initial surface region.

[0139] Specifically, estimate the inflection point data of the surface sample data in the initial surface regions corresponding to the four quadrants respectively through the surface central difference formula, and the surface central difference formula is:

[0140]

[0141] Among them, D is the surface difference result, Δ x is the first surface interval, Δ y is the second surface interval, P i, k is the surface distribution probability;

[0142] The minimum value corresponding points (x , 0), (x inflexion1 , 0), (0, y inflexion2 ), (0, y inflexion1 ) that satisfy the surface center difference formula inflexion2 are determined as inflection points.

[0143] In the embodiments of the present invention, the inflection points of the probability distribution change are found in the surface intervals of the surface sample data through the surface center difference formula, and the surface intervals of the data distribution are divided into multiple regions according to each inflection point, so as to improve the fitting accuracy for constructing the curve probability density function.

[0144] Preferably, for the construction of the initial paraboloid equation, the initial paraboloid equation is calculated respectively through the surface interval sequences of multiple said surface regions and the corresponding surface distribution probabilities to obtain the surface probability density function, including:

[0145] Multiple different elliptic equations are constructed according to the surface inflection points of each said surface region, and multiple said surface sample data are respectively substituted into each elliptic equation for calculation. According to the calculation results, the number of the surface sample data in each said surface region is counted to obtain the surface region sample numbers corresponding to each said surface region;

[0146] For any one of said surface regions, an initial paraboloid equation is constructed. The surface region sample number in the surface region and the surface distribution probability corresponding to the surface interval sequence where it is located are respectively substituted into the initial paraboloid equation for calculation to obtain a second design matrix, a second response vector, and a second coefficient matrix. The second design matrix, the second response vector, and the second coefficient matrix are calculated through the least squares method to obtain a surface coefficient matrix. The initial paraboloid equation is updated according to the surface coefficient matrix to obtain a paraboloid equation; in this process, all surface regions are processed to obtain multiple paraboloid equations, and the surface probability density function is composed of multiple paraboloid equations.

[0147] Specifically, as Figure 4 and Figure 5 shown, considering the inflection point asymmetry, the surface region is divided in the form of an elliptic equation and inflection points, four inflection points are found, and the probability distribution surface is divided into 8 regions. Let a1 = (x inflexion1 , 0), a2 = (x inflexion2 , 0), b1 = (0, y inflexion1), b2 = (0, y inflexion2 ).

[0148] If then the surface sample data falling within this region belongs to the first surface region D 11 ;

[0149] If then the surface sample data falling within this region belongs to the second surface region D 12 ;

[0150] If then the surface sample data falling within this region belongs to the third surface region D 21 ;

[0151] If then the surface sample data falling within this region belongs to the fourth surface region D 22 ;

[0152] If then the surface sample data falling within this region belongs to the fifth surface region D 31 ;

[0153] If then the surface sample data falling within this region belongs to the sixth surface region D 32 ;

[0154] If then the surface sample data falling within this region belongs to the seventh surface region D 41 ;

[0155] If then the surface sample data falling within this region belongs to the eighth surface region D 42 . In each surface region, construct a paraboloid equation:

[0156] z = ax 2 + by 2 + cxy + dx + ey + f,

[0157] The values of a, b, c, d, e, and f in the paraboloid equation for each surface region are different and need to be solved separately. Construct a design matrix and a response vector in each surface region.

[0158] Substitute the corresponding multiple surface sample data in each surface region into the paraboloid equation corresponding to this surface region to obtain the design matrix X and the response vector P. The design matrix X is expressed as:

[0159]

[0160] The response vector P is expressed as:

[0161]

[0162] The coefficient matrix of the parabola equation of each curve area is solved by the least square method to obtain the surface coefficient matrix. The parabola equation is updated according to the surface coefficient matrix to obtain the parabola equation. The least square method calculation expression is:

[0163] β4=(X T P) -1 X T P.

[0164] In the above least squares calculation expression, X T P is the product of the transpose of the design matrix and the response vector, (X T P) -1 For X T The inverse matrix of P.

[0165] The surface probability density function is composed of the parabolic equations of multiple surface areas. The expression of the surface probability density function is:

[0166]

[0167] In an embodiment of the present invention, during the calibration phase, an eight-segment parabola equation is used to fit the collected data for the surface sample data. This segmented fitting method not only significantly reduces the amount of calculation, but also effectively reduces the overall error by optimizing the fitting accuracy of each segment, thereby achieving higher modeling accuracy.

[0168] Preferably, the weighting of the plurality of curve target data is respectively performed using the curve probability density function to obtain a plurality of reconstructed curve target data, including:

[0169] The curve probability density function is used to calculate the multiple curve target data respectively to obtain the weighted average corresponding to each curve target data, each weighted average is normalized to obtain multiple curve data weights, and the multiple curve data weights are respectively assigned to the corresponding curve target data to obtain multiple reconstructed curve target data (i.e., curve target position estimation).

[0170] Specifically, if the target object is a curve in three-dimensional space, for the same position x k Measure a small number of N times to obtain the curve data (x k1 , x k2 , …, x kN ). These small amounts of collected data still follow the distribution of the calibration phase. ki Substitute the curve probability density function obtained in the calibration phase to obtain the probability value f(x ki ); if x kiFalling within regions D1, D2, D3, and D4, using the probability value as the weighted average, and assigning corresponding weights to the curve data according to the weighted average can more accurately reflect the true situation of the position distribution of one-dimensional curve data. If x ki Falls outside regions D1, D2, D3, and D4, due to low confidence, this data is discarded.

[0171] For the weight f(x ki ), use to perform a normalization operation to obtain a complete probability space as:

[0172]

[0173] The probability space represents the weight of each measured value x ki , where this weight reflects the importance or reliability of each measured value in the distribution.

[0174] Multiply each measured value x ki by its corresponding weight and sum them to obtain an optimal estimate of the weighted average position Calculated through the curve correction formula in three-dimensional space, and the expression is:

[0175]

[0176] In the embodiments of the present invention, the position error of the data is corrected according to the weight of the data through the curve correction formula, improving the accuracy of the data.

[0177] Preferably, weight distribution is respectively performed on multiple surface target data through the surface probability density function to obtain multiple reconstructed surface target data, including:

[0178] Calculating through the surface probability density function respectively for multiple surface target data to obtain the weighted average corresponding to each surface target data, performing normalization processing on each weighted average respectively to obtain multiple surface data weights, and respectively assigning the multiple surface data weights to the corresponding surface target data to obtain multiple reconstructed surface target data (i.e., surface target position estimation).

[0179] Specifically, if the target object is a surface in three-dimensional space, when collecting surface data in three-dimensional space, a small number of N measurements are actually made at the same position (x i , y j ) to obtain surface data {(x1, y1), (x2, y1), …, (x n , y1); (x1, y2), (x2, y2), …, (x n , y2); …; (x1, y m ), (x2, y m ), …, (xn , y m )}. These small amounts of collected data still follow the normal distribution in the calibration stage.

[0180] Substitute (x i , y j ) into the surface probability density function obtained in the calibration stage to obtain the probability value f(x ki , y kj ) at this point. If (x min , y min ) < (x i , y j ) < (x max , y max ), then the probability of this data is the weighted average, which is used as the weight of this data and can more accurately reflect the true situation of the data. If (x min , y min ) ≥ (x i , y j ) or (x max , y max ) ≥ (x ki , y kj ), then the confidence level is low and this data is discarded.

[0181] Perform a normalization operation on the weight f(x ki , y kj ) to obtain a complete probability space as:

[0182]

[0183] The probability space represents the weight of each measurement value (x i , y , y j ), where this weight reflects the importance or reliability of each measurement value in the distribution.

[0184] Multiply each measurement value (x i , y j ) by its corresponding weight and sum them to obtain the optimal estimated position of the weighted average Calculated by the surface correction formula in three-dimensional space, and the expression is:

[0185]

[0186] Calculate the two-dimensional target object data collected by using the correction formula, and the obtained data is the correction value. The set of multiple correction points together constitutes the geometric representation of the target object.

[0187] In the embodiment of the present invention, during the actual measurement stage of the target object, based on the probability density function constructed in the calibration stage, only a small amount of target data needs to be collected for the target object, and the position of the data point in the three-dimensional model can be quickly estimated through the probability density function, and the position of the data point can be further optimized through the correction formula to achieve reconstruction, ensuring the reliability and accuracy of the finally output target data.

[0188] As Figure 6 shown, a three-dimensional model reconstruction system provided by an embodiment of the present invention includes:

[0189] A three-dimensional model reconstruction system includes:

[0190] A curve calibration module, configured to import a plurality of collected one-dimensional curve sample data, perform one-dimensional interval division on the plurality of curve sample data to obtain a curve probability distribution, and construct a curve probability density function based on the curve probability distribution;

[0191] A surface calibration module, configured to import a plurality of collected two-dimensional surface sample data, perform two-dimensional interval division on the plurality of surface sample data to obtain a surface probability distribution, and construct a surface probability density function based on the surface probability distribution;

[0192] A data acquisition module, configured to perform data acquisition on the target object to obtain a plurality of target data;

[0193] A curve reconstruction module, configured to, if the plurality of target data are a plurality of one-dimensional curve target data, respectively perform weight assignment on the plurality of curve target data through the curve probability density function to obtain a plurality of reconstructed curve target data, and construct a three-dimensional model of the target object from the plurality of reconstructed curve target data;

[0194] A surface reconstruction module, configured to, if the plurality of target data are a plurality of two-dimensional surface target data, respectively perform weight assignment on the plurality of surface target data through the surface probability density function to obtain a plurality of reconstructed surface target data, and construct a three-dimensional model of the target object from the plurality of reconstructed surface target data.

[0195] The fields to which the present invention can be applied include: in the industrial field, it is used to detect industrial products with complex shapes, can effectively identify and locate potential defects, and make up for the deficiencies of three-dimensional imaging. In the military field, it can quickly and accurately map the topographic maps of combat areas, including three-dimensional information such as mountains, valleys, rivers, and buildings, providing detailed geographical data support for military operations. In the medical field, it can provide three-dimensional structural information of human organs and tissues, enabling doctors to observe the diseased parts from different angles, thereby improving the accuracy of diagnosis and the treatment effect.

[0196] For the above three-dimensional model reconstruction system, reference may be made to the implementation content and its beneficial effects described in detail above for a three-dimensional model reconstruction method, which will not be elaborated herein.

[0197] It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0198] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems and modules can refer to the corresponding processes in the foregoing method embodiments, which will not be elaborated herein.

[0199] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of modules is only a logical function division, and there may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0200] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the objectives of the embodiments of the present invention.

[0201] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A three-dimensional model reconstruction method, characterized in that, Including the following steps: Importing multiple one-dimensional curve sample data collected, performing one-dimensional interval division on the multiple curve sample data to obtain a curve probability distribution, and constructing a curve probability density function based on the curve probability distribution; Importing multiple two-dimensional surface sample data collected, performing two-dimensional interval division on the multiple surface sample data to obtain a surface probability distribution, and constructing a surface probability density function based on the surface probability distribution; Performing data collection on the target object to obtain multiple target data; If the multiple target data are multiple one-dimensional curve target data, respectively performing weight assignment on the multiple curve target data through the curve probability density function to obtain multiple reconstructed curve target data, and constructing a three-dimensional model of the target object from the multiple reconstructed curve target data; If the multiple target data are multiple two-dimensional surface target data, respectively performing weight assignment on the multiple surface target data through the surface probability density function to obtain multiple reconstructed surface target data, and constructing a three-dimensional model of the target object from the multiple reconstructed surface target data.

2. The three-dimensional model reconstruction method according to claim 1, wherein The performing one-dimensional interval division on the multiple curve sample data to obtain a curve probability distribution includes: Screening out the maximum curve value and the minimum curve value from the multiple one-dimensional curve sample data, calculating the difference between the maximum curve value and the minimum curve value to obtain the total curve interval length; Calculating the ratio of the total curve interval length to the set number of curve intervals to obtain the curve interval interval, and dividing the total curve interval length into multiple curve intervals according to the curve interval interval, where each curve interval includes multiple corresponding curve sample data; Taking the midpoint value of any one of the curve intervals as the curve interval sequence, counting the number of curve sample data in the curve interval to obtain the curve distribution number, calculating the ratio of the curve distribution number to the number of the multiple curve sample data to obtain the curve distribution probability; through this process, calculating for each of the curve intervals to obtain the curve distribution probability corresponding to each of the curve intervals, and the curve interval sequence and the corresponding curve distribution probability of each curve interval form a curve probability distribution.

3. The three-dimensional model reconstruction method according to claim 2, characterized in that The constructing a curve probability density function based on the curve probability distribution includes: Screening out the starting point, the first curve inflection point, the curve peak point, the second curve inflection point, and the ending point from the curve probability distribution; Respectively taking the starting point, the first curve inflection point, the curve peak point, the second curve inflection point, and the ending point as regional nodes, and dividing the curve probability distribution into multiple curve regions; Constructing an initial parabola equation, and respectively calculating the initial parabola equation through the curve interval sequence and the corresponding curve distribution probability of the multiple curve regions to obtain a curve probability density function.

4. The three-dimensional model reconstruction method according to claim 3, wherein, The screening out the starting point, the first curve inflection point, the curve peak point, the second curve inflection point, and the ending point from the curve probability distribution includes: Select the maximum curve distribution probability from the curve probability distribution, take the curve interval sequence corresponding to the maximum curve distribution probability as the curve peak point, take the first curve interval sequence as the starting point, and take the last curve interval sequence as the ending point; Calculate multiple pieces of the curve sample data within the range from the starting point to the curve peak point through a pre-constructed curve central difference formula to obtain multiple first curve difference results, screen the minimum value of the multiple first curve difference results to obtain the first difference minimum value, and take the curve interval sequence corresponding to the first difference minimum value as the first curve inflection point; Calculate multiple pieces of the curve sample data within the range from the curve peak point to the ending point through a pre-constructed curve central difference formula to obtain multiple second curve difference results, screen the minimum value of the multiple second curve difference results to obtain the second difference minimum value, and take the curve interval sequence corresponding to the second curve difference minimum value as the second curve inflection point.

5. The three-dimensional model reconstruction method according to claim 3, characterized in that Construct the initial parabola equation, and calculate the curve probability density function through the curve interval sequence and the corresponding curve distribution probability of multiple curve regions respectively, including: Count the number of the curve sample data in each curve region respectively to obtain the region sample number corresponding to each curve region; For any one of the curve regions, construct the initial parabola equation, substitute the region sample number in the curve region and the curve distribution probability corresponding to the curve interval sequence into the initial parabola equation for calculation respectively to obtain the first design matrix, the first response vector and the first coefficient matrix, calculate the first design matrix, the first response vector and the first coefficient matrix through the least squares method to obtain the curve coefficient matrix, and update the initial parabola equation according to the curve coefficient matrix to obtain the parabola equation; in this process, process all curve regions to obtain multiple parabola equations, and the curve probability density function is composed of multiple parabola equations.

6. The three-dimensional model reconstruction method according to claim 1, characterized in that Perform two-dimensional interval division on multiple pieces of the surface sample data to obtain the surface probability distribution, including: Select the first maximum surface value, the first minimum surface value, the second maximum surface value and the second minimum surface value from multiple two-dimensional surface sample data, calculate the difference between the first maximum surface value and the first minimum surface value to obtain the first total surface interval length, and calculate the difference between the second maximum surface value and the second minimum surface value to obtain the second total surface interval length; Calculate the ratio of the total interval length of the first surface to the set number of intervals of the first surface to obtain the interval spacing of the first surface. Calculate the ratio of the total interval length of the second surface to the set number of intervals of the second surface to obtain the interval spacing of the second surface. Divide the total interval length of the first surface into multiple first surface intervals according to the interval spacing of the first surface. Divide the total interval length of the second surface into multiple second surface intervals according to the interval spacing of the second surface. Each first surface interval and each second surface interval include a plurality of corresponding surface sample data; Take the midpoint value of any one of the first surface intervals as the first surface interval sequence, take the midpoint value of any one of the second surface intervals as the second surface interval sequence, count the number of surface sample data in the first surface interval and the second surface interval to obtain the surface distribution quantity, and calculate the ratio of the surface distribution quantity to the number of a plurality of the surface sample data to obtain the surface distribution probability; In this process, calculate for each of the first surface intervals and each of the second surface intervals to obtain the surface distribution probability corresponding to each of the surface intervals, and the surface interval sequence and the corresponding surface distribution probability of each of the surface intervals form a surface probability distribution.

7. The three-dimensional model reconstruction method according to claim 6, characterized in that, The constructing a surface probability density function based on the surface probability distribution includes: Select surface peak points from the surface probability distribution, establish a spatial rectangular coordinate system based on the surface peak points, divide the surface probability distribution into multiple initial surface regions according to the coordinate axes of the spatial rectangular coordinate system, and select surface inflection points corresponding to each initial surface region from the surface probability distribution; Divide the corresponding initial surface region into multiple surface regions according to any one of the surface inflection points. In this process, process the initial surface region to obtain multiple surface regions corresponding to each initial surface region; Construct an initial parabolic equation, and calculate the initial parabolic equation respectively through the surface interval sequences and the corresponding surface distribution probabilities of multiple surface regions to obtain a surface probability density function.

8. The three-dimensional model reconstruction method according to claim 7, wherein The selecting surface inflection points corresponding to each initial surface region from the surface probability distribution includes: Calculate multiple surface difference results by using a pre-constructed surface central difference formula for multiple curve sample data in any one of the initial surface regions, select the minimum surface difference result from the multiple surface difference results, and use the surface interval sequence corresponding to the minimum surface difference result as the surface inflection point; In this process, process all the initial surface regions to obtain surface inflection points corresponding to each initial surface region.

9. The three-dimensional model reconstruction method according to claim 7, wherein The constructing an initial parabolic equation, and calculating the initial parabolic equation respectively through the surface interval sequences and the corresponding surface distribution probabilities of multiple surface regions to obtain a surface probability density function includes: Construct a plurality of different elliptic equations according to the inflection points of the surface of each of the surface regions, substitute the plurality of surface sample data into each elliptic equation for calculation respectively, and count the number of the surface sample data in each of the surface regions according to the calculation results to obtain the surface region sample numbers corresponding to the respective surface regions; For any one of the surface regions, construct an initial parabolic equation, substitute the surface region sample number in the surface region and the surface distribution probability corresponding to the surface interval sequence where it is located into the initial parabolic equation for calculation respectively to obtain a second design matrix, a second response vector and a second coefficient matrix, calculate the second design matrix, the second response vector and the second coefficient matrix by the least square method to obtain a surface coefficient matrix, and update the initial parabolic equation according to the surface coefficient matrix to obtain a parabolic equation; In this process, process all the surface regions to obtain a plurality of parabolic equations, and form a surface probability density function from the plurality of parabolic equations.

10. A three-dimensional model reconstruction system, characterized in that, Including: A curve calibration module, configured to import a plurality of collected one-dimensional curve sample data, perform one-dimensional interval division on the plurality of curve sample data to obtain a curve probability distribution, and construct a curve probability density function based on the curve probability distribution; A surface calibration module, configured to import a plurality of collected two-dimensional surface sample data, perform two-dimensional interval division on the plurality of surface sample data to obtain a surface probability distribution, and construct a surface probability density function based on the surface probability distribution; A data acquisition module, configured to perform data acquisition on a target object to obtain a plurality of target data; A curve reconstruction module, configured to, if the plurality of target data are a plurality of one-dimensional curve target data, perform weight assignment on the plurality of curve target data respectively through the curve probability density function to obtain a plurality of reconstructed curve target data, and construct a three-dimensional model of the target object from the plurality of reconstructed curve target data; A surface reconstruction module, configured to, if the plurality of target data are a plurality of two-dimensional surface target data, perform weight assignment on the plurality of surface target data respectively through the surface probability density function to obtain a plurality of reconstructed surface target data, and construct a three-dimensional model of the target object from the plurality of reconstructed surface target data.

Images