Parameter adaptive density filtering method and device for silo grain pile volume point cloud data

Abstract

The invention relates to a parameter adaptive density filtering method and device for silo grain pile volume point cloud data. The parameter adaptive density filtering method for the silo grain pile volume point cloud data comprises the following steps: S1, obtaining the measured point cloud data of the grain pile volume in a silo; s2, generating random samples from the point cloud data through a Monte Carlo method, calculating according to the random samples to obtain an optimal total cluster number Kbest, and calculating according to the Kbest to obtain an adaptive average Euclidean distance mean value # imgabs0 # S3, calculating according to # imgabs1 # to obtain a global density mean value # imgabs2 # of the point cloud data and local density rho 1-rho N of N original points; and calculating according to # imgabs3 # and rho1-rhoN to obtain a self-adaptive density threshold rhoy. According to the method, adaptive calculation of neighborhood radius parameters is realized by adopting a clustering thought, and then the neighborhood density of each original point is calculated, and the adaptive density threshold is dynamically adjusted by utilizing statistical characteristics, so that low-density noise points are eliminated, and therefore, the accuracy of point cloud data filtering is ensured, and the accuracy of subsequent granary reserve calculation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud data processing, and particularly to a method and device for parameter adaptive density filtering of silo grain pile volume point cloud data. Background Art

[0002] With the continuous development of automated management of grain silos, how to efficiently and accurately extract the characteristic parameters of the grain pile volume in the grain silo has become an important issue in the intelligent grain silo management system. Traditional grain pile structure detection and modeling mainly rely on traditional manual measurement and pressure sensor methods. However, the manual measurement method is not only time-consuming and laborious, but also its accuracy cannot be guaranteed when facing the complex measurement environment inside the grain silo. The pressure sensor measurement method has a high cost and high failure rate due to the large number of sensors required, and its usability is increasingly reduced especially when the size and complexity of the grain silo increase.

[0003] In recent years, with the increasing maturity of three-dimensional laser technology, three-dimensional modeling methods based on point cloud data have gradually emerged. These technologies can obtain the three-dimensional point cloud data of the upper surface of the grain pile in the grain silo with high precision, and then the grain storage can be calculated based on the three-dimensional point cloud data, providing a new way for the structural analysis and feature extraction of the stored grain inside the grain silo to improve the accuracy of subsequent grain storage calculation. However, point cloud data usually has problems such as noise, redundancy, and data missing, resulting in difficult feature extraction, especially in a large-scale and complex environment such as a grain silo.

[0004] Currently, the commonly used point cloud filtering methods mainly include statistical filtering, radius filtering, voxel filtering, etc. Among them, statistical filtering filters based on the statistical features of the neighborhood of points, but the parameter setting lacks self - adaptability and is prone to loss of effective data; radius filtering filters by setting the point number threshold within a fixed radius range, but it is difficult to adapt to the situation of uneven point cloud density; voxel filtering divides the point cloud data into regular grids for downsampling, but it may lose important detail information. These methods have improved the quality of point cloud data to a certain extent, but there are still the following deficiencies: First, when the point cloud density is too large, traditional filtering methods may lose some important structural details; second, the existing point cloud filtering methods have poor adaptability to complex structures in different environments, especially in a scenario like the interior of a granary with various complex factors, and often cannot effectively distinguish noise from real structural features; finally, existing methods often cannot consider the noise source distribution in a specific scenario when processing point clouds, resulting in a large deviation in the processing results and making it difficult to achieve accurate feature extraction, thus affecting the accuracy of subsequent grain storage calculation. In addition, the existing filtering algorithms generally have the problems of lack of self - adaptability in parameter setting and the need for repeated manual debugging. At the same time, these algorithms are difficult to ensure a high filtering rate for the special noise of the point cloud data of the grain pile volume inside the granary, and insufficient consideration is given to the particularity of the point cloud data of the grain pile volume in the granary. Summary of the Invention

[0005] Based on this, in view of the problem of low filtering accuracy of the existing filtering methods for the feature point cloud of grain pile volume, it is necessary to provide a parameter - adaptive density filtering method and device for the point cloud data of the silo grain pile volume.

[0006] In the first aspect, the present invention proposes a parameter - adaptive density filtering method for the point cloud data of the silo grain pile volume, which includes the following steps:

[0007] S1. Obtain the point cloud data measured for the grain pile volume in the silo; the point cloud data includes: N original points.

[0008] S2. Generate random samples from the point cloud data through the Monte Carlo method, and calculate the optimal total number of clusters K according to the random samples best , and then calculate the adaptive average Euclidean distance mean according to K best

[0009] S3. First, calculate the global density mean of the point cloud data according to and the local density ρ of the N original points 1 ~ρ N ; n ∈ [1, N].

[0010] Then, according to and ρ 1 ~ρN Calculate the adaptive density threshold ρ y .

[0011] S4. Determine whether ρ 1 ~ρ N is greater than ρ y .

[0012] If so, retain the original points corresponding to the point cloud data as information points; if not, filter out the points corresponding to the point cloud data as noise points.

[0013] Among them, S2 includes the following steps:

[0014] S21. Perform M clustering processes on the random samples.

[0015] The m-th clustering process generates K + m clusters; m ∈ [1, M].[[]]

[0016] S22. Calculate the tight support measure D of the (k + m)-th cluster in the m-th clustering process m,k+m ; k ∈ [1, K + m];

[0017] Traverse the K + m clusters in the m-th clustering process to obtain K + m tight support measures D m,2 ~D m,K+m ;

[0018] According to D m,2 ~D m,K+m Calculate the sum of squared errors W in the m-th clustering process m,K+m .

[0019] S23. Perform M clustering processes to obtain M sums of squared errors W 1,K+1 ~W M,K+M , and calculate the average value of their logarithms to obtain E(logW m,K+m ).

[0020] S24. Construct the interval value function Gap(K + m), and take the K + m corresponding to its maximum value as the optimal total number of clusters K best .

[0021]

[0022] S25. Divide N by K best Calculate the number of neighborhood points t.

[0023] Calculate the average Euclidean distance from the n-th original point to the t adjacent original points

[0024] Traverse the N original points to obtain N average Euclidean distances

[0025] Pair Take the average to obtain the mean value

[0026] In a second aspect, the present invention also proposes a parameter adaptive density filtering device for silo grain pile volume point cloud data, which uses the parameter adaptive density filtering method for silo grain pile volume point cloud data in the first aspect. The parameter adaptive density filtering device for silo grain pile volume point cloud data includes an acquisition module, an adaptive module, a threshold module, and a judgment module.

[0027] The acquisition module is used to acquire the point cloud data measured for the grain pile volume in the silo.

[0028] The adaptive module is used to generate random samples from the point cloud data by the Monte Carlo method, and calculate the optimal total number of clusters K according to the random samples best , and according to K best Calculate the adaptive mean value of the average Euclidean distance

[0029] The threshold module is used to calculate the global density mean value of the point cloud data according to Calculate the global density mean value of the point cloud data And the local density ρ of N original points 1 ~ρ N , and according to And ρ 1 ~ρ N Calculate the adaptive density threshold ρ y .

[0030] The judgment module is used to judge whether ρ 1 ~ρ N Is greater than ρ y ; if so, retain the original point corresponding to the point cloud data as an information point; if not, filter out the point corresponding to the point cloud data as a noise point.

[0031] In a third aspect, the present invention also proposes a software program product. The software program product includes program instructions, which when running on an electronic device, cause the electronic device to execute the steps of the parameter adaptive density filtering method for silo grain pile volume point cloud data in the first aspect.

[0032] The beneficial effects of the present invention are as follows:

[0033] 1. The present invention uses the clustering idea to realize the adaptive calculation of the neighborhood radius parameter, then calculates the neighborhood density of each original point, and uses statistical characteristics to dynamically adjust the adaptive density threshold to eliminate low-density noise points, so as to ensure the accuracy of point cloud data filtering and improve the accuracy of subsequent grain storage calculation.

[0034] 2. The present invention obtains an adaptive density threshold from the statistical characteristics of the point cloud data itself. This dynamic adjustment of the adaptive density threshold setting takes into account the overall density distribution law of the point cloud data. Compared with the traditional fixed threshold method, this way of determining the adaptive threshold is more flexible and can better adapt to the characteristics of point cloud data in different scenarios.

[0035] 3. The invention can filter lidar point cloud data for complex structures and large-scale environments, and the parameter adaptability in the filtering method makes it applicable to most actual cylindrical granaries, eliminating the need for manual adjustment of filtering parameters, reducing labor costs while enhancing the algorithm's adaptability to different scenarios, and improving the efficiency and intelligent level of granary point cloud data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 It is a flowchart of the parameter adaptive density filtering method for the point cloud data of the grain pile volume in the silo in the embodiment;

[0038] Figure 2 It is the effect diagram of the original grain pile volume point cloud data in the experiment;

[0039] Figure 3 It is the effect diagram of the point cloud data obtained after denoising using the traditional statistical filtering algorithm in the experiment;

[0040] Figure 4 It is the effect diagram of the point cloud data obtained after denoising using the point cloud denoising algorithm based on statistical filtering and bilateral filtering in the experiment;

[0041] Figure 5 It is the effect diagram of the point cloud data obtained after denoising using the parameter adaptive statistical filtering denoising algorithm in the experiment;

[0042] Figure 6 It is the effect diagram of the point cloud data obtained after denoising using the present invention in the experiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying 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. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0045] Please refer to Figure 1 , this embodiment provides a parameter adaptive density filtering method for point cloud data of the volume of a silo grain pile, which includes the following steps:

[0046] S1. Obtain the point cloud data measured for the volume of the grain pile in the silo; the point cloud data includes: N original points.

[0047] In this step, the environment of the grain bin can be scanned by a laser sensor to automatically generate a point cloud data set about the surface of the grain pile. The number information of the original points is included in this data set.

[0048] S2. Generate random samples from the point cloud data by the Monte Carlo method, and calculate the optimal total number of clusters K best , and calculate the adaptive average Euclidean distance mean E(d) according to K best .

[0049] In this step, this embodiment adopts the clustering idea to divide the point cloud data into different clusters. Among them, the point cloud representing the main information and the noise points with less information generally belong to different cluster classes. In this embodiment, the interval statistic method with less computational complexity and insensitive to the density and sparsity of the data set is adopted. Specifically, S2 includes the following steps:

[0050] S21. The random samples generated by the Monte Carlo method are uniformly distributed in the region where they are located, and randomly generate the same number of random points as the original points in the point cloud data. These random points are subjected to M times of clustering processing. The m-th clustering processing will generate K + m clusters. Where m ∈ [1, M].

[0051] S22. Calculate the tight support measure D m,k+m of the (k + m)-th cluster in the m-th clustering processing. Where k ∈ [1, K + m].

[0052] Dm,k+m Represented by the Euclidean distance, the smaller it is, the better the compactness of the clustering.

[0053] Specifically, D m,k+m The calculation formula of is:

[0054]

[0055] In the formula, Q k+m is all the original points in the (k + m)-th cluster; P q is the q-th original point in Q k+m ; μ is the center point of the (k + m)-th cluster.

[0056] Subsequently, traverse the K + m clusters in the m-th clustering process to obtain the compactness measure D m,2 ~D m,K+m , and then calculate the sum of squared errors W m,2 ~D m,K+m in the m-th clustering process. Its calculation formula is: m,K+m The calculation formula of is:

[0057]

[0058] S23. Take the logarithm log(W m,K+m ). During the M-th clustering process, M logarithms log(W m,K+m )~log(W 1,K+1 ) are obtained. Take the average value of log(W M,K+M )~log(W 1,K+1 ) to obtain E(logW M,K+M ). m,K+m )

[0059] S24. Construct the gap value function Gap(K + m), and take the K + m corresponding to its maximum value as the optimal total number of clusters K best :

[0060]

[0061] S25. Cluster the point cloud data into K best clusters according to K best . At this time, divide N by K best to calculate the number of neighborhood points t. This t is an average number of the original points in each cluster. Take t as the number of neighborhood points required by the KNN algorithm, and thus the average Euclidean distance d n from the n-th original point to t adjacent original points can be calculated. Subsequently, traverse the N original points to obtain N average Euclidean distances Sum and average to obtain the required mean value

[0062] S3. According to calculate the global density mean value of the point cloud data and the local density ρ of N original points 1 ~ρ N . n ∈ [1, N]. According to and ρ 1 ~ρ N calculate the adaptive density threshold ρ y .

[0063] In this step, establish the filtering parameters with adaptive density, which mainly consider the local density characteristics of each original point in the point cloud data. First, take as the search radius of the nth original point in the point cloud data. This search radius forms a spherical space, and calculate the number M of all original points located within the spherical space. The calculation method of M includes the following steps:

[0064] S61. Calculate the Euclidean distance d j between the nth original point and any original point P nj .

[0065] S62. Compare the size of d nj with .

[0066] If then increment the count value by 1. If then increment the count value by 0.

[0067] S63. After traversing N - 1 original points, the obtained count value is M.

[0068] In another embodiment, further optimize the calculation method of M. The difference from the above embodiment is that: traverse the N - 1 original points in the order from near to far from the nth original point until and then stop. Compared with the calculation of the above embodiment, this embodiment only needs to calculate the original points near the nth original point, and does not need to retrieve all the original points, thus greatly reducing the data calculation amount.

[0069] After obtaining M, calculate the local density ρ of the nth original point according to n , and its calculation formula is:

[0070]

[0071] Subsequently, traverse N original points to obtain the local densities ρ 1 ~ρ NThis local density calculation fully considers the spatial distribution characteristics around each original point and can accurately reflect the density of the point cloud in different regions. Then, it is judged whether ρ 1 ~ρ N satisfies the Gaussian distribution condition. If it does, the truncation threshold multiple λ is obtained according to the properties of the Gaussian distribution (such as querying the normal distribution table). In order to obtain a suitable ρ y , it is also necessary to sum and average ρ 1 ~ρ N to obtain Obtain and λ, and then ρ y can be calculated. Its calculation formula is:

[0072]

[0073] In the formula, σ is the standard deviation of the point cloud data.

[0074] Among them, the Jarque-Bera statistical method can be used to evaluate whether ρ 1 ~ρ N obeys the Gaussian distribution. Specifically, first calculate the density variance S of the point cloud data. Its calculation formula is:

[0075]

[0076] Then the above σ is Subsequently, calculate the statistic JB. Its calculation formula is:

[0077]

[0078] In the formula, T is the skewness, L is the kurtosis, and E[·] represents the expected value of [·].

[0079] If N > 30 and JB~χ 2 (2), then ρ 1 ~ρ N satisfies the Gaussian distribution condition.

[0080] S4. For those original points with local density lower than the adaptive density threshold, since the distribution of the surrounding original points is sparse and there are obvious differences from the distribution characteristics of normal original points, they are identified as noise points or outliers and need to be removed from the data. And those original points with higher local density often reflect the true characteristics of the object surface and will be retained as valid point cloud data. Therefore, it is necessary to judge whether ρ n is greater than ρ y . If so, the original points corresponding to the point cloud data are retained as information points; if not, the points corresponding to the point cloud data are filtered out as noise points. Finally, the point cloud data composed of the retained information points is the filtered point cloud data.

[0081] In summary, the main principle of the present invention is to determine the number of neighborhoods according to the interval statistic, and obtain the neighborhood search radius and the truncation threshold multiple from the statistics of the point cloud data itself. Then, based on the neighborhood density of each original point in the point cloud data, the local density of each original point is evaluated by calculating the average neighborhood density. Then, an adaptive density threshold is calculated according to the statistical characteristics. All points below the adaptive density threshold are considered as noise points and removed, and the remaining points are retained as valid information points. This adaptive filtering method based on clustering ideas and density characteristics not only ensures the denoising effect of the algorithm, prevents the uncertainty and low efficiency caused by manually input parameters, but also avoids the loss of valid information caused by over-filtering. At the same time, due to the full utilization of the statistical characteristics of the point cloud data, this method has strong robustness and adaptability, can effectively process point cloud data with different density distribution characteristics, and provides a more reliable basis for subsequent data analysis and applications. The present invention is not only applicable to silos, but also to flat warehouses, and can be widely used in fields such as three-dimensional modeling and intelligent management inside grain warehouses. Especially in automated warehousing management and intelligent monitoring systems, it can provide important support for the storage calculation and intelligent management of grain warehouses. With the continuous development of intelligent technologies, based on the technology of the present invention, the grain warehouse management system will be more intelligent and efficient, promoting the grain warehouse industry to develop in a more automated and informatized direction.

[0082] To verify the effectiveness of the denoising effect of the present invention, the present invention is compared with existing different algorithms for the denoising effect of a simulated cylindrical silo grain surface model type with added composite noise, and then denoising experiments on the point cloud model under different noise intensities are carried out and the results are analyzed. The specific analysis process is as follows:

[0083] The experiments of the present invention were carried out in the Ubuntu 24.04 LTS operating system environment, and the point cloud library point Cloudlibrary (PCL 1.14.0) was used to implement the algorithm. For the visualization part, the built-in visualization module PCLVisualizer of PCL was used for point cloud rendering and display. As Figure 2 shown, the original data used in this paper is three-dimensional point cloud data modeled based on the relevant mathematical parameters of an actual cylindrical silo and the actual situation in the reference grain warehouse using the PCL point cloud library. It consists of the point cloud data on the surface of a cone with a bottom radius of 8 m and a height of 4 m and several simulated temperature measurement cable noises. The total number of points in the point cloud after adding the initial simulated noise is 105,000, as Figure 2 shown. Four groups of test schemes were set to perform noise reduction processing on the original point cloud data. To facilitate observing the denoising effect, the point cloud data obtained by traditional statistical filtering denoising is as Figure 3 shown. The point cloud data obtained by denoising using the point cloud denoising algorithm based on statistical filtering and bilateral filtering (Algorithm 1) is as Figure 4As shown. The point cloud data obtained by denoising using the parameter adaptive statistical filtering and denoising algorithm (Algorithm 2) is as follows Figure 5 As shown. The point cloud data obtained by denoising using the algorithm of the present invention is as follows Figure 6 As shown. It can be seen from the data in Table 1 that the denoising effect achieved by the method used in the present invention is better than that of other traditional algorithms and the algorithms mentioned in the literature.

[0084] Table 1: Comparison of denoising results

[0085]

[0086] In some other embodiments, a parameter adaptive density filtering device for the point cloud data of the silo grain pile volume is further proposed, which uses the parameter adaptive density filtering method for the point cloud data of the silo grain pile volume in the above embodiments.

[0087] The parameter adaptive density filtering device includes an acquisition module, an adaptive module, a threshold module and a judgment module. Among them:

[0088] The acquisition module is configured to: acquire the point cloud data measured for the grain pile volume in the silo. The adaptive module is configured to: generate random samples from the point cloud data by the Monte Carlo method, and calculate the optimal total number of clusters K best , and calculate the adaptive average Euclidean distance mean according to K best The threshold module is configured to: calculate the global density mean of the point cloud data according to and the local density ρ of N original points ~ρ , and calculate the adaptive density threshold ρ according to 1 ~ρ N and ρ ~ρ 1 ~ρ N . The judgment module is configured to: judge whether ρ y ~ρ 1 ~ρ N is greater than ρ y ; if so, retain the original point corresponding to the point cloud data as an information point; if not, filter out the point corresponding to the point cloud data as a noise point.

[0089] In some other embodiments, an electronic device is further proposed. The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the parameter adaptive density filtering method for the point cloud data of the silo grain pile volume in the above embodiments are implemented.

[0090] In some other embodiments, a computer-readable storage medium is also proposed. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the parameter adaptive density filtering method for the silo grain pile volume point cloud data in the above embodiments are implemented.

[0091] In some other embodiments, a software program product is also proposed. The software program product includes program instructions that, when running on an electronic device, cause the electronic device to execute the steps of the parameter adaptive density filtering method for the silo grain pile volume point cloud data in the above embodiments.

[0092] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0093] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent should be subject to the appended claims.

Claims

1. A parameter adaptive density filtering method for silo grain volume point cloud data, characterized in that: It includes the following steps: S1. Obtaining point cloud data of the grain pile volume in the silo; the point cloud data includes: N original points; S2. Generate random samples from the point cloud data through the Monte Carlo method, and calculate the optimal total number of clusters K based on the random samples best , and then according to K best Calculate the adaptive average Euclidean distance mean S3, first according to Calculate the global density mean of the point cloud data and the local density of N original points; n∈[1,N]; According to And calculate the adaptive density threshold ρ y ; S4. Determine ρ1~ρ N Is it greater than ρ y ; If yes, the original point corresponding to the point cloud data is retained as an information point; if no, the point corresponding to the point cloud data is filtered out as a noise point; Wherein, S2 includes the following steps: S21, performing M-times clustering processing on the random samples; The mth clustering process generates K+m clusters; m∈[1,M]; S22. Calculate the compact support measure D of the k+mth cluster in the mth clustering process m,k+m ; k∈[1,K+m]; Traverse the K+m clusters in the mth clustering process and obtain K+m compact support measures D m,2 ~D m,K+m ; According to D m,2 ~D m,K+m Calculate the error sum of squares W in the mth clustering process m,K+m ; S23, perform M clustering processes to obtain M error square sums W 1,K+1 ~W M,K+M , and calculate the average of its logarithms to obtain E(logW m,K+m ); S24. Construct the interval value function Gap(K+m), and take the K+m corresponding to its maximum value as the optimal total number of clusters K best ; S25. Divide N by K best Calculate the number of neighborhood points t; Calculate the average Euclidean distance from the nth original point to the t adjacent original points Traverse N original points and get N average Euclidean distances right Find the average and get the mean 2. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 1 is characterized in that: In S3, ρ1~ρ N The calculation method includes the following steps: First As the search radius of the nth original point in the point cloud data, and calculate the number M of all original points within the search radius; According to and M calculate the local density ρ of the nth original point n ; Traverse N original points and obtain N local densities ρ1~ρ N ; For ρ1~ρ N Taking the average, we get 3. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 2 is characterized in that: The calculation method of M includes the following steps: S61, calculate the nth original point and any original point P j The Euclidean distance d nj ; S62, Compare d nj and size; like Then the count value increases by 1; S63. Traverse N-1 original points and obtain a count value of M.

4. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 3 is characterized in that: The S63 also includes: Traverse N-1 original points in order from nearest to farthest from the nth original point until Then stop.

5. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 2 is characterized in that: A spherical space is formed according to the search radius of the nth original point; ρ n The calculation formula is:

6. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 1, characterized in that: In S3, ρ y The calculation method includes the following steps: S71, determine ρ1~ρ N Whether the Gaussian distribution condition is met; If it is satisfied, the cutoff threshold multiple λ is obtained according to the properties of Gaussian distribution; S72, calculate the adaptive density threshold ρ according to λ y , and its calculation formula is: Where σ is the standard deviation of the point cloud data.

7. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 6 is characterized in that: In S71, it is determined that ρ1~ρ N The method of whether the Gaussian distribution condition is satisfied includes the following steps: Calculate ρ1~ρ N The density variance S; The Jarque-Bera statistical method is used to calculate the statistic JB, and its calculation formula is: In the formula, T is skewness and L is kurtosis; If N>30, and JB~χ 2 (2), then ρ1~ρ N Satisfies Gaussian distribution conditions.

8. The parameter adaptive density filtering method for silo grain pile volume point cloud data according to claim 1, characterized in that: In S22, D m,k+m The calculation formula is: In the formula, Q k+m are all the original points in the k+mth cluster; P q Q k+m The qth original point in ; μ is the center point of the k+mth cluster.

9. A parameter adaptive density filtering device for silo grain volume point cloud data, characterized in that: It uses the parameter adaptive density filtering method for the volume point cloud data of the silo grain pile as described in any one of claims 1 to 8; The parameter adaptive density filtering device for the volume point cloud data of the silo grain pile includes: An acquisition module is used to acquire point cloud data of the grain pile volume in the silo; The adaptive module is used to generate random samples from point cloud data through the Monte Carlo method, and calculate the optimal total number of clusters K based on the random samples. best , and according to K best Calculate the adaptive average Euclidean distance mean The threshold module is used to Calculate the global density mean of the point cloud data and the local density of N original points ρ1~ρ N , and according to and ρ1~ρ N Calculate the adaptive density threshold ρ y ; The judgment module is used to judge ρ1~ρ N Is it greater than ρ y ; If so, the original point corresponding to the point cloud data is retained as an information point; if not, the point corresponding to the point cloud data is filtered out as a noise point.

10. A software program product, characterized in that The software program product includes program instructions, which, when running on an electronic device, enable the electronic device to execute the steps of the parameter adaptive density filtering method for volume point cloud data of a silo grain pile as described in any one of claims 1 to 8.

Images