A method and device for enhancing point cloud data of irregular stockpile based on diffusion model

Abstract

The present invention relates to point cloud data enhancement technology, and discloses a method for enhancing point cloud data of irregular stockpiles based on a diffusion model, including: collecting original point cloud data of the stockpiles and constructing a point cloud diffusion model; adding noise to the original point cloud data of the stockpiles and training the point cloud diffusion model to obtain a target point cloud diffusion model; extracting multi-level geometric features of the original point cloud data of the stockpiles using a multi-scale neural network pre-trained with the target point cloud diffusion model, and generating point cloud data using the multi-level geometric features to obtain generated point cloud data; integrating the generated point cloud data with the original point cloud data of the stockpiles to obtain enhanced point cloud data of the stockpiles. The present invention also proposes a device, electronic device, and storage medium for enhancing point cloud data of irregular stockpiles based on a diffusion model. The present invention can improve the data enhancement effect of stockpile data in terms of scale and diversity in a mining environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data enhancement, and in particular relates to a method and a device for enhancing point cloud data of irregular stockpiles based on a diffusion model. Background Art

[0002] With the development of computer vision and the continuous improvement of industrial automation, point cloud data is increasingly used in unmanned operation scenarios. In unmanned operation scenarios in mining areas, it is usually necessary to quickly and accurately identify bulk piles, loading equipment, pedestrians in mining areas, etc. However, due to the complex environment of mining areas, the collected point cloud data often has problems such as noise, sparsity and incompleteness, and it is often difficult to obtain sufficiently comprehensive and high-quality point cloud data. These problems not only affect the accuracy and reliability of the data, but also increase the complexity and cost of data processing, making the actual data set appear to be stretched in terms of data volume and unable to effectively meet the needs of model training. In practical applications, the scale and diversity of the data set directly affect the performance and generalization ability of the recognition perception model. If the data set is insufficient, the model may not be able to fully learn the various scenes and object features that may appear in the mining area, thus affecting its performance in practical applications. For example, there are a large number of changing terrains, different types of equipment, and bulk piles of various sizes in the mining environment. These features may not be fully reflected in small-scale data sets, resulting in reduced recognition accuracy and stability of the model. Therefore, the data enhancement task in mining scenarios is particularly important, and the point cloud data enhancement of typical irregular material piles in mining scenarios is a more difficult problem to solve in the entire task. Among them, irregular material piles refer to material piles in mining areas with irregular geometric features and large volume differences.

[0003] Current point cloud data enhancement methods mainly include traditional filtering methods and interpolation algorithms. However, traditional filtering and interpolation methods cannot generate new data samples, but can only optimize existing data, making it impossible for traditional methods to generate more diverse samples based on existing data, resulting in poor data enhancement effects in terms of scale and diversity of stockpile data in mining environments. Summary of the invention

[0004] The present invention provides a method and device for enhancing point cloud data of irregular stockpiles based on a diffusion model, which can improve the data enhancement effect of stockpile data in terms of scale and diversity in a mining environment.

[0005] To achieve the above object, the present invention provides a method for enhancing point cloud data of irregular stockpiles based on a diffusion model, comprising:

[0006] Collect raw point cloud data of stockpiles from the mining environment and build a point cloud diffusion model;

[0007] After adding noise to the original point cloud data of the stockpile using an adaptive noise adding strategy, the point cloud diffusion model is trained, and after the training is completed, a target point cloud diffusion model is obtained;

[0008] Extracting multi-level geometric features of the original point cloud data of the stockpile using the pre-trained multi-scale neural network of the target point cloud diffusion model, and generating point cloud data using the multi-level geometric features to obtain generated point cloud data;

[0009] The generated point cloud data and the original point cloud data of the material pile are integrated to obtain enhanced material pile point cloud data.

[0010] Optionally, before training the point cloud diffusion model after adding noise to the original point cloud data of the pile using an adaptive noise addition strategy, it also includes performing region segmentation processing on the original point cloud data to obtain multiple sub-region point cloud data, and extracting geometric features from each sub-region point cloud data to obtain sub-region geometric features.

[0011] Optionally, the performing region segmentation processing on the original point cloud data to obtain a plurality of sub-region point cloud data includes:

[0012] Standardizing the original point cloud data to obtain standard original point cloud data;

[0013] Initializing a clustering center of the standard original point cloud data, and clustering the standard point cloud data into a plurality of point cloud data clusters according to the clustering center;

[0014] Each of the point cloud data clusters is regarded as a sub-region point cloud data.

[0015] Optionally, extracting geometric features from each of the sub-region point cloud data includes:

[0016] Extracting point cloud curvature information from the sub-region point cloud data to obtain point cloud curvature features;

[0017] A data point is randomly selected from the sub-region point cloud data, and the point cloud density information of the sub-region point cloud data is calculated using a K-nearest neighbor algorithm to obtain a point cloud density feature.

[0018] Optionally, the adding noise to the original point cloud data of the stockpile by using an adaptive noise adding strategy includes:

[0019] An adaptive noise addition rate is constructed, and noise is added to the original point cloud data of the stockpile using a pre-constructed noise addition formula and the adaptive noise addition rate.

[0020] Optionally, the multi-scale neural network includes:

[0021] An input layer receives the original point cloud data of the pile;

[0022] A multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile;

[0023] The feature fusion layer performs feature fusion on the local geometric features, mesoscale geometric features and global geometric features to obtain multi-level geometric features.

[0024] Optionally, the multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile, including:

[0025] Performing a neighborhood search on each point cloud data in the original point cloud data of the stockpile according to a K-nearest neighbor algorithm to obtain neighborhood point cloud data of each point cloud data;

[0026] Performing a local convolution operation on the neighborhood point cloud data to obtain local geometric features of the original point cloud data of the pile;

[0027] Performing a dilated convolution operation on the original point cloud data of the material pile to obtain mesoscale geometric features of the original point cloud data of the material pile;

[0028] The global self-attention mechanism is used to perform global feature extraction on the original point cloud data of the pile to obtain global geometric features.

[0029] In order to solve the above problems, the present invention also provides a device for enhancing point cloud data of irregular stockpiles based on a diffusion model, the device comprising:

[0030] The point cloud diffusion model acquisition module is used to collect the original point cloud data of the stockpile from the mining environment and construct a point cloud diffusion model; after adding noise to the geometric features of the sub-region using an adaptive noise addition strategy, the point cloud diffusion model is trained, and after the training is completed, a target point cloud diffusion model is obtained;

[0031] A point cloud data generation module, used to extract multi-level geometric features of the original point cloud data of the stockpile using the pre-trained multi-scale neural network of the target point cloud diffusion model, and generate point cloud data using the multi-level geometric features to obtain generated point cloud data;

[0032] The enhanced material pile point cloud data acquisition module is used to integrate the generated point cloud data and the original point cloud data of the material pile to obtain the enhanced material pile point cloud data.

[0033] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0034] at least one processor; and,

[0035] a memory communicatively connected to the at least one processor; wherein,

[0036] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the above-mentioned method for enhancing point cloud data of irregular stockpiles based on a diffusion model.

[0037] In order to solve the above problems, the present invention also provides a computer-readable storage medium, in which at least one computer program is stored. The at least one computer program is executed by a processor in an electronic device to implement the above-mentioned method for enhancing point cloud data of irregular stockpiles based on a diffusion model.

[0038] The present invention obtains multiple sub-region point cloud data by performing region segmentation processing on the original point cloud data, which can reduce the calculation scale of point cloud data and improve the accuracy of geometric feature extraction in subsequent point cloud data. In addition, by using a pre-trained multi-scale neural network to extract multi-level geometric features of the original point cloud data of the material pile, richer geometric features can be obtained to ensure the richness of the subsequently generated point cloud data. Finally, by integrating the generated point cloud data and the original point cloud data of the material pile to obtain enhanced material pile point cloud data, the integrity and consistency of the generated enhanced material pile point cloud data can be ensured, thereby improving the data enhancement effect of the scale and diversity of the material pile data in the mining environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a flow chart of a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention;

[0040] Figure 2 The overall structure flow chart of the algorithm of the irregular stockpile point cloud data enhancement method based on the diffusion model provided in one embodiment of the present invention;

[0041] Figure 3 A schematic diagram of a point cloud diffusion model of a point cloud data enhancement method for irregular stockpiles based on a diffusion model provided in one embodiment of the present invention;

[0042] Figure 4 A multi-scale neural network structure diagram of a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention;

[0043] Figure 5 A functional module diagram of a point cloud data enhancement device for irregular stockpiles based on a diffusion model provided by an embodiment of the present invention;

[0044] Figure 6 A schematic diagram of the structure of an electronic device for implementing a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention.

[0045] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0047] The embodiment of the present application provides a method for enhancing point cloud data of irregular material piles based on a diffusion model. The execution subject of the method for enhancing point cloud data of irregular material piles based on a diffusion model includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided in the embodiment of the present application. In other words, the method for enhancing point cloud data of irregular material piles based on a diffusion model can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0048] Reference Figure 1 FIG. 1 is a flow chart of a method for enhancing point cloud data of irregular material piles based on a diffusion model according to an embodiment of the present invention. In this embodiment, the method for enhancing point cloud data of irregular material piles based on a diffusion model includes:

[0049] S1. Collect the original point cloud data of the stockpile from the mining environment and build a point cloud diffusion model.

[0050] In the embodiment of the present invention, the raw point cloud data of the material pile refers to the collected point cloud data of the loading equipment, pedestrians in the mining area, and the material pile in the mining area environment.

[0051] In the embodiment of the present invention, the point cloud diffusion model refers to a model for generating point cloud data. The point cloud diffusion model can synthesize point cloud data and perform data enhancement on the point cloud data.

[0052] The embodiment of the present invention collects raw point cloud data of a stockpile from a mining environment, and may use different types of sensors such as a laser radar, a structured light scanner, and a depth camera to collect and acquire the data, so as to ensure the diversity of the point cloud data.

[0053] S2. After adding noise to the geometric features of the sub-region using an adaptive noise adding strategy, the point cloud diffusion model is trained, and a target point cloud diffusion model is obtained after the training is completed.

[0054] As an embodiment of the present invention, before training the point cloud diffusion model after adding noise to the original point cloud data of the material pile using an adaptive noise addition strategy, it also includes performing region segmentation processing on the original point cloud data to obtain multiple sub-region point cloud data, and extracting geometric features from each of the sub-region point cloud data to obtain sub-region geometric features.

[0055] Furthermore, the region segmentation processing is performed on the original point cloud data to obtain a plurality of sub-region point cloud data, including:

[0056] Standardizing the original point cloud data to obtain standard original point cloud data;

[0057] Initializing a clustering center of the standard original point cloud data, and clustering the standard point cloud data into a plurality of point cloud data clusters according to the clustering center;

[0058] Each of the point cloud data clusters is regarded as a sub-region point cloud data.

[0059] In the embodiment of the present invention, the standardization processing of the original point cloud data includes:

[0060] The original point cloud data of the stockpile is subjected to noise removal, and each point cloud data in the point cloud data after the noise removal is subjected to coordinate alignment processing.

[0061] In the embodiment of the present invention, the original point cloud data of the pile is Indicates that x i represents the i-th data point, and the standard original point cloud data after the original point cloud data is standardized is represented as

[0062] As an embodiment of the present invention, the step of extracting geometric features from the point cloud data of each sub-region includes:

[0063] Extracting point cloud curvature information from the sub-region point cloud data to obtain point cloud curvature features;

[0064] A data point is randomly selected from the sub-region point cloud data, and the point cloud density information of the sub-region point cloud data is calculated using a K-nearest neighbor algorithm to obtain a point cloud density feature.

[0065] In the embodiment of the present invention, the point cloud curvature information refers to parameters describing local features in the point cloud data, and reflects the curvature degree of the point cloud data.

[0066] In the embodiment of the present invention, the point cloud density information refers to information describing the density of the spatial distribution of point clouds in point cloud data, reflecting the number of point clouds per unit area.

[0067] In the embodiment of the present invention, the K-nearest neighbor algorithm refers to a basic classification and regression method, and different distance measurement methods can be used to determine the nearest neighbor.

[0068] As an embodiment of the present invention, the step of adding noise to the original point cloud data of the pile using an adaptive noise adding strategy includes:

[0069] An adaptive noise addition rate is constructed, and noise is added to the original point cloud data of the stockpile using a pre-constructed noise addition formula and the adaptive noise addition rate.

[0070] Further, the adding noise to the original point cloud data of the stockpile by using the pre-built noise adding formula and the adaptive noise adding rate includes:

[0071] The following noise adding formula is used to add noise to the original point cloud data of the pile:

[0072]

[0073] in, is the noise generation result of the tth step in the forward diffusion process, represents the point cloud data of the tth step in the forward diffusion process, D i is the point cloud density feature, C i is the point cloud curvature feature, is a normal distribution, β μ is the basic noise addition rate, I is the identity matrix, β t (D i ,C i ) is the calculation formula of the adaptive noise rate.

[0074] Furthermore, the calculation formula of the adaptive noise rate is calculated using the following formula:

[0075]

[0076] Among them, α is the first weight parameter, γ is the second weight parameter, C max It is the maximum value of the point cloud curvature information.

[0077] The embodiment of the present invention can better preserve the geometric information of the stockpile by adaptively adjusting the noise adding rate and dynamically adjusting the size and distribution of the noise according to the geometric characteristics of the sub-region.

[0078] As an embodiment of the present invention, training the point cloud diffusion model includes:

[0079] The maximization of likelihood estimation is taken as the optimization goal of the point cloud diffusion model, and the loss function of the variational lower bound is used to calculate the loss value of the noise generation result predicted by the point cloud diffusion model in the t-th step and the noise generation result in the t-th step in the forward diffusion process. When the loss value meets the preset loss value threshold, the training of the point cloud diffusion model is completed to obtain the target point cloud diffusion model.

[0080] In the embodiment of the present invention, the maximum likelihood estimation refers to a solution for making an optimal decision under uncertain conditions.

[0081] Furthermore, the loss function using the variational lower bound is used to calculate the loss value of the noise generation result predicted by the point cloud diffusion model at the tth step and the noise generation result at the tth step in the forward diffusion process, including:

[0082] The following loss function formula is used to calculate the loss value L(θ) between the noise generation result predicted by the point cloud diffusion model at the tth step and the noise generation result at the tth step in the forward diffusion process:

[0083]

[0084] Among them, E q Take the expectation of the distribution q, T refers to the total number of time steps, D KL represents KL divergence, θ is the multi-scale neural network parameter, Refers to the noise generation result predicted by the point cloud diffusion model at the tth step.

[0085] S3. Using the pre-trained multi-scale neural network of the target point cloud diffusion model to extract multi-level geometric features of the original point cloud data of the stockpile, and using the multi-level geometric features to generate point cloud data to obtain generated point cloud data.

[0086] As an embodiment of the present invention, the multi-scale neural network includes:

[0087] An input layer receives the original point cloud data of the pile;

[0088] A multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile;

[0089] The feature fusion layer performs feature fusion on the local geometric features, mesoscale geometric features and global geometric features to obtain multi-level geometric features.

[0090] In the embodiment of the present invention, the pre-trained multi-scale neural network also includes a denoising prediction module, including:

[0091] The denoising prediction module is used to perform mean prediction, and based on the fused multi-level geometric features, the mean value μ of the point cloud learned by the multi-scale neural network parameter θ after denoising is predicted. θ ;

[0092] The denoising prediction module is used to perform variance prediction to predict the variance learned by the multi-scale neural network parameter θ in the denoising process.

[0093] Furthermore, the multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile, including:

[0094] Performing a neighborhood search on each point cloud data in the original point cloud data of the stockpile according to a K-nearest neighbor algorithm to obtain neighborhood point cloud data of each point cloud data;

[0095] Performing a local convolution operation on the neighborhood point cloud data to obtain local geometric features of the original point cloud data of the pile;

[0096] Performing a dilated convolution operation on the original point cloud data of the material pile to obtain mesoscale geometric features of the original point cloud data of the material pile;

[0097] The global self-attention mechanism is used to perform global feature extraction on the original point cloud data of the pile to obtain global geometric features.

[0098] In the embodiment of the present invention, the multi-level geometric features are used to perform reverse denoising on the original generated point cloud data, and the following formula can be used:

[0099]

[0100] in, represents the data generation process at step t, is the mean value learned by the multi-scale neural network parameter θ, which indicates the prediction of the next step of data. Represents the variance learned by the multi-scale neural network parameter θ, and multi-scale features refer to the multi-level geometric features used in the inverse denoising process.

[0101] S4. Integrate the generated point cloud data and the original point cloud data of the material pile to obtain enhanced material pile point cloud data.

[0102] Exemplarily, the step of integrating the generated point cloud data with the original point cloud data of the stockpile to obtain the enhanced stockpile point cloud data may be implemented by the following steps:

[0103] Random noise is input into the trained target point cloud diffusion model, and the data is restored from the Gaussian distribution state to a state close to the original data through the inverse process, and new stockpile point cloud data is generated from the noise.

[0104] The generated point cloud data is integrated with the original point cloud data of the material pile, and the data is cleaned and deduplicated to obtain the enhanced point cloud data of the material pile.

[0105] The embodiment of the present invention integrates the generated point cloud data and the original point cloud data of the material pile to obtain enhanced material pile point cloud data, thereby ensuring the integrity and consistency of the generated enhanced material pile point cloud data.

[0106] In the embodiment of the present invention, the generated point cloud data has similar features to the original point cloud data of the stockpile, but contains more changes and diversity.

[0107] The present invention obtains multiple sub-region point cloud data by performing region segmentation processing on the original point cloud data, which can reduce the calculation scale of point cloud data and improve the accuracy of geometric feature extraction in subsequent point cloud data. In addition, by using a pre-trained multi-scale neural network to extract multi-level geometric features of the original point cloud data of the material pile, richer geometric features can be obtained to ensure the richness of the subsequently generated point cloud data. Finally, by integrating the generated point cloud data and the original point cloud data of the material pile to obtain enhanced material pile point cloud data, the integrity and consistency of the generated enhanced material pile point cloud data can be ensured, thereby improving the data enhancement effect of the scale and diversity of the material pile data in the mining environment.

[0108] Reference Figure 2 , which is a flowchart of the overall algorithm structure of a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention.

[0109] Reference Figure 3 , which is a schematic diagram of the point cloud diffusion model principle structure of a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention.

[0110] Reference Figure 4 , which is a multi-scale neural network structure diagram of a method for enhancing point cloud data of irregular stockpiles based on a diffusion model provided in one embodiment of the present invention.

[0111] like Figure 5 , which is a functional module diagram of a point cloud data enhancement device for irregular stockpiles based on a diffusion model provided by an embodiment of the present invention.

[0112] The device 100 for enhancing point cloud data of irregular stockpiles based on a diffusion model of the present invention can be installed in an electronic device. According to the functions to be implemented, the device 100 for enhancing point cloud data of irregular stockpiles based on a diffusion model can include a point cloud diffusion model acquisition module 101, a point cloud data generation module 102, and an enhanced stockpile point cloud data acquisition module 103.

[0113] The module described in the present invention may also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and is stored in a memory of the electronic device.

[0114] In this embodiment, the functions of each module / unit are as follows:

[0115] The point cloud diffusion model acquisition module 101 is used to collect original point cloud data of the stockpile from the mining environment and construct a point cloud diffusion model; after adding noise to the geometric features of the sub-region using an adaptive noise addition strategy, the point cloud diffusion model is trained, and a target point cloud diffusion model is obtained after the training is completed.

[0116] In the embodiment of the present invention, the raw point cloud data of the material pile refers to the collected point cloud data of the loading equipment, pedestrians in the mining area, and the material pile in the mining area environment.

[0117] In the embodiment of the present invention, the point cloud diffusion model refers to a model for generating point cloud data. The point cloud diffusion model can synthesize point cloud data and perform data enhancement on the point cloud data.

[0118] The embodiment of the present invention collects raw point cloud data of a stockpile from a mining environment, and may use different types of sensors such as a laser radar, a structured light scanner, and a depth camera to collect and acquire the data, so as to ensure the diversity of the point cloud data.

[0119] As an embodiment of the present invention, before training the point cloud diffusion model after adding noise to the original point cloud data of the material pile using an adaptive noise addition strategy, it also includes performing region segmentation processing on the original point cloud data to obtain multiple sub-region point cloud data, and extracting geometric features from each of the sub-region point cloud data to obtain sub-region geometric features.

[0120] Furthermore, the region segmentation processing is performed on the original point cloud data to obtain a plurality of sub-region point cloud data, including:

[0121] Standardizing the original point cloud data to obtain standard original point cloud data;

[0122] Initializing a clustering center of the standard original point cloud data, and clustering the standard point cloud data into a plurality of point cloud data clusters according to the clustering center;

[0123] Each of the point cloud data clusters is regarded as a sub-region point cloud data.

[0124] In the embodiment of the present invention, the standardization processing of the original point cloud data includes:

[0125] The original point cloud data of the stockpile is subjected to noise removal, and each point cloud data in the point cloud data after the noise removal is subjected to coordinate alignment processing.

[0126] In the embodiment of the present invention, the original point cloud data of the pile is Indicates that x i represents the i-th data point, and the standard original point cloud data after the original point cloud data is standardized is represented as

[0127] As an embodiment of the present invention, the step of extracting geometric features from the point cloud data of each sub-region includes:

[0128] Extracting point cloud curvature information from the sub-region point cloud data to obtain point cloud curvature features;

[0129] A data point is randomly selected from the sub-region point cloud data, and the point cloud density information of the sub-region point cloud data is calculated using a K-nearest neighbor algorithm to obtain a point cloud density feature.

[0130] In the embodiment of the present invention, the point cloud curvature information refers to parameters describing local features in the point cloud data, and reflects the curvature degree of the point cloud data.

[0131] In the embodiment of the present invention, the point cloud density information refers to information describing the density of the spatial distribution of point clouds in point cloud data, reflecting the number of point clouds per unit area.

[0132] In the embodiment of the present invention, the K-nearest neighbor algorithm refers to a basic classification and regression method, and different distance measurement methods can be used to determine the nearest neighbor.

[0133] As an embodiment of the present invention, the step of adding noise to the sub-region geometric features by using an adaptive noise adding strategy includes:

[0134] An adaptive noise addition rate is constructed, and noise is added to the original point cloud data of the stockpile using a pre-constructed noise addition formula and the adaptive noise addition rate.

[0135] Further, the adding noise to the original point cloud data of the stockpile by using the pre-built noise adding formula and the adaptive noise adding rate includes:

[0136] The following noise adding formula is used to add noise to the original point cloud data of the pile:

[0137]

[0138] in, is the noise generation result of the tth step in the forward diffusion process, represents the point cloud data of the tth step in the forward diffusion process, D i is the point cloud density feature, C i is the point cloud curvature feature, is a normal distribution, β μ is the basic noise addition rate, I is the identity matrix, β t (D i ,C i ) is the calculation formula of the adaptive noise rate.

[0139] Furthermore, the calculation formula of the adaptive noise rate is calculated using the following formula:

[0140]

[0141] Among them, α is the first weight parameter, γ is the second weight parameter, C max It is the maximum value of the point cloud curvature information.

[0142] The embodiment of the present invention can better preserve the geometric information of the stockpile by adaptively adjusting the noise adding rate and dynamically adjusting the size and distribution of the noise according to the geometric characteristics of the sub-region.

[0143] As an embodiment of the present invention, training the point cloud diffusion model includes:

[0144] The maximization of likelihood estimation is taken as the optimization goal of the point cloud diffusion model, and the loss function of the variational lower bound is used to calculate the loss value of the noise generation result predicted by the point cloud diffusion model in the t-th step and the noise generation result in the t-th step in the forward diffusion process. When the loss value meets the preset loss value threshold, the training of the point cloud diffusion model is completed to obtain the target point cloud diffusion model.

[0145] In the embodiment of the present invention, the maximum likelihood estimation refers to a solution for making an optimal decision under uncertain conditions.

[0146] Furthermore, the loss function using the variational lower bound is used to calculate the loss value of the noise generation result predicted by the point cloud diffusion model at the tth step and the noise generation result at the tth step in the forward diffusion process, including:

[0147] The following loss function formula is used to calculate the loss value L(θ) between the noise generation result predicted by the point cloud diffusion model at the tth step and the noise generation result at the tth step in the forward diffusion process:

[0148]

[0149] Among them, E q Take the expectation of the distribution q, T refers to the total number of time steps, D KL represents KL divergence, θ is the multi-scale neural network parameter, Refers to the noise generation result predicted by the point cloud diffusion model at the tth step.

[0150] The point cloud data generation module 102 is used to extract multi-level geometric features of the original point cloud data of the stockpile using the pre-trained multi-scale neural network of the target point cloud diffusion model, and generate point cloud data using the multi-level geometric features to obtain generated point cloud data.

[0151] As an embodiment of the present invention, the multi-scale neural network includes:

[0152] An input layer receives the original point cloud data of the pile;

[0153] A multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile;

[0154] The feature fusion layer performs feature fusion on the local geometric features, mesoscale geometric features and global geometric features to obtain multi-level geometric features.

[0155] In the embodiment of the present invention, the pre-trained multi-scale neural network also includes a denoising prediction module, including:

[0156] The denoising prediction module is used to perform mean prediction, and based on the fused multi-level geometric features, the mean value μ of the point cloud learned by the multi-scale neural network parameter θ after denoising is predicted. θ ;

[0157] The denoising prediction module is used to perform variance prediction to predict the variance learned by the multi-scale neural network parameter θ in the denoising process.

[0158] Furthermore, the multi-scale feature extraction module extracts local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the stockpile, including:

[0159] Performing a neighborhood search on each point cloud data in the original point cloud data of the stockpile according to a K-nearest neighbor algorithm to obtain neighborhood point cloud data of each point cloud data;

[0160] Performing a local convolution operation on the neighborhood point cloud data to obtain local geometric features of the original point cloud data of the pile;

[0161] Performing a dilated convolution operation on the original point cloud data of the material pile to obtain mesoscale geometric features of the original point cloud data of the material pile;

[0162] The global self-attention mechanism is used to perform global feature extraction on the original point cloud data of the pile to obtain global geometric features.

[0163] In the embodiment of the present invention, the multi-level geometric features are used to perform reverse denoising on the original generated point cloud data, and the following formula can be used:

[0164]

[0165] in, represents the data generation process at step t, is the mean value learned by the multi-scale neural network parameter θ, which indicates the prediction of the next step of data. Represents the variance learned by the multi-scale neural network parameter θ, and multi-scale features refer to the multi-level geometric features used in the inverse denoising process.

[0166] The enhanced material pile point cloud data acquisition module 103 is used to integrate the generated point cloud data and the original point cloud data of the material pile to obtain enhanced material pile point cloud data.

[0167] The generated point cloud data and the original point cloud data of the material pile are integrated to obtain enhanced material pile point cloud data.

[0168] like Figure 6 , which is a schematic diagram of the structure of an electronic device for implementing a method for enhancing point cloud data of irregular stockpiles based on a diffusion model according to an embodiment of the present invention.

[0169] The electronic device may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a method program for enhancing point cloud data of irregular stockpiles based on a diffusion model.

[0170] In some embodiments, the processor 10 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 10 is the control core (ControlUnit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (for example, executing a method program for enhancing point cloud data of irregular stockpiles based on a diffusion model, etc.), and calls the data stored in the memory 11 to execute various functions of the electronic device and process data.

[0171] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 can also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 11 can also include both an internal storage unit of the electronic device and an external storage device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device, such as a code of a program for a method for enhancing point cloud data of an irregular stockpile based on a diffusion model, but can also be used to temporarily store data that has been output or is to be output.

[0172] The communication bus 12 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0173] The communication interface 13 is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device and to display a visual user interface.

[0174] Figure 6 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 6 The structure shown does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0175] For example, although not shown, the electronic device may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that the power management device can realize functions such as charging management, discharging management, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0176] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0177] The program of the method for enhancing point cloud data of irregular stockpiles based on a diffusion model stored in the memory 11 of the electronic device is a combination of multiple instructions. When running in the processor 10, the following can be achieved:

[0178] Collect raw point cloud data of stockpiles from the mining environment and build a point cloud diffusion model;

[0179] After adding noise to the geometric features of the sub-region using an adaptive noise adding strategy, the point cloud diffusion model is trained, and after the training is completed, a target point cloud diffusion model is obtained;

[0180] Extracting multi-level geometric features of the original point cloud data of the stockpile using the pre-trained multi-scale neural network of the target point cloud diffusion model, and generating point cloud data using the multi-level geometric features to obtain generated point cloud data;

[0181] The generated point cloud data and the original point cloud data of the material pile are integrated to obtain enhanced material pile point cloud data.

[0182] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, which will not be repeated here.

[0183] Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0184] The present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor of an electronic device, the computer program can implement:

[0185] Collect raw point cloud data of stockpiles from the mining environment and build a point cloud diffusion model;

[0186] After adding noise to the geometric features of the sub-region using an adaptive noise adding strategy, the point cloud diffusion model is trained, and after the training is completed, a target point cloud diffusion model is obtained;

[0187] Extracting multi-level geometric features of the original point cloud data of the stockpile using the pre-trained multi-scale neural network of the target point cloud diffusion model, and generating point cloud data using the multi-level geometric features to obtain generated point cloud data;

[0188] The generated point cloud data and the original point cloud data of the material pile are integrated to obtain enhanced material pile point cloud data.

[0189] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0190] 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 units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0192] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0193] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0194] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0195] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0196] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for enhancing point cloud data of irregular stockpiles based on a diffusion model, characterized in that: The method comprises: Collect raw point cloud data of stockpiles from the mining environment and build a point cloud diffusion model; After adding noise to the original point cloud data of the stockpile using an adaptive noise adding strategy, the point cloud diffusion model is trained, and after the training is completed, a target point cloud diffusion model is obtained; The multi-scale neural network pre-trained by the target point cloud diffusion model is used to extract the multi-level geometric features of the original point cloud data of the pile, and the multi-level geometric features are used to generate point cloud data to obtain generated point cloud data, wherein the multi-scale neural network includes: an input layer, receiving the original point cloud data of the pile; a multi-scale feature extraction module, extracting local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the pile; a feature fusion layer, fusing the local geometric features, mesoscale geometric features and global geometric features to obtain multi-level geometric features, wherein, Extracting local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the pile, including: performing a neighborhood search on each point cloud data in the original point cloud data of the pile according to the K-nearest neighbor algorithm to obtain the neighborhood point cloud data of each point cloud data; performing a local convolution operation on the neighborhood point cloud data to obtain the local geometric features of the original point cloud data of the pile; performing a hole convolution operation on the original point cloud data of the pile to obtain the mesoscale geometric features of the original point cloud data of the pile; and performing global feature extraction on the original point cloud data of the pile using a global self-attention mechanism to obtain the global geometric features; The generated point cloud data and the original point cloud data of the material pile are integrated to obtain enhanced material pile point cloud data.

2. The method for enhancing point cloud data of irregular stockpiles based on a diffusion model according to claim 1, characterized in that: Before training the point cloud diffusion model after adding noise to the original point cloud data of the pile using the adaptive noise addition strategy, the method also includes performing region segmentation processing on the original point cloud data to obtain multiple sub-region point cloud data, and extracting geometric features from each sub-region point cloud data to obtain sub-region geometric features.

3. The method for enhancing point cloud data of irregular stockpiles based on diffusion model according to claim 2, characterized in that: The performing region segmentation processing on the original point cloud data to obtain a plurality of sub-region point cloud data includes: Standardizing the original point cloud data to obtain standard original point cloud data; Initializing a clustering center of the standard original point cloud data, and clustering the standard original point cloud data into a plurality of point cloud data clusters according to the clustering center; Each of the point cloud data clusters is regarded as a sub-region point cloud data.

4. The method for enhancing point cloud data of irregular stockpiles based on a diffusion model according to claim 2, characterized in that: The step of extracting geometric features from each sub-region point cloud data comprises: Extracting point cloud curvature information from the sub-region point cloud data to obtain point cloud curvature features; A data point is randomly selected from the sub-region point cloud data, and the point cloud density information of the sub-region point cloud data is calculated using a K-nearest neighbor algorithm to obtain a point cloud density feature.

5. The method for enhancing point cloud data of irregular stockpiles based on a diffusion model according to any one of claims 1 to 4, characterized in that: The step of adding noise to the original point cloud data of the stockpile by using an adaptive noise adding strategy includes: An adaptive noise addition rate is constructed, and noise is added to the original point cloud data of the stockpile using a pre-constructed noise addition formula and the adaptive noise addition rate.

6. A point cloud data enhancement device for irregular stockpiles based on a diffusion model, characterized in that: The device can implement the irregular stockpile point cloud data enhancement method based on the diffusion model as claimed in any one of claims 1 to 5, and the device comprises: The point cloud diffusion model acquisition module is used to collect the original point cloud data of the material pile from the mining environment and construct a point cloud diffusion model; add noise to the original point cloud data of the material pile using an adaptive noise addition strategy, and then train the point cloud diffusion model, and obtain the target point cloud diffusion model after the training is completed; The point cloud data generation module is used to extract the multi-level geometric features of the original point cloud data of the pile using the multi-scale neural network pre-trained by the target point cloud diffusion model, and generate point cloud data using the multi-level geometric features to obtain generated point cloud data, wherein the multi-scale neural network includes: an input layer, receiving the original point cloud data of the pile; a multi-scale feature extraction module, extracting local geometric features, mesoscale geometric features and global geometric features from the original point cloud data of the pile; a feature fusion layer, fusing the local geometric features, mesoscale geometric features and global geometric features to obtain a multi-level geometric feature. Features, wherein the local geometric features, mesoscale geometric features and global geometric features in the original point cloud data of the material pile are extracted, including: performing a neighborhood search on each point cloud data in the original point cloud data of the material pile according to the K-nearest neighbor algorithm to obtain the neighborhood point cloud data of each point cloud data; performing a local convolution operation on the neighborhood point cloud data to obtain the local geometric features of the original point cloud data of the material pile; performing a hole convolution operation on the original point cloud data of the material pile to obtain the mesoscale geometric features of the original point cloud data of the material pile; using a global self-attention mechanism to perform global feature extraction on the original point cloud data of the material pile to obtain global geometric features; The enhanced material pile point cloud data acquisition module is used to integrate the generated point cloud data and the original point cloud data of the material pile to obtain the enhanced material pile point cloud data.

7. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the irregular material pile point cloud data enhancement method based on the diffusion model as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for enhancing point cloud data of irregular stockpiles based on a diffusion model as described in any one of claims 1 to 5 is implemented.

Images