A neighborhood structure-aware self-attention landslide segmentation method for point clouds

Abstract

This invention discloses a self-attention landslide segmentation method based on point cloud neighborhood structure perception, belonging to the field of landslide detection technology. It involves acquiring laser point cloud data of a landslide area; using the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model for computation, obtaining the landslide classification and recognition results for each point in the laser point cloud data; the landslide segmentation model includes multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer. The multiple neighborhood perception and feature encoding layers are used to sample the point cloud data at different scales, extract features, determine the neighborhood perception coefficients of the sampled points, and update the features of the sampled points; the multiple neighborhood perception and feature decoding layers are used to decode and concatenate the updated features so that they can be used as input to the output layer to obtain the landslide classification and recognition results. The self-attention landslide segmentation method based on point cloud neighborhood structure perception disclosed in this invention can improve the accuracy of landslide detection.

Description

Technical Field

[0001] This invention belongs to the field of landslide detection technology, specifically relating to a self-attention landslide segmentation method based on point cloud neighborhood structure perception. Background Technology

[0002] Landslides are a common natural disaster that poses a serious threat to human life and property. Effective landslide detection is of great significance for disaster relief, reconstruction and vegetation restoration planning.

[0003] Currently, landslide detection is often performed by landslide segmentation based on point cloud data. However, due to the diverse landslide morphologies and complex boundaries in natural scenes, landslide segmentation based on point cloud data often lacks the ability to extract features of the overall outline and local details of the landslide and fails to utilize the landslide boundary structure. As a result, landslide segmentation based on point cloud data is difficult to achieve the expected results, and the accuracy of landslide detection is poor.

[0004] Therefore, how to provide an effective solution to improve the accuracy of landslide detection has become a pressing problem to be solved in existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide a self-attention slippery segmentation method for point cloud neighborhood structure perception, in order to solve the above-mentioned problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a self-attention slippage segmentation method for point cloud neighborhood structure awareness, comprising:

[0008] Acquire laser point cloud data of the landslide area;

[0009] The point cloud data of each point in the laser point cloud data is used as the input of a pre-trained landslide segmentation model for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

[0010] The landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer.

[0011] The multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales.

[0012] Calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points;

[0013] Feature extraction is performed on the features of each sampling point in point cloud data at different scales, and the features of each sampling point and its neighboring sampling points in point cloud data at different scales are extracted.

[0014] Based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales is determined. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points.

[0015] Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales.

[0016] The multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in the point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

[0017] Based on the above-disclosed content, this invention acquires laser point cloud data of a landslide area; uses the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model for computation, and obtains the landslide classification and identification results of each point in the laser point cloud data; wherein, the landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer; the multiple neighborhood perception and feature encoding layers are used to sample and extract features from the input point cloud data at different scales; calculate the slope corresponding to each sampling point in the point cloud data at different scales and the slope variation between each sampling point and its neighboring sampling points; extract features from the features of each sampling point in the point cloud data at different scales, and extract the features of each sampling point and its neighboring sampling points in the point cloud data at different scales; based on the point cloud data at different scales... Based on the features corresponding to each sampling point and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales are determined. The neighborhood perception coefficients are used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points. Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales. Multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and recognition results of each point in the laser point cloud data. Thus, during landslide detection, by extracting features from sampling points and their neighboring sampling points, a neighborhood perception coefficient representing the degree of influence of neighboring sampling points on corresponding sampling points is determined, and the features of the sampling points are updated. This takes into account the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation. At the same time, by sampling and feature extraction at different scales, the ability to extract the overall contour and local details of the landslide can be improved. Therefore, during landslide segmentation, both the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation and the overall contour and local details of the landslide can be taken into account, thereby improving the accuracy of landslide detection.

[0018] In one possible design, the calculation of the slope corresponding to each sampling point in point cloud data at different scales and the slope variation between each sampling point and its neighboring sampling points includes:

[0019] The slope corresponding to each sampling point in point cloud data at different scales was calculated by fitting a plane using principal component analysis.

[0020] Based on the slope corresponding to each sampling point in point cloud data at different scales, and the slope corresponding to the neighboring sampling points of each sampling point in point cloud data at different scales, the slope variation between each sampling point and its neighboring sampling points in point cloud data at different scales is calculated.

[0021] In one possible design, the slope corresponding to the sampling point is calculated according to the following formula (1);

[0022]

[0023] Where, n x n y and n z These represent the components of the minimum eigenvector of the point cloud data obtained by principal component analysis in the three coordinate directions, respectively.

[0024] The slope variation between a sampling point and its neighboring sampling points is calculated according to the following formula (2);

[0025]

[0026] Where, θ i Let θ represent the slope corresponding to the i-th sampling point. ij This represents the slope corresponding to the j-th neighboring sampling point of the i-th sampling point.

[0027] In one possible design, the neighborhood perception coefficient of the neighborhood sampling points corresponding to each sampling point in the point cloud data at different scales is determined according to the following formula (3);

[0028]

[0029] Where α and β represent linear transformations, f i f represents the feature corresponding to the i-th sampling point. ij This represents the feature of the j-th neighboring sampling point of the i-th sampling point. Let λ represent the position encoding of the i-th sampling point and its j-th neighboring sampling point in three-dimensional space, and d represent the multilayer perceptron. k Let ξ represent the dimension of the key vector of the neighborhood-aware self-attention module in the neighborhood-aware and feature-encoding layer, and let ξ represent the neighborhood-aware encoding. Linear represents a linear transformation, σ ReLU This represents the ReLU activation function.

[0030] In one possible design, the features corresponding to the sampling points are updated according to the following formula (4);

[0031]

[0032] Where ⊙ represents the Hadamard product, and p represents the neighborhood sampling point corresponding to the sampling point. f represents the set of neighboring sampling points corresponding to the sampling point. ij This represents the characteristics of the neighboring sampling points corresponding to the sampling point.

[0033] In one possible design, the plurality of neighborhood sensing and feature encoding layers include a first neighborhood sensing and feature encoding layer, a second neighborhood sensing and feature encoding layer, a third neighborhood sensing and feature encoding layer, a fourth neighborhood sensing and feature encoding layer, and a fifth neighborhood sensing and feature encoding layer; the plurality of neighborhood sensing and feature decoding layers include a first neighborhood sensing and feature decoding layer, a second neighborhood sensing and feature decoding layer, a third neighborhood sensing and feature decoding layer, a fourth neighborhood sensing and feature decoding layer, and a fifth neighborhood sensing and feature decoding layer; the first neighborhood sensing and feature encoding layer includes a first neighborhood sensing layer and a first encoding layer. The second neighborhood perception and feature encoding layer includes a second neighborhood perception layer and a second encoding layer; the third neighborhood perception and feature encoding layer includes a third neighborhood perception layer and a third encoding layer; the fourth neighborhood perception and feature encoding layer includes a fourth neighborhood perception layer and a fourth encoding layer; the fifth neighborhood perception and feature encoding layer includes a fifth neighborhood perception layer and a fifth encoding layer; the first neighborhood perception and feature decoding layer includes a first decoding layer and a first neighborhood perception layer; the second neighborhood perception and feature decoding layer includes a second decoding layer and a second neighborhood perception layer; and the third neighborhood perception and feature decoding layer includes a third decoding layer. The fourth neighborhood perception and feature decoding layer includes a fourth decoding layer and the fourth neighborhood perception layer, and the fifth neighborhood perception and feature decoding layer includes a fifth decoding layer and the fifth neighborhood perception layer. The first neighborhood perception layer, the first encoding layer, the second neighborhood perception layer, the second encoding layer, the third neighborhood perception layer, the third encoding layer, the fourth neighborhood perception layer, the fourth encoding layer, the fifth neighborhood perception layer, and the fifth encoding layer are sequentially connected. The first decoding layer, the fifth neighborhood perception layer, the second decoding layer, and the fourth neighborhood perception layer are... The first decoding layer, the third decoding layer, the third neighborhood sensing layer, the fourth decoding layer, the second neighborhood sensing layer, the fifth decoding layer, and the first neighborhood sensing layer are connected in sequence. The output of the first decoding layer is also connected to the input of the fifth decoding layer, the output of the second decoding layer is also connected to the input of the fourth decoding layer, the output of the third decoding layer is also connected to the input of the third decoding layer, the output of the fourth decoding layer is also connected to the input of the second decoding layer, and the output of the fifth decoding layer is connected to the input of the first decoding layer through multiple sensing layers.

[0034] In one possible design, the first neighborhood perception and feature encoding layer, the second neighborhood perception and feature encoding layer, the third neighborhood perception and feature encoding layer, the fourth neighborhood perception and feature encoding layer, and the fifth neighborhood perception and feature encoding layer sample the input point cloud data at different scales according to sampling numbers of N, N / 4, N / 16, N / 64, and N / 256, respectively, where N represents the total amount of data corresponding to the point cloud data. The first neighborhood perception and feature decoding layer, the second neighborhood perception and feature decoding layer, the third neighborhood perception and feature decoding layer, the fourth neighborhood perception and feature decoding layer, and the fifth neighborhood perception and feature decoding layer upscale the concatenated features according to multiples of 1, 2, 4, 8, and 16, respectively.

[0035] In a second aspect, the present invention provides a self-attention slippage segmentation device for point cloud neighborhood structure perception, comprising:

[0036] The acquisition unit is used to acquire laser point cloud data of the landslide area;

[0037] The computing unit is used to take the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model and perform calculations to obtain the landslide classification and recognition results of each point in the laser point cloud data.

[0038] The landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer.

[0039] The multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales.

[0040] Calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points;

[0041] Feature extraction is performed on the features of each sampling point in point cloud data at different scales, and the features of each sampling point and its neighboring sampling points in point cloud data at different scales are extracted.

[0042] Based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales is determined. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points.

[0043] Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales.

[0044] The multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in the point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

[0045] Thirdly, the present invention provides an electronic device comprising a memory, a processor, and a transceiver connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the self-attention slippage segmentation method for point cloud neighborhood structure awareness as described in the first aspect or any possible design of the first aspect.

[0046] Fourthly, the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, perform the self-attention slippery segmentation method for point cloud neighborhood structure perception as described in the first aspect or any possible design of the first aspect.

[0047] Fifthly, the present invention provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform a self-attention slippery segmentation method for point cloud neighborhood structure awareness as described in the first aspect or any possible design of the first aspect.

[0048] Beneficial effects:

[0049] The self-attention landslide segmentation method based on point cloud neighborhood structure perception provided by this invention, during landslide detection, extracts features of sampling points and their neighboring sampling points, determines a neighborhood perception coefficient that characterizes the degree of influence of neighboring sampling points on corresponding sampling points, and updates the features of the sampling points. This takes into account the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation. At the same time, by sampling and feature extraction at different scales, the ability to extract the overall contour and local details of the landslide can be improved. Thus, during landslide segmentation, both the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation and the overall contour and local details of the landslide can be considered, thereby improving the accuracy of landslide detection and facilitating practical application and promotion. Attached Figure Description

[0050] Figure 1 A flowchart of a self-attention landslide segmentation method for point cloud neighborhood structure perception provided in an embodiment of this application;

[0051] Figure 2 This is a network structure diagram of the landslide segmentation model provided in the embodiments of this application;

[0052] Figure 3A block diagram of a self-attention landslide segmentation device for point cloud neighborhood structure perception provided in an embodiment of this application;

[0053] Figure 4 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0055] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0056] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0057] To achieve landslide detection, this application provides a self-attention landslide segmentation method based on point cloud neighborhood structure perception, which can improve the accuracy of landslide detection.

[0058] The point cloud neighborhood structure-aware self-attention slippery slope segmentation method provided in this application can be applied to user terminals or servers. It is understood that the execution entity described does not constitute a limitation on the embodiments of this application.

[0059] The self-attention landslide segmentation method for point cloud neighborhood structure perception provided in the embodiments of this application will be described in detail below.

[0060] like Figure 1The diagram shown is a flowchart of a self-attention slippery segmentation method for point cloud neighborhood structure perception provided in the first aspect of the present application. This self-attention slippery segmentation method for point cloud neighborhood structure perception may include, but is not limited to, the following steps S101-S102.

[0061] Step S101. Obtain laser point cloud data of the landslide area.

[0062] The laser point cloud data includes point cloud data of multiple points. The point cloud data includes three-dimensional spatial coordinates and may also include color attributes, such as the values ​​of the R, G, and B color channels.

[0063] Step S102. Use the point cloud data of each point in the laser point cloud data as input to the pre-trained landslide segmentation model for calculation, and obtain the landslide classification and recognition results of each point in the laser point cloud data.

[0064] In this embodiment of the application, a landslide segmentation model for landslide detection is pre-trained. After obtaining the laser point cloud data of the landslide area, the point cloud data of each point in the laser point cloud data can be used as the input of the pre-trained landslide segmentation model for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data. The landslide classification and identification results can be the probability that each point in the laser point cloud data is a landslide point or a non-landslide point.

[0065] The landslide segmentation model includes an input layer, multiple neighborhood sensing and feature encoding layers, multiple neighborhood sensing and feature decoding layers, and an output layer. The input layer is connected to the multiple neighborhood sensing and feature encoding layers, and the multiple neighborhood sensing and feature decoding layers are connected to the output layer.

[0066] Multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales.

[0067] To ensure that more points are sampled in areas of high curvature and to control the uniformity of the spatial distribution of the sampled points, this application designs a curvature-based sampling method. First, for a point cloud of N points, principal component analysis is used to fit a plane to each point and calculate the average curvature. Then, the point cloud is sorted according to curvature from largest to smallest. When M points need to be sampled, a threshold η (ranging from 0 to 1) can be set, and the sampling interval is the first (1-η)N+ηM points in the sorted point cloud. Then, within this sampling interval, the farthest point sampling algorithm is used to sample M more points.

[0068] Multiple neighborhood perception and feature encoding layers are also used to calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points.

[0069] In this embodiment, the slope corresponding to each sampling point in point cloud data at different scales can be calculated by fitting a plane using Principal Component Analysis (PCA). Then, based on the slope corresponding to each sampling point in point cloud data at different scales, and the slope corresponding to the neighboring sampling points of each sampling point in point cloud data at different scales, the slope variation between each sampling point and its neighboring sampling points in point cloud data at different scales can be calculated.

[0070] When calculating the slope corresponding to each sampling point, the slope corresponding to the sampling point can be calculated according to the following formula (1);

[0071]

[0072] Where, n x n y and n z These represent the components of the minimum eigenvector of the point cloud data obtained by principal component analysis in the three coordinate directions.

[0073] When calculating the slope variation between a sampling point and its neighboring sampling points, the slope variation between a sampling point and its neighboring sampling points can be calculated according to the following formula (2);

[0074]

[0075] Where, θ i Let θ represent the slope corresponding to the i-th sampling point. ij This represents the slope corresponding to the j-th neighboring sampling point of the i-th sampling point. This indicates the gradient variation between the i-th sample and its j-th neighboring sample points.

[0076] Multiple neighborhood perception and feature encoding layers are also used to extract features from each sampling point in point cloud data at different scales, thereby extracting the features of each sampling point and its neighboring sampling points in point cloud data at different scales.

[0077] In this embodiment of the application, given point cloud data, the features of each sampling point and its neighboring sampling points in point cloud data of different scales can be obtained through linear transformation.

[0078] Multiple neighborhood perception and feature encoding layers are also used to determine the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales.

[0079] In this embodiment, an attention mechanism can be used to calculate the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points. Specifically, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales can be determined according to the following formula (3);

[0080]

[0081] Where α and β represent linear transformations, f i f represents the feature corresponding to the i-th sampling point. ij This represents the feature of the j-th neighboring sampling point of the i-th sampling point. Let λ represent the position encoding of the i-th sampling point and its j-th neighboring sampling point in three-dimensional space, and d represent the multilayer perceptron. k Let ξ represent the dimension of the key vector of the neighborhood-aware self-attention module in the neighborhood-aware and feature-encoding layer, and let ξ represent the neighborhood-aware encoding. Linear represents a linear transformation, σ ReLU This represents the ReLU activation function.

[0082] Multiple neighborhood perception and feature encoding layers are also used to update the features corresponding to each sampling point in point cloud data at different scales based on the neighborhood perception coefficients of the neighborhood sampling points corresponding to each sampling point in point cloud data at different scales, so as to obtain the updated features corresponding to each sampling point in point cloud data at different scales. In this embodiment, the neighborhood perception coefficients can be weighted and aggregated to update the features of the sampling points. Specifically, the features corresponding to the sampling points can be updated according to the following formula (4);

[0083]

[0084] Where ⊙ represents the Hadamard product, and p represents the neighborhood sampling point corresponding to the sampling point. f represents the set of neighboring sampling points corresponding to the sampling point. ij This represents the characteristics of the neighboring sampling points corresponding to the sampling point.

[0085] Multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

[0086] like Figure 2In the embodiments described herein, the multiple neighborhood perception and feature encoding layers include a first neighborhood perception and feature encoding layer, a second neighborhood perception and feature encoding layer, a third neighborhood perception and feature encoding layer, a fourth neighborhood perception and feature encoding layer, and a fifth neighborhood perception and feature encoding layer. The first neighborhood perception and feature encoding layer, the second neighborhood perception and feature encoding layer, the third neighborhood perception and feature encoding layer, the fourth neighborhood perception and feature encoding layer, and the fifth neighborhood perception and feature encoding layer can sequentially sample the input point cloud data at different scales according to sampling numbers of N, N / 4, N / 16, N / 64, and N / 256, where N represents the total amount of data corresponding to the point cloud data. Multiple neighborhood perception and feature decoding layers include a first neighborhood perception and feature decoding layer, a second neighborhood perception and feature decoding layer, a third neighborhood perception and feature decoding layer, a fourth neighborhood perception and feature decoding layer, and a fifth neighborhood perception and feature decoding layer. The first neighborhood perception and feature decoding layer, the second neighborhood perception and feature decoding layer, the third neighborhood perception and feature decoding layer, the fourth neighborhood perception and feature decoding layer, and the fifth neighborhood perception and feature decoding layer can sequentially increase the dimensionality of the spliced ​​features in multiples of 1, 2, 4, 8, and 16.

[0087] The first neighborhood perception and feature encoding layer includes a first neighborhood perception layer and a first encoding layer; the second neighborhood perception and feature encoding layer includes a second neighborhood perception layer and a second encoding layer; the third neighborhood perception and feature encoding layer includes a third neighborhood perception layer and a third encoding layer; the fourth neighborhood perception and feature encoding layer includes a fourth neighborhood perception layer and a fourth encoding layer; the fifth neighborhood perception and feature encoding layer includes a fifth neighborhood perception layer and a fifth encoding layer; the first neighborhood perception and feature decoding layer includes a first decoding layer and a first neighborhood perception layer; the second neighborhood perception and feature decoding layer includes a second decoding layer and a second neighborhood perception layer; the third neighborhood perception and feature decoding layer includes a third decoding layer and the third neighborhood perception layer; the fourth neighborhood perception and feature decoding layer includes a fourth decoding layer and the fourth neighborhood perception layer; and the fifth neighborhood perception and feature decoding layer includes a fifth decoding layer and the fifth neighborhood perception layer. The first encoding layer, the second neighborhood sensing layer, the second encoding layer, the third neighborhood sensing layer, the third encoding layer, the fourth neighborhood sensing layer, the fourth encoding layer, the fifth neighborhood sensing layer, and the fifth encoding layer are connected in sequence. The first decoding layer, the fifth neighborhood sensing layer, the second decoding layer, the fourth neighborhood sensing layer, the third decoding layer, the third neighborhood sensing layer, the fourth decoding layer, the second neighborhood sensing layer, the fifth decoding layer, and the first neighborhood sensing layer are connected in sequence. The output of the first encoding layer is also connected to the input of the fifth decoding layer. The output of the second encoding layer is also connected to the input of the fourth decoding layer. The output of the third encoding layer is also connected to the input of the third decoding layer. The output of the fourth encoding layer is also connected to the input of the second decoding layer. The output of the fifth encoding layer and the input of the first decoding layer are connected through multiple sensing layers.

[0088] During the encoding phase, point cloud data (including 3D coordinates and color attributes) of N points is input, with an input feature dimension D=6. First, a multilayer perceptron (MLP) layer is used for encoding, abstracting the initial 6-dimensional features to a higher dimension C=32. In subsequent encoding phases, each layer embeds neighborhood-aware self-attention, sampling the point data according to N→N / 4→N / 16→N / 64→N / 256. The sampled points in each layer are grouped using the k-nearest neighbor algorithm. Finally, a shared MLP is used to increase the feature dimension according to C→2C→4C→8C→16C.

[0089] In the decoding phase, multi-scale features extracted from each encoding layer are fused and propagated to classify the point cloud. During decoding, each layer fuses low-level and high-level features through residual connections and embeds a neighborhood-aware self-attention module in each layer, employing inverse distance interpolation to upsample the point cloud. In the final step, a multilayer perceptron is used to predict point cloud labels using a normalized exponential (Softmax) function, with an output dimension of 2, representing the probabilities of predicting a landslide point as a landslide point and a non-landslide point as a non-landslide point.

[0090] In summary, the self-attention landslide segmentation method based on point cloud neighborhood structure perception provided by this invention acquires laser point cloud data of a landslide area; uses the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model for computation, and obtains the landslide classification and recognition results of each point in the laser point cloud data; wherein, the landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer; the multiple neighborhood perception and feature encoding layers are used to sample and extract features from the input point cloud data at different scales; calculate the slope corresponding to each sampling point in the point cloud data at different scales and the slope variation between each sampling point and its neighboring sampling points; extract features from the features of each sampling point in the point cloud data at different scales, and extract the features of each sampling point and its neighboring sampling points in the point cloud data at different scales; Based on the features corresponding to each sampling point in point cloud data at different scales, and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales are determined. The neighborhood perception coefficients are used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points. Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales. Multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and recognition results of each point in the laser point cloud data. Thus, in landslide detection, by extracting features from sampling points and their neighboring sampling points, a neighborhood perception coefficient representing the degree of influence of neighboring sampling points on corresponding sampling points is determined, and the features of the sampling points are updated. This takes into account the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation. At the same time, by sampling and feature extraction at different scales, the ability to extract the overall contour and local details of the landslide can be improved. Therefore, in landslide segmentation, the influence of the neighborhood structure of the landslide boundary point cloud on landslide segmentation can be taken into account, as well as the overall contour and local details of the landslide, thereby improving the accuracy of landslide detection and facilitating practical application and promotion.

[0091] Please see Figure 3The second aspect of this application provides a self-attention landslide segmentation device for point cloud neighborhood structure perception, the self-attention landslide segmentation device for point cloud neighborhood structure perception includes:

[0092] The acquisition unit is used to acquire laser point cloud data of the landslide area;

[0093] The computing unit is used to take the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model and perform calculations to obtain the landslide classification and recognition results of each point in the laser point cloud data.

[0094] The landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer.

[0095] The multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales.

[0096] Calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points;

[0097] Feature extraction is performed on the features of each sampling point in point cloud data at different scales, and the features of each sampling point and its neighboring sampling points in point cloud data at different scales are extracted.

[0098] Based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales is determined. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points.

[0099] Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales.

[0100] The multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in the point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

[0101] The working process, working details and technical effects of the point cloud neighborhood structure perception self-attention landslide segmentation device provided in the second aspect of this embodiment can be found in the first aspect of the embodiment, and will not be repeated here.

[0102] like Figure 4As shown, a third aspect of this application provides an electronic device, including a memory, a processor, and a transceiver that are sequentially and communicatively connected. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the self-attention slippage segmentation method for point cloud neighborhood structure awareness as described in the first aspect of the application.

[0103] Specifically, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or last-in-first-out (FILO) memory, etc.; the processor may not be limited to microprocessors of the STM32F105 series, ARM (Advanced RISC Machines), x86 architecture processors, or processors with integrated NPU (neural-network processing units); the transceiver may be, but is not limited to, WiFi (Wireless Fidelity) wireless transceivers, Bluetooth wireless transceivers, General Packet Radio Service (GPRS) wireless transceivers, ZigBee (a low-power LAN protocol based on the IEEE 802.15.4 standard), 3G transceivers, 4G transceivers, and / or 5G transceivers, etc.

[0104] This fourth aspect of the embodiment provides a computer-readable storage medium storing instructions containing the self-attention slippery segmentation method for point cloud neighborhood structure awareness as described in the first aspect of the embodiment. Specifically, the computer-readable storage medium stores instructions that, when executed on a computer, perform the self-attention slippery segmentation method for point cloud neighborhood structure awareness as described in the first aspect. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or Memory Sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0105] The fifth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the self-attention slippery segmentation method for point cloud neighborhood structure awareness as described in the first aspect of this embodiment, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0106] It should be understood that specific details are provided in the following description to facilitate a complete understanding of the exemplary embodiments. However, those skilled in the art will understand that the exemplary embodiments can be implemented without these specific details. For example, the system may be shown in block diagrams to avoid obscuring the example with unnecessary details. In other instances, well-known processes, structures, and techniques may be shown without unnecessary details to avoid obscuring the exemplary embodiments.

[0107] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A self-attention slippage segmentation method based on point cloud neighborhood structure perception, characterized in that, include: Acquire laser point cloud data of the landslide area; The point cloud data of each point in the laser point cloud data is used as the input of a pre-trained landslide segmentation model for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data. The landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer. The multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales. Calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points; Feature extraction is performed on the features of each sampling point in point cloud data at different scales, and the features of each sampling point and its neighboring sampling points in point cloud data at different scales are extracted. Based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales is determined. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points. Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales. The multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in the point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

2. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 1, characterized in that, The calculation of the slope corresponding to each sampling point in point cloud data at different scales and the slope variation between each sampling point and its neighboring sampling points includes: The slope corresponding to each sampling point in point cloud data at different scales was calculated by fitting a plane using principal component analysis. Based on the slope corresponding to each sampling point in point cloud data at different scales, and the slope corresponding to the neighboring sampling points of each sampling point in point cloud data at different scales, the slope variation between each sampling point and its neighboring sampling points in point cloud data at different scales is calculated.

3. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 2, characterized in that, Calculate the slope corresponding to the sampling point according to the following formula (1); Where, n x n y and n z These represent the components of the minimum eigenvector of the point cloud data obtained by principal component analysis in the three coordinate directions, respectively. The slope variation between a sampling point and its neighboring sampling points is calculated according to the following formula (2); Where, θ i Let θ represent the slope corresponding to the i-th sampling point. ij This represents the slope corresponding to the j-th neighboring sampling point of the i-th sampling point.

4. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 1, characterized in that, The neighborhood perception coefficients of the neighborhood sampling points corresponding to each sampling point in the point cloud data at different scales are determined according to the following formula (3); Where α and β represent linear transformations, f i f represents the feature corresponding to the i-th sampling point. ij This represents the feature of the j-th neighboring sampling point of the i-th sampling point. Let λ represent the position encoding of the i-th sampling point and its j-th neighboring sampling point in three-dimensional space, and d represent the multilayer perceptron. k Let ξ represent the dimension of the key vector of the neighborhood-aware self-attention module in the neighborhood-aware and feature-encoding layer, and let ξ represent the neighborhood-aware encoding. Linear represents a linear transformation, σ ReLU This represents the ReLU activation function.

5. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 4, characterized in that, The features corresponding to the sampling points are updated according to the following formula (4); Where ⊙ represents the Hadamard product, and p represents the neighborhood sampling point corresponding to the sampling point. f represents the set of neighboring sampling points corresponding to the sampling point. ij This represents the characteristics of the neighboring sampling points corresponding to the sampling point.

6. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 1, characterized in that, The plurality of neighborhood perception and feature encoding layers include a first neighborhood perception and feature encoding layer, a second neighborhood perception and feature encoding layer, a third neighborhood perception and feature encoding layer, a fourth neighborhood perception and feature encoding layer, and a fifth neighborhood perception and feature encoding layer. The plurality of neighborhood perception and feature decoding layers include a first neighborhood perception and feature decoding layer, a second neighborhood perception and feature decoding layer, a third neighborhood perception and feature decoding layer, a fourth neighborhood perception and feature decoding layer, and a fifth neighborhood perception and feature decoding layer. The first neighborhood perception and feature encoding layer includes a first neighborhood perception layer and a first encoding layer. The second neighborhood perception and feature decoding layer... The feature encoding layer includes a second neighborhood perception layer and a second encoding layer; the third neighborhood perception and feature encoding layer includes a third neighborhood perception layer and a third encoding layer; the fourth neighborhood perception and feature encoding layer includes a fourth neighborhood perception layer and a fourth encoding layer; the fifth neighborhood perception and feature encoding layer includes a fifth neighborhood perception layer and a fifth encoding layer; the first neighborhood perception and feature decoding layer includes a first decoding layer and a first neighborhood perception layer; the second neighborhood perception and feature decoding layer includes a second decoding layer and a second neighborhood perception layer; and the third neighborhood perception and feature decoding layer includes a third decoding layer and a third... The neighborhood perception layer, the fourth neighborhood perception and feature decoding layer includes a fourth decoding layer and the fourth neighborhood perception layer, the fifth neighborhood perception and feature decoding layer includes a fifth decoding layer and the fifth neighborhood perception layer, the first neighborhood perception layer, the first encoding layer, the second neighborhood perception layer, the second encoding layer, the third neighborhood perception layer, the third encoding layer, the fourth neighborhood perception layer, the fourth encoding layer, the fifth neighborhood perception layer and the fifth encoding layer are connected in sequence, the first decoding layer, the fifth neighborhood perception layer, the second decoding layer, the fourth neighborhood perception layer, the third decoding layer, the third neighborhood perception layer, the fourth decoding layer, the second neighborhood perception layer, the fifth decoding layer and the first neighborhood perception layer are connected in sequence, the output of the first encoding layer is also connected to the input of the fifth decoding layer, the output of the second encoding layer is also connected to the input of the fourth decoding layer, the output of the third encoding layer is also connected to the input of the third decoding layer, the output of the fourth encoding layer is also connected to the input of the second decoding layer, and the output of the fifth encoding layer is connected to the input of the first decoding layer through a multi-layer perception layer.

7. The self-attention slippage segmentation method for point cloud neighborhood structure perception according to claim 6, characterized in that, The first neighborhood perception and feature encoding layer, the second neighborhood perception and feature encoding layer, the third neighborhood perception and feature encoding layer, the fourth neighborhood perception and feature encoding layer, and the fifth neighborhood perception and feature encoding layer sample the input point cloud data at different scales according to the sampling numbers N, N / 4, N / 16, N / 64, and N / 256, respectively, where N represents the total amount of data corresponding to the point cloud data. The first neighborhood perception and feature decoding layer, the second neighborhood perception and feature decoding layer, the third neighborhood perception and feature decoding layer, the fourth neighborhood perception and feature decoding layer, and the fifth neighborhood perception and feature decoding layer upscale the spliced ​​features according to multiples of 1, 2, 4, 8, and 16, respectively.

8. A self-attention landslide segmentation device for point cloud neighborhood structure perception, characterized in that, include: The acquisition unit is used to acquire laser point cloud data of the landslide area; The computing unit is used to take the point cloud data of each point in the laser point cloud data as input to a pre-trained landslide segmentation model and perform calculations to obtain the landslide classification and recognition results of each point in the laser point cloud data. The landslide segmentation model includes an input layer, multiple neighborhood perception and feature encoding layers, multiple neighborhood perception and feature decoding layers, and an output layer. The multiple neighborhood sensing and feature encoding layers are used to sample and extract features from the input point cloud data at different scales. Calculate the slope corresponding to each sampling point in point cloud data at different scales, as well as the slope variation between each sampling point and its neighboring sampling points; Feature extraction is performed on the features of each sampling point in point cloud data at different scales, and the features of each sampling point and its neighboring sampling points in point cloud data at different scales are extracted. Based on the features corresponding to each sampling point in point cloud data at different scales and the features of the neighboring sampling points of each sampling point in point cloud data at different scales, the neighborhood perception coefficient of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales is determined. The neighborhood perception coefficient is used to characterize the degree of influence of the neighboring sampling points on the corresponding sampling points. Based on the neighborhood perception coefficients of the neighboring sampling points corresponding to each sampling point in point cloud data at different scales, the features corresponding to each sampling point in point cloud data at different scales are updated to obtain the updated features corresponding to each sampling point in point cloud data at different scales. The multiple neighborhood perception and feature decoding layers are used to decode and stitch together the updated features corresponding to each sampling point in the point cloud data at different scales to obtain stitched features, so that the stitched features can be used as input to the output layer for calculation to obtain the landslide classification and identification results of each point in the laser point cloud data.

9. An electronic device, characterized in that, The device includes a memory, a processor, and a transceiver that are sequentially and communicatively connected. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the self-attention slippage segmentation method for point cloud neighborhood structure awareness as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or the instructions are executed by the computer, they implement the self-attention slippage segmentation method for point cloud neighborhood structure awareness as described in any one of claims 1 to 7.

Images