LiDar point cloud data extraction method and device suitable for glacier

Abstract

The invention discloses a LiDar point cloud data extraction method and device suitable for a glacier. The method comprises the following steps: S1, acquiring laser radar point cloud data and a high-resolution remote sensing image of a target glacier area in a historical period; s2, constructing a glacier recognition model according to the laser radar point cloud data and the high-resolution remote sensing image of the target glacier region in the historical period; wherein the glacier identification model is used for describing a time sequence neural radiation model of feature change; s3, inputting the laser radar point cloud data and the high-resolution remote sensing image of the target glacier area in the current period into the time sequence neural radiation model, and extracting LiDar point cloud data; wherein the LiDar point cloud data is used for describing glacier area change. By adopting the technical scheme of the invention, the glacier data can be automatically identified and detected from the high-resolution remote sensing image.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser radar, and in particular relates to a method and device for extracting LiDar point cloud data suitable for glaciers. Background Art

[0002] The cryosphere is one of the five major spheres of the Earth's climate system. It is composed of various glaciers, cirques, glacial erosion, and frozen soil, among which glaciers account for the majority. However, due to the global warming caused by the continuous accumulation of the greenhouse effect in recent years, the loss of global glaciers and glacial materials has accelerated sharply, bringing many potential crises to human life in the future. At present, the main method for measuring glacier data is field surveys. This method not only consumes a lot of manpower and material resources, but also cannot conduct large-scale dynamic monitoring of global glaciers with only a few field survey data. There is no completely objective and true data to specifically reflect the retreat trend, speed and degree of glaciers, resulting in the limitations of the obtained glacier change trends and analysis conclusions in the assessment of global climate change. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a LiDar point cloud data extraction method and device suitable for glaciers, so as to realize automatic identification and detection of glacier data, such as glacier area changes, from high-resolution remote sensing images.

[0004] To achieve the above object, the present invention adopts the following technical solution:

[0005] A LiDar point cloud data extraction method applicable to glaciers, comprising:

[0006] Step S1, obtaining lidar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period;

[0007] Step S2, constructing a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes;

[0008] Step S3, inputting the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

[0009] Preferably, step S1 further includes cleaning the historical lidar point cloud data and the high-resolution remote sensing images, and in step S2, a glacier recognition model is constructed based on the cleaned historical lidar point cloud data and the high-resolution remote sensing images.

[0010] Preferably, in step S2, a deep learning convolutional neural network model based on a deep residual network is trained according to the historical lidar point cloud data and high-resolution remote sensing images after data cleaning to obtain a time series neural radiation model.

[0011] Preferably, step S3 comprises:

[0012] Based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period, the LiDar point cloud characteristics of the glacier in the historical period are obtained;

[0013] Based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period, the LiDar point cloud of the glacier in the current period is obtained;

[0014] The historical glacier LiDar point cloud data and the current glacier LiDar point cloud are input into the time series neural radiation model to calculate the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud.

[0015] The present invention also provides a LiDar point cloud data extraction device suitable for glaciers, comprising:

[0016] An acquisition module is used to obtain lidar point cloud data and high-resolution remote sensing images of the target glacier area in historical periods;

[0017] A construction module is used to construct a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes;

[0018] The extraction module is used to input the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

[0019] Preferably, the acquisition module is also used to clean the historical lidar point cloud data and high-resolution remote sensing images, and the construction module is used to construct a glacier recognition model based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning.

[0020] Preferably, the construction module trains a deep learning convolutional neural network model based on a deep residual network according to the historical lidar point cloud data and high-resolution remote sensing images after data cleaning to obtain a time series neural radiation model.

[0021] Preferably, the extraction module comprises:

[0022] The first extraction unit is used to obtain the LiDar point cloud features of the glacier in the historical period based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period;

[0023] The second extraction unit is used to obtain the LiDar point cloud of the glacier in the current period based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period;

[0024] The calculation unit is used to input the historical glacier LiDar point cloud data and the current glacier LiDar point cloud into the time series neural radiation model, and calculate the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud.

[0025] The present invention constructs a glacier recognition model based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes; step S3, input the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model, and extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area. The technical solution of the present invention is adopted to realize the automatic recognition and detection of glacier data from high-resolution remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0027] Figure 1 The figure is a flow chart of a method for extracting LiDar point cloud data applicable to glaciers according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Embodiment 1:

[0031] like Figure 1 As shown, an embodiment of the present invention provides a LiDar point cloud data extraction method applicable to glaciers, comprising:

[0032] Step S1, obtaining lidar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period;

[0033] Step S2, constructing a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes;

[0034] Step S3, inputting the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

[0035] As an implementation method of an embodiment of the present invention, step S1 also includes data cleaning of historical lidar point cloud data and high-resolution remote sensing images. In step S2, a glacier recognition model is constructed based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning.

[0036] Furthermore, data cleaning of historical LiDAR point cloud data and high-resolution remote sensing images includes:

[0037] By cleaning the historical LiDAR point cloud data and high-resolution remote sensing images, garbled data and offside data are screened out to obtain screening index data;

[0038] Performing vectorization operation on the screening index data to obtain a screening index vector set, and adding a time-space vector to each screening index vector in the screening index vector set to obtain a time-space vector set;

[0039] Performing feature clustering on the space-time vector set to obtain a cluster center data set;

[0040] Fill each empty value data in the screening index data according to the cluster center data set.

[0041] As an implementation method of an embodiment of the present invention, in step S2, time series data for describing changes in image features are extracted based on historical lidar point cloud data and high-resolution remote sensing images after data cleaning, and a deep learning convolutional neural network model based on a deep residual network is trained using the time series data to obtain a time series neural radiation model.

[0042] Furthermore, the process of using a deep residual network to construct a deep learning convolutional neural network model based on a deep residual network is as follows: input the time series data into the first convolutional layer of the deep residual network, perform convolution, regularization, activation function, and maximum pooling calculations on the input, change the number of channels and size of the original image, and obtain a deep learning convolutional neural network model based on a deep residual network.

[0043] Furthermore, a deep learning convolutional neural network model based on a deep residual network is trained to obtain a time series neural radiation model, specifically:

[0044] The deep residual network includes four residual blocks, each of which has three layers of convolution. The time series data is input into the residual block, and firstly, a 1*1 convolution is performed to reduce the dimension, and then a convolution operation with a pooling kernel of 3*3 is performed, and finally, a 1*1 convolution is performed to restore its dimension. After the corresponding convolution operation is performed in each residual block, the output feature map is superimposed with the feature map of the previous layer. After the corresponding convolution operation of the residual block, a global average pooling operation is finally performed to complete the training of the deep learning convolutional neural network model based on the deep residual network.

[0045] Furthermore, the deep residual network is ResNet50, which includes forty-nine convolutional layers and one global average pooling layer.

[0046] Training process: For the time series data used to describe the changes in image features, the training set and the test set are divided into a ratio of 8:2, and then the deep learning convolutional neural network model based on the deep residual network is trained with the training set; the relevant parameters of the deep learning convolutional neural network model based on the deep residual network are debugged, and the quality of the trained deep learning convolutional neural network model based on the deep residual network is judged according to the model evaluation indicators until the time series neural radiation model is obtained.

[0047] The embodiment of the present invention can further use the coefficient of determination R2 and the mean squared error MSE (mean squared error) to evaluate the training and testing results of the model to obtain the optimal time series neural radiation model.

[0048] As an implementation of the embodiment of the present invention, step S3 includes:

[0049] Based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period, the LiDar point cloud characteristics of the glacier in the historical period are obtained;

[0050] Based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period, the LiDar point cloud of the glacier in the current period is obtained;

[0051] The historical glacier LiDar point cloud data and the current glacier LiDar point cloud are input into the time series neural radiation model, the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud is calculated, and the motion displacement of the feature points is calculated to realize the automatic identification and detection of glacier area changes from high-resolution remote sensing images.

[0052] Embodiment 2:

[0053] The embodiment of the present invention further provides a LiDar point cloud data extraction device applicable to glaciers, comprising:

[0054] An acquisition module is used to obtain lidar point cloud data and high-resolution remote sensing images of the target glacier area in historical periods;

[0055] A construction module is used to construct a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes;

[0056] The extraction module is used to input the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

[0057] As an implementation method of an embodiment of the present invention, the acquisition module is also used to clean the historical lidar point cloud data and high-resolution remote sensing images, and the construction module is used to construct a glacier recognition model based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning.

[0058] As an implementation method of an embodiment of the present invention, a construction module trains a deep learning convolutional neural network model based on a deep residual network according to historical lidar point cloud data and high-resolution remote sensing images after data cleaning to obtain a time series neural radiation model.

[0059] As an implementation of an embodiment of the present invention, the extraction module includes:

[0060] The first extraction unit is used to obtain the LiDar point cloud features of the glacier in the historical period based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period;

[0061] The second extraction unit is used to obtain the LiDar point cloud of the glacier in the current period based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period;

[0062] The calculation unit is used to input the historical glacier LiDar point cloud data and the current glacier LiDar point cloud into the time series neural radiation model, and calculate the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud.

[0063] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A LiDar point cloud data extraction method suitable for glaciers, characterized in that: include: Step S1, obtaining lidar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; Step S2, constructing a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes; Step S3, inputting the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

2. The LiDar point cloud data extraction method applicable to glaciers according to claim 1, characterized in that: Step S1 also includes data cleaning of historical lidar point cloud data and high-resolution remote sensing images. In step S2, a glacier recognition model is constructed based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning.

3. The LiDar point cloud data extraction method applicable to glaciers according to claim 2, characterized in that: In step S2, a deep learning convolutional neural network model based on a deep residual network is trained based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning to obtain a time series neural radiation model.

4. The LiDar point cloud data extraction method applicable to glaciers according to claim 3, characterized in that: Step S3 includes: Based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period, the LiDar point cloud characteristics of the glacier in the historical period are obtained; Based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period, the LiDar point cloud of the glacier in the current period is obtained; The historical glacier LiDar point cloud data and the current glacier LiDar point cloud are input into the time series neural radiation model to calculate the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud.

5. A LiDar point cloud data extraction device suitable for glaciers, characterized in that: include: An acquisition module is used to obtain lidar point cloud data and high-resolution remote sensing images of the target glacier area in historical periods; A construction module is used to construct a glacier recognition model based on the laser radar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; wherein the glacier recognition model is used to describe the time series neural radiation model of feature changes; The extraction module is used to input the lidar point cloud data and high-resolution remote sensing images of the target glacier area in the current period into the time series neural radiation model to extract the LiDar point cloud data; wherein the LiDar point cloud data is used to describe the change in glacier area.

6. The LiDar point cloud data extraction device suitable for glaciers according to claim 5, characterized in that: The acquisition module is also used to clean the historical lidar point cloud data and high-resolution remote sensing images, and the construction module is used to construct a glacier recognition model based on the historical lidar point cloud data and high-resolution remote sensing images after data cleaning.

7. The LiDar point cloud data extraction device suitable for glaciers according to claim 6, characterized in that: The construction module trains a deep learning convolutional neural network model based on a deep residual network according to the historical lidar point cloud data and high-resolution remote sensing images after data cleaning to obtain a time series neural radiation model.

8. The LiDar point cloud data extraction device for glaciers according to claim 7, characterized in that: The extraction module includes: The first extraction unit is used to obtain the LiDar point cloud features of the glacier in the historical period based on the historical LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the historical period; The second extraction unit is used to obtain the LiDar point cloud of the glacier in the current period based on the LiDar point cloud data and high-resolution remote sensing images of the target glacier area in the current period; The calculation unit is used to input the historical glacier LiDar point cloud data and the current glacier LiDar point cloud into the time series neural radiation model, and calculate the position of the historical glacier LiDar point cloud features in the current glacier LiDar point cloud.

Images