Landslide susceptibility assessment methods, devices, electronic equipment, storage media and products

Abstract

The present invention provides a landslide susceptibility assessment method, device, electronic device, storage medium and product, the method comprising: obtaining a remote sensing image to be measured; inputting the remote sensing image to be measured into a landslide susceptibility assessment model, and obtaining a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model; wherein the landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide. Smoothed samples are added through the Gaussian heat map of the landslide to optimize the distribution of samples, further improving the performance and efficiency of the assessment model, and obtaining more accurate prediction results.

Description

Technical Field

[0001] The present invention relates to the technical field of landslide remote sensing identification, and in particular to a landslide susceptibility assessment method, device, electronic equipment, storage medium and product. Background Art

[0002] Landslides are one of the most destructive natural disasters in the world caused by both human and natural factors. Earthquakes, as one of the triggering factors of landslides, are prone to cause thousands of landslides due to the special geomorphic environment in mountainous areas. High spatial resolution remote sensing images and drones have been widely used in landslide investigation and identification. However, it is difficult to predict the probability of future landslides based on image band information alone. Therefore, landslide susceptibility mapping is of great significance for post-earthquake emergency response and disaster prevention and mitigation.

[0003] Landslide susceptibility is a prediction of the likelihood of a landslide at a given location based on factors such as topography, geology, and hydrology. Several methods have been applied to landslide susceptibility mapping and have been tested and proven to be available and effective, including geomorphic mapping, physics-based models, heuristic terrain and susceptibility zoning, and machine learning methods. Traditional machine learning methods are shallow structures and cannot learn more representative deep features. Deep learning, especially convolutional neural networks, has a deep network structure and powerful feature learning capabilities, and is increasingly used in the field of landslide susceptibility mapping.

[0004] In the landslide sample construction method based on convolutional neural network, positive samples represented by landslides are marked as "1", while negative samples represented by non-landslides are marked as "0". Among them, positive samples are often real and reliable, while the selection of negative samples is often subjective or selected according to certain rules. Since the training of convolutional neural network models is simply based on positive and negative samples, the results often tend to be close to the maximum and minimum values ​​of the susceptibility index. At the same time, existing studies mainly solve the overfitting problem by simplifying the network structure and reducing the number of iterations. These strategies may cause the model to be unable to learn rich features, thereby affecting the model performance. Therefore, the existing methods have the risk of being unable to balance model performance and overfitting caused by single sample training. Summary of the invention

[0005] The present invention provides a landslide susceptibility assessment method, device, electronic device, storage medium and product, which are used to solve the defects of the evaluation model in the prior art and realize the accurate assessment of landslide susceptibility.

[0006] The present invention provides a landslide susceptibility assessment method, comprising:

[0007] Acquire the remote sensing image to be measured;

[0008] Inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model;

[0009] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0010] According to a landslide susceptibility assessment method provided by the present invention, the landslide sample set is obtained based on the Gaussian heat map of the landslide, including:

[0011] Acquire the landslide remote sensing image and extract the center point of the landslide;

[0012] A two-dimensional Gaussian kernel function is used to generate a Gaussian thermal map of the landslide at each center point;

[0013] Obtaining coordinate points corresponding to different landslide susceptibility according to the Gaussian heat map of the landslide;

[0014] Coordinate points corresponding to the different landslide susceptibilities are randomly selected to obtain a landslide sample set.

[0015] According to a landslide susceptibility assessment method provided by the present invention, the step of randomly extracting coordinate points corresponding to different landslide susceptibilities to obtain a landslide sample set comprises:

[0016] A landslide sample set is obtained by randomly selecting corresponding coordinate points from each landslide susceptibility interval, wherein the different landslide susceptibility intervals include five intervals with equal intervals between 0 and 1 in terms of landslide susceptibility.

[0017] According to a landslide susceptibility assessment method provided by the present invention, the Gaussian heat map of the landslide is generated point by center point using a two-dimensional Gaussian kernel function, comprising:

[0018] For a single landslide center point, a two-dimensional Gaussian kernel function is used to generate a circle with a radius of R at each center point to obtain the Gaussian thermal map of the landslide;

[0019] For multiple landslide center points, a two-dimensional Gaussian kernel function is used to generate a circle with a radius R for each center point, and after eliminating the overlapping areas of different landslide center points, a Gaussian thermal map of the landslide is obtained;

[0020] The landslide susceptibility of the center point is 1, the landslide susceptibility of the boundary of the circle is 0.1, and the landslide susceptibility outside the boundary is 0.

[0021] According to a landslide susceptibility assessment method provided by the present invention, the types of landslide susceptibility influencing factors marked on the landslide sample set include: slope, curvature, slope aspect, slope position, distance to water system, distance to road, lithology, NDVI and land utilization rate.

[0022] According to a landslide susceptibility assessment method provided by the present invention, the landslide susceptibility assessment model includes a convolutional layer, an scSE module, a maximum pooling layer, a dropout layer, a Flatten layer and a fully connected layer connected in sequence.

[0023] The present invention also provides a landslide susceptibility assessment device, comprising:

[0024] An acquisition module is used to obtain the remote sensing image to be measured;

[0025] An evaluation module, used for inputting the remote sensing image to be measured into a landslide susceptibility evaluation model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility evaluation model;

[0026] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0027] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned landslide susceptibility assessment methods are implemented.

[0028] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned landslide susceptibility assessment methods are implemented.

[0029] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of any of the above-mentioned landslide susceptibility assessment methods are implemented.

[0030] The present invention provides a landslide susceptibility assessment method, device, electronic device, storage medium and product, which add smooth samples through the Gaussian heat map of the landslide to optimize the distribution of samples. This strategy takes into account the rich environmental information and the possibility of landslides in the surrounding environment due to disasters in the future, further improves the performance and efficiency of the assessment model, and obtains more accurate prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 is a flow chart of a landslide susceptibility assessment method provided by an embodiment of the present invention;

[0033] Figure 2 It is a schematic diagram of the process of generating a landslide Gaussian thermodynamic map provided by an embodiment of the present invention;

[0034] Figure 3 It is a schematic diagram of a process for generating single and multiple landslide Gaussian thermal maps provided by an embodiment of the present invention;

[0035] Figure 4 is a schematic diagram of the structure of a landslide susceptibility assessment model provided by an embodiment of the present invention;

[0036] Figure 5 is a comparison diagram of multi-model PR curves provided by an embodiment of the present invention;

[0037] FIG6( a ) is a landslide susceptibility mapping result based on CNN provided in an embodiment of the present invention;

[0038] FIG6( b ) is a landslide susceptibility mapping result based on a landslide susceptibility assessment model provided by an embodiment of the present invention;

[0039] FIG. 7( a) is a frequency histogram of the susceptibility index based on CNN provided in an embodiment of the present invention;

[0040] FIG7( b ) is a frequency histogram of the susceptibility index based on the landslide susceptibility assessment model provided by an embodiment of the present invention;

[0041] Figure 8 is a schematic structural diagram of a landslide susceptibility assessment device provided by an embodiment of the present invention;

[0042] Fig. 9 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0044] Deep learning, especially convolutional neural networks, has a deep network structure and powerful feature learning capabilities, and is increasingly used in the field of landslide susceptibility mapping. Landslide susceptibility mapping based on convolutional neural networks takes into account rich environmental information and the influence of various factors, thereby achieving better performance than the baseline model. There are some schemes in the existing methods that take into account the subjectivity of negative sample selection. For example, the selection principles for negative samples are random sampling in non-landslide areas, random sampling in low-slope areas, and sampling based on fractal theory, or methods that select negative samples by buffering at a certain distance. And these schemes prove the accuracy of the model by optimizing negative samples.

[0045] Although some specific rules for selecting negative samples can improve the performance of the model, the existing methods of directly applying positive and negative samples to landslide susceptibility mapping have the following two main disadvantages: (1) Relying solely on positive and negative samples to participate in the training of the convolutional neural network model often leads to results that are close to the maximum and minimum values ​​of the susceptibility index. This is because these specific rules allow the model to focus on the training of easy-to-classify samples and the robust fitting of the model. (2) Existing studies mainly solve the overfitting problem by simplifying the network structure and reducing the number of iterations. These strategies may cause the model to be unable to learn rich features, thereby affecting model performance.

[0046] Therefore, a landslide susceptibility assessment method is proposed in the embodiment of the present invention, which uses the powerful spatial generalization ability of Gaussian heat map to improve the target detection accuracy. Figure 1-Figure 6(b) The present invention describes a method for assessing landslide susceptibility, comprising:

[0047] Step 101, obtaining a remote sensing image to be measured;

[0048] Step 102: inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model;

[0049] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0050] The landslide susceptibility assessment method of the embodiment of the present invention augments the landslide sample set through the Gaussian heat map of the landslide, adds smooth samples to optimize the distribution of samples, further improves the performance and efficiency of the assessment model, and obtains more accurate prediction results.

[0051] In at least one embodiment of the present invention, obtaining the landslide sample set based on the Gaussian heat map of the landslide includes:

[0052] Step 201, obtaining the landslide remote sensing image and extracting the center point of the landslide;

[0053] Step 203, using a two-dimensional Gaussian kernel function to generate a Gaussian heat map of the landslide center point by center point;

[0054] Step 202, obtaining coordinate points corresponding to different landslide susceptibility according to the Gaussian heat map of the landslide;

[0055] Step 204: Obtain a landslide sample set according to the coordinate points corresponding to the different landslide susceptibilities.

[0056] The landslide susceptibility assessment method of the embodiment of the present invention can expand some credible samples when there are insufficient samples through the Gaussian heat map generated by the Gaussian kernel function, and consider the environmental information around the landslide. At the same time, these smooth samples are considered difficult to classify, thereby enhancing the spatial generalization ability of the model and avoiding the extreme susceptibility index distribution. The generated new samples avoid the risk of overfitting while ensuring the performance of the model.

[0057] like Figure 2 The figure shows a method for generating a Gaussian heat map of a landslide, where the landslide image taken by GF-1 in figure a is the remote sensing image of the landslide. At the same time, since the closer to the location of the landslide, the greater the probability of a landslide, the remote sensing image is used to interpret the range of the landslide in figure b and extract the coordinates of the center point of the plane landslide. Combined with the Gaussian kernel function in figure c, for each center point of the landslide in figure b, the corresponding Gaussian heat map of the landslide is generated, as shown in figure d.

[0058] In at least one embodiment of the present invention, the step of randomly selecting coordinate points corresponding to different landslide susceptibilities to obtain a landslide sample set comprises:

[0059] A landslide sample set is obtained by randomly selecting corresponding coordinate points from each landslide susceptibility interval, wherein the different landslide susceptibility intervals include five intervals with equal intervals between 0 and 1 in terms of landslide susceptibility.

[0060] It should be noted that many decimals with landslide susceptibility between 0 and 1 can be obtained from the Gaussian heat map, and these decimals are divided into 5 equal intervals. Some samples are randomly selected in each interval as the landslide sample set, and the ratio of the number of samples selected is 1:1:1:1:1.

[0061] In at least one embodiment of the present invention, the step of generating the Gaussian heat map of the landslide point by center point using a two-dimensional Gaussian kernel function comprises:

[0062] For a single landslide center point, a two-dimensional Gaussian kernel function is used to generate a circle with a radius of R at each center point to obtain the Gaussian thermal map of the landslide;

[0063] For multiple landslide center points, a two-dimensional Gaussian kernel function is used to generate a circle with a radius R for each center point, and after eliminating the overlapping areas of different landslide center points, a Gaussian thermal map of the landslide is obtained;

[0064] The landslide susceptibility of the center point is 1, the landslide susceptibility of the boundary of the circle is 0.1, and the landslide susceptibility outside the boundary is 0.

[0065] like Figure 3 As shown in the figure, some problems may also be encountered in the process of generating the Gaussian heat map of the landslide in the entire area. The radius of the Gaussian kernel is the same for the same batch of samples. If the radius of the Gaussian kernel is too large, the heat map will overlap. The overlapping areas of different landslide center points are eliminated through the following methods. According to the principle that the closer the distance, the greater the impact, it is reasonable to take the maximum value of the overlap. As shown in the following formula:

[0066] F max (O1,O2,O3…O n )

[0067] O1,O2…O n =LGH1,LGH2…LGH n ∩LGH i (i=1,2…n)

[0068] LGH n is the Gaussian heat map of the landslide intersecting with the target landslide; n Representing LGH n and LGH i The intersection area, F max Represents the maximum value of the function, that is, the intersection area O n The maximum value of .

[0069] In at least one embodiment of the present invention, the types of landslide susceptibility influencing factors marked on the landslide sample set include: slope, curvature, slope aspect, slope position, distance to water system, distance to road, lithology, NDVI and land utilization rate.

[0070] It should be noted that the landslide susceptibility influencing factors are stored in the form of vectors. For example, the values ​​after classification are (5, 6, 4, 5, 11, 1, 6, 3, 3). Each type of influencing factor is standardized to obtain the corresponding score and is stored in the vector.

[0071] In at least one embodiment of the present invention, the landslide susceptibility influencing factor needs to be converted from a one-dimensional vector form to a two-dimensional image form before training.

[0072] It should be noted that the conversion method is the CNN-2D method, which follows one-hot encoding.

[0073] In at least one embodiment of the present invention, the landslide susceptibility assessment model includes a convolutional layer, a scSE module, a maximum pooling layer, a dropout layer, a Flatten layer and a fully connected layer connected in sequence.

[0074] It should be noted that if Figure 3 As shown in the figure, the landslide susceptibility assessment model is an improved model based on CNN. The network includes operations such as convolution (Conv2D), spatial and channel compression and excitation (scSE) modules, maximum pooling layers (Maxpooling), dropout layers, flattening (Flatten) layers and fully connected layers (Dense). The convolution layer is used to extract different features of the input. The first convolution layer may only extract some low-level features such as edges, lines and corners. More layers of the network can iteratively extract more complex features from low-level features. The scSE module is used to make the model pay more attention to important features and ignore minor features; the maximum pooling layer is used to reduce dimension, remove redundant information, compress features, simplify network complexity, reduce calculation amount, and reduce memory consumption; the dropout layer is used to prevent overfitting and improve efficiency; the flattening layer is used to adapt the model output; the fully connected layer is used to map the learned "distributed feature representation" to the sample label space. The model skeleton is mainly constructed through the first four operations. In order to improve the efficiency and accuracy of the model, batch normalization is added after each convolution layer of the network. The early stopping strategy can save the optimal model with high efficiency to obtain better accuracy. Unlike traditional models, Gaussian heatmaps can generate more reliable difficult-to-classify samples, improving the robustness and generalization ability of the model.

[0075] In the embodiment of the present invention, the hardware configuration is: the processor is Intel (R) Xeon (R) Silver 4210R CPU @ 2.40GHZ, the graphics card is NVIDIA GeForce RTX 3090, and the memory (RAM) is 64.0GB. The software environment is Windows 10 Professional Edition 64-bit operating system, the programming language used is Python, and tensorflow-gpu-1.4.0 and Keras-2.0.8DL frameworks are selected as tools for building models, which corresponds to Python-3.6 version.

[0076] In an embodiment of the present invention, 3592 landslide remote sensing image samples are extracted, and a Gaussian kernel function is used to generate a thermal map of the entire area. The values ​​of the corresponding pixels in the thermal map are evenly divided into 5 parts and randomly selected as new samples at equal intervals. It should be noted that since the selection of landslide and non-landslide is based on a credible traditional method, this embodiment only needs to generate new samples in the middle three intervals. After augmentation, a total of 10376 samples are obtained, and they are divided into a training set and a test set according to a ratio of 7:3. Finally, these samples are converted into two-dimensional images to participate in the training of the model.

[0077] For the landslide susceptibility assessment model (LSGH-Net), the attention mechanism of the scSE module is added after each convolution operation. By comparing the losses of the training set and the validation set, this embodiment sets the drought rate to 0.5. Considering the size of the data set, the batch size in batch normalization is set to 32. In order to ensure sufficient training, this embodiment sets the maximum number of iterations to 200. When the accuracy of the validation set does not increase significantly in 50 iterations, it is terminated early to avoid invalid training. At the same time, the Adam algorithm is used as an optimizer to accelerate model training. Since landslide susceptibility mapping is a regression problem between a factor and a sample, this embodiment selects MSE as the loss function. In the deep learning model, the final feature map is obtained by the activation function to obtain the landslide susceptibility index (between 0 and 1).

[0078] The accuracy of the landslide susceptibility assessment in this embodiment is compared below:

[0079] First, the differences between the LSGH-Net and CNN models proposed in this invention are compared. The structure of CNN is consistent with the proposed network structure. The difference is that the Gaussian heat map regression used to generate smooth samples is not applied to CNN. At the same time, the effects of the improved strategies (such as scSE) adopted in the proposed model are also compared. The results of the comparative experiment are shown in Table 1.

[0080] Table 1 Comparison of indicators of three models

[0081]

[0082] In terms of overall accuracy, the model proposed in the embodiment of the present invention and the improved strategy using Gaussian heat map regression performed better than the traditional CNN (93.67%). LSGH-Net outperformed CNN in all indicators, especially sensitivity (improved by about 3%). The FI score is an indicator that reflects the overall performance of a deep learning model. The score of CNN (93.34%) is lower than that of LSGH-Net (95.13%), indicating that the proposed method has better predictive performance in landslide susceptibility mapping. Compared with LSGH-Net, removing the scSE strategy will cause some changes in the model, namely higher specificity (99.23%) and lower sensitivity (89.28%). The implementation of the scSE strategy improved the overall accuracy of the model (1.04%) and F1 score (1.17%).

[0083] At the same time, if Figure 5 As shown in the figure, the comparison of PR curves and AP values ​​of different models is shown. If the PR curve of classifier A completely contains another classifier B, then A performs better than B. Through the analysis and comparison of AP values, the regression performance of the three models has achieved excellent results, among which LSGH-Net is the best and CNN is relatively poor.

[0084] The present invention is to use Python software packages Keras and Tensorflow as the backend for landslide susceptibility mapping. The generated results are imported into ArcGIS 10.6, and the susceptibility index is divided into 5 categories from low to high using the Jenks natural break point method. It is worth noting that since the embodiment of the present invention mainly studies the influence of smooth samples generated by Gaussian heat map regression on landslide susceptibility. Therefore, although the improvement strategies such as the scSE module improve the model performance, the difference in the mapping results is not obvious. And they have similar susceptibility index frequency distribution. Therefore, the present invention only compares the mapping results of the optimized LSGH-Net and the traditional CNN.

[0085] As shown in Figure 6(a) and Figure 6(b), the spatial distribution of landslides in the two mapping results is similar. The landslides are mainly concentrated in the northwest and southeast of the earthquake zone. From the perspective of landslide susceptibility index, there are obvious differences between the two models. For CNN, high and extremely high landslide susceptibility account for about 52% (108.16 km2) of the total area. Pourghasemi et al. (2013) proposed that landslide susceptibility should be considered for the rationality of mapping results. That is, the areas where landslides have occurred should be located in high and extremely high susceptibility areas as much as possible. At the same time, the proportion of high susceptibility areas and extremely high susceptibility areas should be as small as possible, because the incidence of landslides is still small relative to the entire area. Compared with the results of CNN, the proportion of high susceptibility areas and extremely high susceptibility areas in LSGH-Net (about 33%) is more reasonable.

[0086] As shown in Figure 7(a) and Figure 7(b), the quantitative relationship between the index and the frequency can be more intuitively reflected through the frequency histogram of the landslide susceptibility index. For CNN, most of the susceptibility indexes are between 0 and 0.2 and 0.8 and 1.0. This phenomenon quantitatively explains why high-susceptibility areas and extremely high-susceptibility areas account for a large proportion. For LSGH-Net, the frequency distribution of the susceptibility index is more reasonable, that is, there is still a relatively large frequency distribution in the medium-susceptibility area. Therefore, the Gaussian heat map regression method can improve the frequency distribution of the susceptibility index. This result can provide more precise mapping details and accurate prediction results.

[0087] The landslide susceptibility assessment device provided by the present invention is described below. The landslide susceptibility assessment device described below and the landslide susceptibility assessment method described above can be referred to each other. Figure 8 As shown, an embodiment of the present invention discloses a landslide susceptibility assessment device, comprising:

[0088] The acquisition module 801 is used to acquire the remote sensing image to be measured;

[0089] An evaluation module 802 is used to input the remote sensing image to be measured into a landslide susceptibility evaluation model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility evaluation model;

[0090] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0091] The landslide susceptibility assessment device of the embodiment of the present invention augments the landslide sample set through the Gaussian heat map of the landslide, adds smooth samples to optimize the distribution of samples, further improves the performance and efficiency of the assessment model, and obtains more accurate prediction results.

[0092] In at least one embodiment of the present invention, the evaluation module 802 obtains the landslide sample set based on the Gaussian heat map of the landslide, including:

[0093] A center point extraction module, used to obtain the landslide remote sensing image and extract the center point of the landslide;

[0094] A heat map generation module, which generates a Gaussian heat map of the landslide by center point using a two-dimensional Gaussian kernel function;

[0095] The augmentation module is used to obtain coordinate points corresponding to different landslide susceptibilities according to the Gaussian heat map of the landslide, and randomly extract the coordinate points corresponding to the different landslide susceptibilities to obtain a landslide sample set.

[0096] In at least one embodiment of the present invention, the step of randomly selecting coordinate points corresponding to different landslide susceptibilities to obtain a landslide sample set comprises:

[0097] A landslide sample set is obtained by randomly selecting corresponding coordinate points from each landslide susceptibility interval, wherein the different landslide susceptibility intervals include five intervals with equal intervals between 0 and 1 in terms of landslide susceptibility.

[0098] In at least one embodiment of the present invention, the step of generating the Gaussian heat map of the landslide point by center point using a two-dimensional Gaussian kernel function comprises:

[0099] For a single landslide center point, a Gaussian thermal map of the landslide is obtained by using a two-dimensional Gaussian kernel function to generate a circumscribed square with a side length of 2R;

[0100] For multiple landslide center points, a two-dimensional Gaussian kernel function is used to generate a circle with a radius R at each center point to obtain a Gaussian thermal map of the landslide;

[0101] The landslide susceptibility of the center point is 1, the landslide susceptibility of the circle or square boundary is 0.1, and the landslide susceptibility outside the boundary is 0.

[0102] In at least one embodiment of the present invention, the types of landslide susceptibility influencing factors marked on the landslide sample set include: slope, curvature, slope aspect, slope position, distance to water system, distance to road, lithology, NDVI and land utilization rate.

[0103] In at least one embodiment of the present invention, the landslide susceptibility assessment model includes a convolutional layer, a scSE module, a maximum pooling layer, a dropout layer, a Flatten layer and a fully connected layer connected in sequence.

[0104] Fig. 9 An example of a physical structure diagram of an electronic device is shown in FIG. Fig. 9 As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930 and a communication bus 940, wherein the processor 910, the communication interface 920 and the memory 930 communicate with each other through the communication bus 940. The processor 910 may call the logic instructions in the memory 930 to execute a landslide susceptibility assessment method, which includes:

[0105] Acquire the remote sensing image to be measured;

[0106] Inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model;

[0107] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0108] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0109] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the landslide susceptibility assessment method provided by the above methods, the method comprising:

[0110] Acquire the remote sensing image to be measured;

[0111] Inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model;

[0112] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0113] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the landslide susceptibility assessment method provided by the above methods, the method comprising:

[0114] Acquire the remote sensing image to be measured;

[0115] Inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model;

[0116] The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide.

[0117] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0118] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A landslide susceptibility assessment method, characterized in that: include: Acquire the remote sensing image to be measured; Inputting the remote sensing image to be measured into a landslide susceptibility assessment model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility assessment model; The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide; The landslide sample set is obtained based on the Gaussian heat map of the landslide, including: Acquire the landslide remote sensing image and extract the center point of the landslide; A two-dimensional Gaussian kernel function is used to generate a Gaussian thermal map of the landslide at each center point; Obtaining coordinate points corresponding to different landslide susceptibility according to the Gaussian heat map of the landslide; Randomly extracting coordinate points corresponding to the different landslide susceptibility to obtain a landslide sample set; The method of generating the Gaussian thermal map of the landslide by center point by using a two-dimensional Gaussian kernel function comprises: For a single landslide center point, a two-dimensional Gaussian kernel function is used to generate a circle with a radius of R at each center point to obtain the Gaussian thermal map of the landslide; For multiple landslide center points, a two-dimensional Gaussian kernel function is used to generate a circle with a radius R for each center point, and after eliminating the overlapping areas of different landslide center points, a Gaussian thermal map of the landslide is obtained; The landslide susceptibility of the center point is 1, the landslide susceptibility of the boundary of the circle is 0.1, and the landslide susceptibility outside the boundary is 0.

2. The landslide susceptibility assessment method according to claim 1, characterized in that: The randomly selecting coordinate points corresponding to the different landslide susceptibilities to obtain a landslide sample set includes: The corresponding coordinate points are randomly selected from each landslide susceptibility interval to obtain a landslide sample set, wherein the landslide susceptibility interval includes five intervals with equal intervals between 0 and 1 in landslide susceptibility.

3. The landslide susceptibility assessment method according to claim 1, characterized in that: The types of landslide susceptibility influencing factors marked on the landslide sample set include: slope, curvature, slope aspect, slope position, distance to water system, distance to road, lithology, NDVI and land utilization rate.

4. The landslide susceptibility assessment method according to claim 1, characterized in that: The landslide susceptibility assessment model includes a convolutional layer, a scSE module, a maximum pooling layer, a dropout layer, a Flatten layer and a fully connected layer which are connected in sequence.

5. A landslide susceptibility assessment device, characterized in that: include: An acquisition module is used to obtain the remote sensing image to be measured; An evaluation module, used for inputting the remote sensing image to be measured into a landslide susceptibility evaluation model to obtain a landslide susceptibility score of the remote sensing image to be measured output by the landslide susceptibility evaluation model; The landslide susceptibility assessment model is obtained by training based on a landslide sample set and landslide susceptibility influencing factors marked on the landslide sample set, and the landslide sample set is obtained based on a Gaussian heat map of the landslide; The evaluation module comprises: A center point extraction module, used to obtain the landslide remote sensing image and extract the center point of the landslide; A heat map generation module, used for generating a Gaussian heat map of the landslide by center point using a two-dimensional Gaussian kernel function; An augmentation module, for obtaining coordinate points corresponding to different landslide susceptibilities according to the Gaussian heat map of the landslide; and randomly extracting coordinate points corresponding to the different landslide susceptibilities to obtain a landslide sample set; The heat map generation module is specifically used to obtain the Gaussian heat map of the landslide by using a two-dimensional Gaussian kernel function to generate a circle with a radius of R from each center point for a single landslide center point; for multiple landslide center points, a two-dimensional Gaussian kernel function is used to generate a circle with a radius of R from each center point, and the overlapping areas of different landslide center points are eliminated to obtain the Gaussian heat map of the landslide; wherein the landslide susceptibility of the center point is 1, the landslide susceptibility value of the boundary of the circle is 0.1, and the landslide susceptibility outside the boundary is 0.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the landslide susceptibility assessment method according to any one of claims 1 to 4 are implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the landslide susceptibility assessment method according to any one of claims 1 to 4 are implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the landslide susceptibility assessment method according to any one of claims 1 to 4 are implemented.

Images