Method and device for identifying risks caused by human activities to natural reserve

Abstract

The invention discloses a method and a device for identifying risks caused by human activities to a natural reserve. The method comprises the following steps: selecting a plurality of grids from all grids corresponding to the natural reserve as to-be-checked grids based on a remote sensing image; determining population time sequence data of each to-be-checked grid based on the mobile phone signaling data corresponding to each to-be-checked grid, determining all human activity types corresponding to each to-be-checked grid based on the population time sequence data of each to-be-checked grid, and determining development trends of various human activity types corresponding to each to-be-checked grid; and determining that the reason that the land coverage type change pattern spot area of each to-be-checked grid exceeds a preset threshold value is a natural factor or a human factor. The method is high in accuracy and efficiency.

Description

Technical Field

[0001] This application belongs to the technical field of ecological protection, and in particular relates to a method and device for identifying risks caused by human activities to nature reserves. Background Art

[0002] Nature reserves play an indispensable role in protecting key ecosystems and biodiversity. These reserves are crucial for maintaining ecological balance. However, since they are usually located in remote areas with complex and variable terrain and inconvenient transportation, this poses challenges to on-site manual verification and makes it difficult for traditional supervision methods to accurately identify and lock in ecological risk clues, thus adding many difficulties to ecological risk management.

[0003] Currently, the monitoring of nature reserves mainly relies on satellite remote sensing technology to monitor the dynamic changes of land cover types. This method can provide high-precision, wide-coverage and short-cycle monitoring data, which helps to understand the ecological conditions within the reserve. However, there are certain timeliness issues with satellite remote sensing images. It takes a certain amount of time from image acquisition to preprocessing and analysis. More importantly, although satellite remote sensing can reveal the changes that have occurred, it is not sensitive to potential risks caused by human activities and cannot distinguish whether the land cover changes are caused by human factors or natural causes. This means that relying solely on satellite remote sensing for monitoring may lead to untimely and incomplete identification of risks brought by human activities, increasing the difficulty and workload of on-site verification.

[0004] Therefore, there is an urgent need to develop a more efficient method for identifying human activity risks to supplement existing monitoring means and improve the effectiveness and response speed of nature reserve management. Summary of the Invention

[0005] This application provides a method and device for identifying risks caused by human activities to nature reserves to solve the above problems or at least partially solve the above problems.

[0006] In a first aspect, a method for identifying risks caused by human activities to nature reserves is disclosed. The method includes:

[0007] Step S1: Select a number of grids from all the grids corresponding to the nature reserve based on remote sensing images as grids to be checked. All the grids to be checked are grids where the area of the land cover type change patch exceeds a preset threshold;

[0008] Step S2: Determine the population time series data of each grid to be checked based on the mobile phone signaling data corresponding to each grid to be checked. Determine all human activity types corresponding to each grid to be checked based on the population time series data of each grid to be checked, and determine the development trend of various human activity types corresponding to each grid to be checked;

[0009] Step S3: Based on the development trends of various human activity types corresponding to each grid to be inspected, determine whether the reason for the area of the land cover type change patches in each grid to be inspected exceeding the preset threshold is a natural factor or a human factor, and generate target verification task information for each grid to be inspected. The target verification task information includes the verification frequency, verification time, and verification area.

[0010] If it is a human factor, allocate the target verification task to the target user, and the target user executes the target verification task.

[0011] Preferably, in step S1, several grids are selected from all the grids corresponding to the nature reserve based on the remote sensing image. The grids to be inspected are all grids with the area of the land cover type change patches exceeding the preset threshold, including:

[0012] Step S11: Divide the nature reserve into grids according to the vector boundary of the nature reserve; regard the grids occupied by the nature reserve and the nature reserve boundary as all the grids corresponding to the nature reserve; obtain multi-phase sub-meter satellite remote sensing images corresponding to the nature reserve through the satellite.

[0013] Step S12: Select several grids from all the grids corresponding to the nature reserve as the grids to be inspected. The grids to be inspected are all grids with the area of the land cover type change patches exceeding the preset threshold; among them, the land cover type change includes vegetation reduction, structure increase, and road construction.

[0014] Preferably, preprocess the remote sensing image. The preprocessing includes completing the preprocessing of the remote sensing image according to the process of orthorectification, radiometric calibration, and Flash atmospheric correction.

[0015] Preferably, use U-Net and Deeplab V3+ to identify the increase of structures, use the U-Net model to identify road construction, and use the maximum likelihood method, artificial neural network method, and random forest method to identify vegetation reduction.

[0016] Preferably, in step S2: Determine the population time series data of each grid to be inspected based on the mobile phone signaling data corresponding to each grid to be inspected, determine all human activity types corresponding to each grid to be inspected based on the population time series data of each grid to be inspected, and determine the development trends of various human activity types corresponding to each grid to be inspected, including:

[0017] Step S21: Determine the sampling time period of the mobile phone signaling data corresponding to each grid to be inspected, sample the mobile phone signaling data within the sampling time period at a frequency of once per hour, and arrange the sampled data in chronological order to obtain the population time series data of each grid to be inspected.

[0018] Step S22: Determine all human activity types corresponding to each grid to be checked based on the population time series data of each grid to be checked, and determine the population quantity corresponding to each human activity type of each grid to be checked. The human activity types include residence, work, and visit.

[0019] Step S23: Use the Theil-Sen median to analyze the development trends of various human activity types corresponding to each grid to be checked.

[0020] Preferably, in step S3: Based on the development trends of various human activity types corresponding to each grid to be checked, determine whether the reason for the area of the patch of land cover type change in each grid to be checked exceeding the preset threshold is a natural factor or a human factor, including:

[0021] For each grid to be checked, extract the image features of the patch of land cover type change of the grid to be checked; obtain the historical patch data with the same development trend as various human activity types corresponding to the grid to be checked, and extract the historical patch features from the historical patch data; determine whether the reason for the area of the patch of land cover type change in the grid to be checked exceeding the preset threshold is a natural factor or a human factor based on the annotation result of the historical patch data; the annotation result of the historical patch data is a natural factor or a human factor.

[0022] In a second aspect, a device for identifying the risks caused by human activities to a nature reserve is disclosed. The device includes:

[0023] Initialization module: Configured to select several grids from all grids corresponding to the nature reserve as grids to be checked based on remote sensing images. The grids to be checked are all grids with the area of the patch of land cover type change exceeding the preset threshold.

[0024] Analysis module: Configured to determine the population time series data of each grid to be checked based on the mobile signaling data corresponding to each grid to be checked, determine all human activity types corresponding to each grid to be checked based on the population time series data of each grid to be checked, and determine the development trends of various human activity types corresponding to each grid to be checked.

[0025] Cause analysis module: Configured to determine whether the reason for the area of the patch of land cover type change in each grid to be checked exceeding the preset threshold is a natural factor or a human factor based on the development trends of various human activity types corresponding to each grid to be checked, and generate target verification task information for each grid to be checked. The target verification task information includes verification frequency, verification time, and verification area.

[0026] If it is a human factor, allocate the target verification task to the target user, and the target user executes the target verification task.

[0027] In a third aspect, an electronic device is disclosed, the electronic device comprising:

[0028] at least one processor; and

[0029] a memory communicatively connected to the at least one processor; wherein,

[0030] the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method as described above.

[0031] In a fourth aspect, a non-transitory computer-readable storage medium storing computer instructions is disclosed, the computer instructions being used to cause the computer to execute the method as described above.

[0032] This application collects mobile signaling data within the scope of the nature reserve, identifies risks for areas where the land cover type change in the nature reserve exceeds a preset threshold based on the grid-based mobile signaling data, determines the reasons for the areas where the land cover type change exceeds the preset threshold, and determines whether the change in the land cover type of this area is caused by human activities. It solves the problems of low accuracy and low efficiency in determining the reasons for the areas where the land cover type change exceeds the preset threshold by only using satellite remote sensing monitoring.

[0033] This application has the following technical effects: It solves the problems of insufficient timeliness in only using satellite remote sensing monitoring and inability to determine whether it is caused by natural or human factors, can detect potential threats that may affect the reserve earlier, and accurately distinguish the reasons for these changes, so as to take appropriate protection measures to ensure that the functions of the nature reserve are fully maintained. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a schematic flowchart of a method for identifying risks caused by human activities to a nature reserve;

[0035] Figure 2 is a schematic structural diagram of a device for identifying risks caused by human activities to a nature reserve. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0037] As Figure 1 shown, the present application provides a method for identifying risks caused by human activities to a nature reserve, the method comprising:

[0038] Step S1: Select several grids from all the grids corresponding to the nature reserve based on the remote sensing image as the grids to be inspected. All the grids to be inspected are grids where the area of the land cover type change patches exceeds a preset threshold;

[0039] Step S2: Determine the population time series data of each grid to be inspected based on the mobile signaling data corresponding to each grid to be inspected. Determine all the human activity types corresponding to each grid to be inspected based on the population time series data of each grid to be inspected, and determine the development trend of various human activity types corresponding to each grid to be inspected;

[0040] Step S3: Based on the development trend of various human activity types corresponding to each grid to be inspected, determine whether the reason for the area of the land cover type change patches in each grid to be inspected exceeding the preset threshold is a natural factor or a human factor, and generate target verification task information for each grid to be inspected. The target verification task information includes the verification frequency, verification time, and verification area;

[0041] If it is a human factor, allocate the target verification task to the target user, and the target user executes the target verification task.

[0042] The step S1, which selects several grids from all the grids corresponding to the nature reserve based on the remote sensing image as the grids to be inspected, and all the grids to be inspected are grids where the area of the land cover type change patches exceeds a preset threshold, includes:

[0043] Step S11: Divide the nature reserve into grids according to the vector boundary of the nature reserve; regard the grids occupied by the nature reserve and the nature reserve boundary as all the grids corresponding to the nature reserve; obtain multi-phase sub-meter satellite remote sensing images corresponding to the nature reserve through the satellite;

[0044] Step S12: Select several grids from all the grids corresponding to the nature reserve as the grids to be inspected. All the grids to be inspected are grids where the area of the land cover type change patches exceeds a preset threshold; among them, the land cover type change includes vegetation reduction, structure increase, and road construction.

[0045] Furthermore, preprocess the remote sensing image. The preprocessing includes completing the preprocessing of the remote sensing image according to the process of orthorectification, radiometric calibration, and Flash atmospheric correction.

[0046] In this application, preprocessing is used to correct the radiation error and geometric distortion of the image. The NNDiffuse PanSharpening model can be used for image fusion. By combining the high-resolution panchromatic band image with the multi-band image, the spatial resolution can be improved, and the color difference of the fused result image can be weakened. Statistical operations related to all bands input to the algorithm will also be performed, which can not only well retain the spectral information and texture features of the multi-spectral image, but also achieve the best effect of the fused result.

[0047] In this application, nature reserves are divided into core areas, buffer zones and experimental areas. The risk sources of human activities in nature reserves include vegetation reduction, increase in structures, and road construction. Further, it can be classified into mine resource development, industrial development, energy development, tourism development, transportation development, aquaculture development, agricultural development, residential areas and other activities.

[0048] The grid to be searched is determined by using deep learning. For example, U-Net and Deeplab V3+ are used to identify the increase in structures, the U-Net model is used to identify road construction, and the maximum likelihood method, artificial neural network method and random forest method are used to identify vegetation reduction. Taking the increase in structures as an example, sample selection is carried out first. The training set includes data from multiple remote sensing change detection data sets (LEVIR-CD, DSIFN, etc.). LEVIR-CD is a large-scale remote sensing building binary change detection data set, consisting of 637 pairs of very high-resolution images, with a size of 1024*1024 pixels. These bi-temporal images with a time span of 5-14 years have significant changes in building distribution, including building growth and decline. LEVIR-CD covers various types of buildings, such as villa houses, high-rise apartments, small garages and large warehouses. The DSIFN data set consists of six large bi-temporal high-resolution images. The large image pairs are cropped into 394 small image patches with a size of 512*512. After data augmentation, a set of 3940 bi-temporal image pairs is obtained.

[0049] In the test set, the main preprocessing of the data is to fill in the default values. Because there are often pixel points in the two-phase images of the same location, where one-phase image is default and the other is normal, then a false positive example will be output for change detection here. Therefore, a Mask is made to process the default values, that is, if one of the two-phase images is default for a certain pixel, the pixel values of both two-phase images at this pixel are set to zero, that is, no change detection is performed on it.

[0050] The data this time includes five-phase building distribution images of a certain place collected by two satellites in different quarters, including different perspectives, scales, illuminations, and quarters, which also brings some difficulties to the change detection of buildings. The size of the original remote sensing images is large. Each original remote sensing image is cropped into image patches of 512*512 size to facilitate subsequent sample annotation and model training. Traverse all the change detection results output. Since there are relatively few image pairs with TP output, all the image patches with TP output are found and observed whether they have obvious building changes. Finally, 50 image pairs are selected for annotation. The main annotation tool is labelme integrated with SAM (segment anything).

[0051] Secondly, the model is selected. Deep learning models are usually used for remote sensing image change detection. Among them, the encoder-decoder structure is one of the common architectures. This structure aims to extract features from the input remote sensing image and output a mask image that identifies the changed area.

[0052] Common encoder-decoder architectures are basically implemented based on fully convolutional networks, including U-Net series, Deeplab series, PSPNet, STANet, etc. In this application, the classic U-Net and Deeplab V3+ are mainly selected as the basic architectures.

[0053] The Encoder is based on the ResNet network. Finally, ResNet34 is selected as the encoder for change detection and is initialized through pre-training on the ImageNet dataset.

[0054] Finally, model training and accuracy verification are carried out. Due to reasons such as the shooting time, location, and angle of remote sensing images, compared with natural images, the ground objects in remote sensing images show characteristics such as small inter-class differences, large intra-class differences, and large illumination changes, which greatly increases the difficulty of semantic segmentation in remote sensing images. According to the task and data distribution characteristics, this application uses a variety of data augmentation methods to expand the data and increase the diversity of the data to reduce the risk of network overfitting. The data augmentation methods include enhancement methods based on the information of the image itself such as random flipping, random brightness adjustment, random HSV, random noise, and random scale cropping.

[0055] The U-Net and Deeplab V3+ were respectively used as the infrastructure to identify the structures in this area, and the test set was used to verify the model accuracy. The accuracies of the two models are shown in the following table. Among them, the U-Net model has a higher recall rate and can basically detect all changes in buildings, but its robustness is poor and it is sensitive to cloud cover and terrain changes; the Deeplab V3+ model has the best comprehensive performance and high accuracy in identifying large-scale demolitions. Therefore, the Deeplab V3+ model was finally selected as the structure identification model.

[0056] Table 1 Comparison of recognition accuracies of different models

[0057]

[0058] In step S2: Based on the mobile signaling data corresponding to each grid to be queried, determine the population time series data of each grid to be queried, and based on the population time series data of each grid to be queried, determine all human activity types corresponding to each grid to be queried, and determine the development trends of various human activity types corresponding to each grid to be queried, including:

[0059] Step S21: Determine the sampling time period of the mobile signaling data corresponding to each grid to be queried, sample the mobile signaling data within the sampling time period at a frequency of once per hour, and arrange the sampled data in chronological order to obtain the population time series data of each grid to be queried;

[0060] Step S22: Based on the population time series data of each grid to be queried, determine all human activity types corresponding to each grid to be queried, and determine the population quantity corresponding to each human activity type of each grid to be queried. The human activity types include residence, work, and visit;

[0061] Step S23: Use the Theil-Sen median to analyze the development trends of various human activity types corresponding to each grid to be queried.

[0062] In this application, the population time series data of each grid to be queried is statistically analyzed. For example, based on the Two-step, k-means clustering algorithm + decision tree combination algorithm, the classification of resident population and working population is obtained, and other signaling data are classified as visiting population.

[0063] Another example is the identification of the working population and the resident population, including:

[0064] 1) Set the residence time from 21:00 to 05:00 (early morning of the next day).

[0065] 2) Set the working time from 06:00 to 20:00 (weekday).

[0066] 3) Locate the area to which the user belongs based on the location information where the communication behavior occurs.

[0067] 4) Obtain the classification rules for resident population and working population based on the Two-step, k-means clustering algorithm + decision tree combination algorithm, and classify other signaling data as visiting population.

[0068] 5) If a user satisfies multiple places of residence, use the location with the most months of stay as the place of residence.

[0069] 6) If a user satisfies multiple workplaces, use the location with the most months of stay as the workplace.

[0070] 7) Obtain the total population, working population, resident population, and visiting population of each grid to be queried during the sampling time period; the sampling time period is, for example, one month.

[0071] The Theil-Sen median trend calculates the median of the slopes of n(n - 1) / 2 data combinations, and its calculation formula is as follows:

[0072]

[0073] 1 ≤ i < j ≤ 31

[0074] In the formula, Median represents taking the median value. If S x > 0, it indicates that the ecological parameter x is in an increasing trend, otherwise it is in a decreasing trend.

[0075] Classify it as significant change and non-significant change according to the significance test results of the MK test at the 0.05 confidence level, and further classify the significant change as slightly significant, generally significant, and extremely significant changes according to whether the trend passes the 90%, 95%, and 99% significance tests, so as to obtain the change trends of the total population, visiting population, working population, and resident population of each grid one by one.

[0076] In step S3: Based on the development trends of various human activity types corresponding to each grid to be queried, determine whether the reason for the area of the land cover type change patch of each grid to be queried exceeding the preset threshold is a natural factor or a human factor, including:

[0077] For each grid to be queried, extract the image features of the land cover type change patch of this grid to be queried; obtain the historical patch data with the same development trend as various human activity types corresponding to this grid to be queried, and extract the historical patch features from the historical patch data; determine whether the reason for the area of the land cover type change patch of this grid to be queried exceeding the preset threshold is a natural factor or a human factor based on the annotation results of the historical patch data; the annotation results of the historical patch data are natural factors or human factors.

[0078] Furthermore, the target user conducts on-site verification of the grids to be inspected in the nature reserve according to the target verification task, and conducts corresponding supervision and rectification according to the on-site verification results.

[0079] This application combines satellite remote sensing data and mobile phone signaling data, and improves the efficiency and accuracy of identifying human activity risks in the ecological protection red line area through multi-source data fusion and analysis.

[0080] This application can sort out the grids to be inspected on-site, and conduct data distribution, on-site verification and result feedback through methods such as emails and supervision platforms. The specific information of the map patches includes: map patch number, district where it is located, township, category, area, remote sensing image before change, remote sensing image after change, longitude coordinate of the center point of the verified map patch, latitude coordinate of the center point of the verified map patch, and situation description. For on-site verification, it is necessary to take field photos and the reasons for the changes in the human activity risk map patches.

[0081] For example, through remote sensing image analysis, it is found that there is a change in land cover from bare land to vegetation cover in the buffer zone of A1 Nature Reserve in 2022. The results calculated by mobile phone signaling show that Grid 335712 in the buffer zone of A1 Nature Reserve in that year is a grid for development and construction activities. There was frequent human activity in this grid from May to December of that year, and the number of visiting population and working population increased significantly, and the change rule of the working population conforms to the early warning characteristics of development and construction activities. Since 2023, there has been no working population in this grid, but there are still visiting populations. It is thus judged that the grid mainly started concentrated development and construction activities in May 2022. On-site verification found that there was ecological restoration activity on the lower slope of the highway here.

[0082] Another example is that according to remote sensing images, a certain point is judged to be for residents and other activities, and there are obvious changes in land cover between 2022 and 2023. Combining the calendar maps of the visiting population grids in 2022 and 2023, it is judged that there are sporadic human activities in this grid during the two years, and the mobile phone signaling data does not meet the requirements of tourism activities and development and construction activities. It is judged that the change in land cover is caused by natural factors, and this human activity risk point does not need to be distributed.

[0083] As Figure 2 shown, this application provides a device for identifying the risks caused by human activities to nature reserves, and the device includes:

[0084] Initialization module: configured to select several grids from all the grids corresponding to the nature reserve based on remote sensing images as the grids to be inspected, and all the grids to be inspected are grids where the area of the land cover type change map patch exceeds a preset threshold;

[0085] Analysis module: Configured to determine the population time series data of each grid to be checked based on the mobile signaling data corresponding to each grid to be checked, determine all human activity types corresponding to each grid to be checked based on the population time series data of each grid to be checked, and determine the development trends of various human activity types corresponding to each grid to be checked;

[0086] Cause analysis module: Configured to determine whether the reason for the area of the land cover type change patch of each grid to be checked exceeding the preset threshold is a natural factor or a human factor based on the development trends of various human activity types corresponding to each grid to be checked, and generate target verification task information for each grid to be checked. The target verification task information includes verification frequency, verification time, and verification area;

[0087] If it is a human factor, allocate the target verification task to the target user, and the target user executes the target verification task.

[0088] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for identifying the risks caused by human activities to nature reserves, characterized in that, The method includes: Step S1: Select several grids from all the grids corresponding to the nature reserve based on the remote sensing image as the grids to be inspected. The grids to be inspected are all grids where the area of the patch with land cover type change exceeds a preset threshold; Step S2: Determine the population time series data of each grid to be inspected based on the mobile phone signaling data corresponding to each grid to be inspected. Determine all the human activity types corresponding to each grid to be inspected based on the population time series data of each grid to be inspected, and determine the development trend of various human activity types corresponding to each grid to be inspected; Step S3: Based on the development trend of various human activity types corresponding to each grid to be inspected, determine whether the reason for the area of the patch with land cover type change exceeding the preset threshold in each grid to be inspected is a natural factor or a human factor, and generate target verification task information for each grid to be inspected. The target verification task information includes the verification frequency, verification time, and verification area; If it is a human factor, allocate the target verification task to the target user, and the target user executes the target verification task.

2. The method according to claim 1, wherein The step S1, which selects several grids from all the grids corresponding to the nature reserve based on the remote sensing image as the grids to be inspected, and the grids to be inspected are all grids where the area of the patch with land cover type change exceeds a preset threshold, includes: Step S11: Divide the nature reserve into grids according to the vector boundary of the nature reserve; regard the grids occupied by the nature reserve and the nature reserve boundary as all the grids corresponding to the nature reserve; obtain multi-phase sub-meter satellite remote sensing images corresponding to the nature reserve through satellites; Step S12: Select several grids from all the grids corresponding to the nature reserve as the grids to be inspected. The grids to be inspected are all grids where the area of the patch with land cover type change exceeds a preset threshold; among them, the land cover type change includes vegetation reduction, construction increase, and road construction.

3. The method according to claim 1, characterized in that, Perform preprocessing on the remote sensing image. The preprocessing includes completing the preprocessing of the remote sensing image according to the processes of orthorectification, radiometric calibration, and Flash atmospheric correction.

4. The method according to claim 2, wherein Use U-Net and Deeplab V3+ to identify the increase in structures, use the U-Net model to identify road construction, and use the maximum likelihood method, artificial neural network method, and random forest method to identify vegetation reduction.

5. The method according to claim 1, characterized in that The step S2: Determine the population time series data of each grid to be inspected based on the mobile phone signaling data corresponding to each grid to be inspected. Determine all the human activity types corresponding to each grid to be inspected based on the population time series data of each grid to be inspected, and determine the development trend of various human activity types corresponding to each grid to be inspected, includes: Step S21: Determine the sampling time period of the mobile phone signaling data corresponding to each grid to be inspected, sample the mobile phone signaling data within the sampling time period at a frequency of once per hour, and arrange the sampled data in chronological order to obtain the population time series data of each grid to be inspected; Step S22: Determine all the human activity types corresponding to each grid to be inspected based on the population time series data of each grid to be inspected, and determine the population quantity corresponding to each human activity type of each grid to be inspected. The human activity types include residence, work, and visit; Step S23: Use Theil - Sen median trend analysis to analyze the development trends of various types of human activities corresponding to each grid to be checked.

6. The method according to claim 1, wherein The said Step S3: Based on the development trends of various types of human activities corresponding to each grid to be checked, determine whether the reason for the area of the land cover type change patches in each grid to be checked exceeding the preset threshold is a natural factor or a human factor, including: For each grid to be checked, extract the image features of the land cover type change patches of this grid to be checked; obtain the historical patch data with the same development trend as various types of human activities corresponding to this grid to be checked, and extract the historical patch features from the historical patch data; based on the annotation results of the historical patch data, determine whether the reason for the area of the land cover type change patches in this grid to be checked exceeding the preset threshold is a natural factor or a human factor; the annotation results of the historical patch data are natural factors or human factors.

7. An identification device for the risks caused by human activities to nature reserves, characterized in that, The said device includes: Initialization module: Configured to select several grids from all the grids corresponding to the nature reserve based on remote sensing images as the grids to be checked, and the grids to be checked are all grids where the area of the land cover type change patches exceeds the preset threshold; Analysis module: Configured to determine the population time - series data of each grid to be checked based on the mobile phone signaling data corresponding to each grid to be checked, determine all types of human activities corresponding to each grid to be checked based on the population time - series data of each grid to be checked, and determine the development trends of various types of human activities corresponding to each grid to be checked; Cause analysis module: Configured to determine whether the reason for the area of the land cover type change patches in each grid to be checked exceeding the preset threshold is a natural factor or a human factor based on the development trends of various types of human activities corresponding to each grid to be checked, and generate target verification task information for each grid to be checked, and the target verification task information includes verification frequency, verification time and verification area; If it is a human factor, allocate the said target verification task to the target user, and the target user executes the said target verification task.

8. A computer - readable storage medium, in which multiple instructions are stored; the multiple instructions are used to be loaded and executed by a processor to perform the method according to any one of claims 1 - 6.

9. An electronic device, characterized in that, The said electronic device includes: A processor, used to execute multiple instructions; A memory, used to store multiple instructions; Among them, the multiple instructions are used to be stored by the memory and loaded and executed by the processor to perform the method according to any one of claims 1 - 6.

Images