Remote sensing data reconstruction method and device and computer equipment

Abstract

The invention relates to a remote sensing data reconstruction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring optical remote sensing data of crops, and generating an NDVI time sequence repair data set according to an optical data missing rule; a periodic interpolation generative adversarial network is called to reconstruct the NDVI time sequence repair data set, and a high-time-resolution NDVI time sequence is obtained; and carrying out image data visualization processing on the NDVI time sequence with the high time resolution to obtain a reconstructed NDVI image time sequence. Through restoration, reconstruction and visualization processing of the optical remote sensing data, the accuracy and practicability of crop growth monitoring are improved.

Description

Technical Field

[0001] The present application relates to the field of optical remote sensing technology, and in particular to a remote sensing data reconstruction method, device, computer equipment and storage medium. Background Art

[0002] With the rapid development of remote sensing technology and data processing technology, remote sensing data acquisition and processing technology has gradually become the core of research in various fields, especially in agriculture, environmental monitoring and disaster warning. The Normalized Difference Vegetation Index (NDVI) is a common vegetation index based on multispectral remote sensing data, which is widely used in vegetation coverage monitoring, drought assessment and ecosystem research. NDVI time series data obtained at different time points can be used to track the dynamic changes of vegetation.

[0003] Affected by satellite rotation period, cloud cover and sensor failure, NDVI data often has large-scale regularity missing and data noise, which significantly affects the accuracy of subsequent analysis tasks. Existing NDVI time series data reconstruction methods are mainly divided into the following three categories: traditional filtering algorithms, machine learning methods and multi-source data fusion methods. However, existing methods still have shortcomings in complex nonlinear relationship modeling and high-resolution time series reconstruction, making it difficult to accurately capture rapidly changing vegetation characteristics. Summary of the invention

[0004] Based on this, it is necessary to provide a remote sensing data reconstruction method, device, computer equipment and storage medium that can combine targeted network design and optimization strategies to significantly improve the reconstruction quality and temporal resolution of NDVI time series data in response to the above technical problems.

[0005] In a first aspect, the present application provides a remote sensing data reconstruction method. The method comprises:

[0006] Obtain optical remote sensing data of crops and generate NDVI time series repair dataset based on the missing rules of optical data;

[0007] The periodic interpolation generative adversarial network is called to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series.

[0008] The image data of the high temporal resolution NDVI time series is visualized to obtain the reconstructed NDVI image time series.

[0009] In one embodiment, the periodic interpolation generative adversarial network includes:

[0010] Acquire optical remote sensing data of historical crops, and obtain optical data and SAR data in the optical remote sensing data of historical crops;

[0011] The input features of optical data and SAR data are screened according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features;

[0012] According to the distribution of missing optical data, the training data set and the test data set are generated by combining low-correlation input features and high-importance input features;

[0013] Generate a periodic interpolation generative adversarial network using a total loss function based on the training and test datasets.

[0014] In one embodiment, the input features of optical data and SAR data are screened according to the Pearson coefficient method and the random forest method, and the low-correlation input features and high-importance input features are obtained, including:

[0015] Based on the dual-stream neural network, features of optical data and SAR data are extracted simultaneously. During the extraction process, the dynamic adjustment and optimization of the model are improved by squeezing and exciting one-dimensional convolution modules and channel-by-channel linear layers.

[0016] In one embodiment, generating a training data set and a test data set based on the distribution of optical data missing patterns and combining low-correlation input features and high-importance input features includes:

[0017] The missing rate of optical remote sensing data is obtained based on the optical remote sensing data of historical crops;

[0018] According to the optical data missing rate, an artificial mask is constructed; the artificial mask is used to mark the optical data missing area and remove the available optical data area;

[0019] According to the correspondence between different types of optical data missing areas and low-correlation input features and high-importance input features under the regular distribution of optical data missing, training data sets and test data sets are constructed.

[0020] In one embodiment, the method further comprises:

[0021] The high temporal resolution NDVI time series reconstructed by periodic interpolation generative adversarial network is discriminated through void convolution; the authenticity of local areas with fixed intervals in fixed-interval optical remote sensing data is judged, and the data missing areas in the regular distribution of missing optical data are ignored.

[0022] In one embodiment, obtaining optical remote sensing data of crops includes:

[0023] Obtain crop spatial coordinates based on land classification dataset;

[0024] The optical remote sensing data under the spatial coordinates of the crops are retrieved according to the spatial coordinates of the crops.

[0025] In one embodiment, the total loss function includes:

[0026] The sum of the basic loss function, the adversarial loss function, and the consistent loss function is used to constrain the periodic interpolation generative adversarial network.

[0027] In a second aspect, the present application also provides a remote sensing data reconstruction device. The device comprises:

[0028] The dataset module is used to obtain optical remote sensing data of crops and generate NDVI time series repair datasets based on the distribution of missing optical data;

[0029] The reconstruction module is used to call the periodic interpolation generative adversarial network to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series;

[0030] The visualization module is used to perform image data visualization processing on the high temporal resolution NDVI time series to obtain the reconstructed NDVI image time series.

[0031] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0032] Obtain optical remote sensing data of crops and generate NDVI time series repair dataset based on the missing rules of optical data;

[0033] The periodic interpolation generative adversarial network is called to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series.

[0034] The image data of the high temporal resolution NDVI time series is visualized to obtain the reconstructed NDVI image time series.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0036] Obtain optical remote sensing data of crops and generate NDVI time series repair dataset based on the missing rules of optical data;

[0037] The periodic interpolation generative adversarial network is called to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series.

[0038] The image data of the high temporal resolution NDVI time series is visualized to obtain the reconstructed NDVI image time series.

[0039] The above remote sensing data reconstruction method, device, computer equipment and storage medium obtain optical remote sensing data of crops and generate NDVI time series repair data sets according to the optical data missing law, so as to effectively identify and repair the data missing areas, ensure the integrity of the data, and avoid the inaccurate impact caused by data missing; call the time series data set repair model to reconstruct the NDVI time series repair data set to obtain a high temporal resolution NDVI time series. The reconstruction process improves the temporal resolution, ensures that the changes in crop growth can be captured accurately, and provides more accurate information for agricultural management. Finally, the image data visualization processing is performed on the reconstructed high temporal resolution NDVI time series to generate a reconstructed NDVI image time series, so that users can intuitively and clearly observe the dynamic changes in crop growth, which is convenient for data analysis and decision support. The accuracy and practicality of crop growth monitoring are improved by repairing, reconstructing and visualizing optical remote sensing data. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A diagram showing an application environment of a remote sensing data reconstruction method in an embodiment;

[0041] Figure 2 A schematic diagram of a process of remote sensing data reconstruction method in one embodiment;

[0042] Figure 3 is a correlation analysis diagram between different feature sensitivity indicators in one embodiment;

[0043] Figure 4 A comparison diagram of NDVI data before and after restoration in an embodiment;

[0044] Figure 5 A schematic diagram of a flow chart of a remote sensing data reconstruction method in another embodiment;

[0045] Figure 6 is a structural block diagram of a remote sensing data reconstruction device in one embodiment;

[0046] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0048] The remote sensing data reconstruction method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0049] In one embodiment, Figure 2 As shown, a remote sensing data reconstruction method is provided, which is applied to Figure 1 The server in the example is used to illustrate the following steps:

[0050] Step 202, obtaining optical remote sensing data of crops, and generating an NDVI time series repair data set according to the optical data missing law.

[0051] Among them, the optical remote sensing data of crops is ground information data about crops obtained by optical remote sensing technology, usually including multi-band data such as visible light and infrared bands, which can provide information about crop growth conditions, land use conditions, etc. The law of missing optical data is that in the process of remote sensing data collection, optical data may be missing due to weather (such as cloud cover) and equipment problems. The law of missing optical data refers to the spatial and temporal distribution of missing data. NDVI (Normalized Difference Vegetation Index) is a remote sensing index used to measure ground vegetation coverage. It is calculated based on the red and near-infrared band values ​​of remote sensing images; NDVI time series refers to a continuous sequence of NDVI values ​​collected in chronological order within a certain time range.

[0052] Step 204, calling the periodic interpolation generative adversarial network to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series.

[0053] Among them, the high-temporal-resolution NDVI time series refers to continuous time series data generated through data repair and reconstruction technology, which can accurately reflect the growth cycle and status of crops.

[0054] Step 206 , performing image data visualization processing on the high temporal resolution NDVI time series to obtain a reconstructed NDVI image time series.

[0055] Among them, image data visualization processing is the process of converting time series data into visual images, which are usually presented in the form of charts, dynamic images, etc., which helps researchers analyze data more intuitively and discover patterns and trends.

[0056] Exemplarily, optical remote sensing data of crop areas are collected by remote sensing equipment (such as satellites), and the optical data include NDVI values ​​(vegetation index) at various time points during the growth of crops. Due to weather factors or equipment problems, some optical data will be missing. After acquiring the data, a preliminary NDVI time series repair data set is generated according to the law of data missing (such as missing patterns of temporal and spatial distribution). The NDVI time series repair data set is reconstructed using a trained periodic interpolation generative adversarial network (PIGAN model), and missing data is predicted and filled to obtain a high-temporal resolution NDVI time series that can more accurately describe the growth status of crops at different time points. The repaired high-temporal resolution NDVI time series is converted into an image form to display a dynamic image of the NDVI change process or a static chart of the time series data, so that users can intuitively observe the growth change trend of crops and the vegetation conditions at each time node.

[0057] In the above remote sensing data reconstruction method, by obtaining the optical remote sensing data of crops and generating the NDVI time series repair data set according to the optical data missing law, the data missing area is effectively identified and repaired to ensure the integrity of the data and avoid the inaccuracy caused by the missing data; the time series data set repair model is called to reconstruct the NDVI time series repair data set to obtain a high temporal resolution NDVI time series. The reconstruction process improves the temporal resolution and ensures that the changes in crop growth can be captured accurately, providing more accurate information for agricultural management. Finally, the image data visualization processing is performed on the reconstructed high temporal resolution NDVI time series to generate a reconstructed NDVI image time series, so that users can intuitively and clearly observe the dynamic changes in crop growth, which is convenient for data analysis and decision support. The accuracy and practicality of crop growth monitoring are improved by repairing, reconstructing and visualizing the optical remote sensing data.

[0058] In one embodiment, a periodic interpolation generative adversarial network includes: acquiring optical remote sensing data of historical crops to obtain optical data and SAR data in the historical optical remote sensing data of crops; screening input features of the optical data and SAR data according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features; generating a training data set and a test data set based on the distribution of missing optical data and combining low-correlation input features and high-importance input features; generating a periodic interpolation generative adversarial network using a total loss function based on the training data set and the test data set.

[0059] Among them, in remote sensing data, optical data and synthetic aperture radar (SAR) data are the two most commonly used data types. Optical remote sensing data can provide high-resolution surface images, which are suitable for the identification and classification of ground objects; SAR data can provide surface information under various weather conditions, and has unique advantages for monitoring surface deformation, humidity, etc. The Pearson coefficient method is a statistical method used to measure the linear correlation between two variables. By calculating the Pearson correlation coefficient, the strength and direction of the relationship between two variables can be evaluated. The random forest method is an ensemble learning method that solves regression and classification problems by constructing multiple decision trees.

[0060] For example, a total of 24 optical vegetation sensitivity indicators and 7 SAR vegetation sensitivity indicators were obtained through literature review. The input features of optical data and SAR data were screened by the Pearson coefficient method, and the features with strong correlation were eliminated to reduce redundant features; the indicators were screened based on the Pearson correlation coefficient, with 0.9 as the threshold, and 12 low-correlation optical vegetation sensitivity indicators and 4 low-correlation SAR vegetation sensitivity indicators were obtained, such as Figure 3 The importance of the remaining features was evaluated using the random forest algorithm, and the input features with high importance for remote sensing data reconstruction were selected; the low-correlation indicators were screened using the importance analysis of the random forest, and 8 low-correlation and high-importance optical sensitivity indicators were obtained: B1, B3, NDVI, CVI, RVI, NGRDI and CI, and 2 low-correlation and high-importance SAR sensitivity indicators: VV and VH.

[0061] According to the distribution of optical data missing patterns, the training data set and the test data set are generated by combining low-correlation input features and high-importance input features; the periodic interpolation generative adversarial network using the total loss function is generated based on the training data set and the test data set. The total loss function includes: the sum of the basic loss function, the adversarial loss function and the consistent loss function, which is used to constrain the periodic interpolation generative adversarial network. The expressions of the basic loss function, the adversarial loss function and the consistent loss function are shown in the following formulas:

[0062]

[0063] L Gen =L base +λ 3 L Gan +λ 4 L CL

[0064] Where, L base is the basic loss function, L Gan To combat the loss function, L CL is a consistent loss function, L Gen is the overall loss function; M fake To simulate artificial mask, M sim Mask the missing data of the entire input time series; The output of the periodic interpolation generative adversarial network and the original monitored discrete NDVI data are respectively the data after interpolation and replacement of the output of the periodic interpolation generative adversarial network and the discrete data; F Dis ,F Gen They refer to the discriminator function and generator function of the periodic interpolation generative adversarial network respectively; X S1 ,X S2 ,T Gen represent SAR and optical satellite data without artificial deletion, T Gen Represents the intermediate indicator variable for the generator discriminative loss calculation, which is calculated by dilated convolution with a fixed dilation rate. sim , and then save the average value or the larger point value at each data point to get; TV(·), Represent the total variation value of the calculated data and the bi-norm of the data respectively; λ 1 ,λ 2 ,λ 3 ,λ 4 ,λ 5 is the balance coefficient of the loss function.

[0065] The periodic interpolation generative adversarial network (PIGAN model) is tested on the test set to compare with the other six commonly used time series analysis methods. The overall performance results of different methods on the test set are shown in the following table:

[0066] Methods MAE RMSE <![CDATA[R 2 ]]> SG filter 0.0590 0.124 0.704 HANTS 0.0724 0.128 0.686 RNN 0.0579 0.091 0.841 LSTM 0.0583 0.092 0.839 Dlinear 0.0744 0.111 0.765 TimesNet 0.0620 0.092 0.780 Ours 0.0514 0.081 0.873

[0067] The performance results of different methods under different data missingness on the test set are shown in the following table:

[0068]

[0069]

[0070] Through the training and testing of the periodic interpolation generative adversarial network, compared with other methods, the periodic interpolation generative adversarial network performed well in various evaluation indicators, especially in the RMSE* indicator, which was improved by about 0.04 compared with the traditional filtering algorithm, equivalent to about 30% of the original performance indicator. Experiments show that this method is better than other methods in different optical data missing situations, and shows excellent performance in NDVI image reconstruction tasks, and can be widely used in downstream tasks.

[0071] Furthermore, the periodic interpolation generative adversarial network also includes discriminating the high-temporal-resolution NDVI time series reconstructed by the periodic interpolation generative adversarial network through dilated convolution; judging the authenticity of the fixed-interval local area in the fixed-interval optical remote sensing data, and ignoring the data missing area in the regular distribution of the optical data missing. That is, based on the dilated convolution technology, the traditional task of judging whether the entire reconstruction result is correct or not is transformed into the task of judging whether the fixed-interval samples of the reconstructed remote sensing data are correct or not, which adapts to the characteristics of large-scale regular missing of remote sensing data.

[0072] In one embodiment, after the input features of optical data and SAR data are screened according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features, the method includes: simultaneously extracting features of the optical data and SAR data based on a two-stream neural network; during the extraction process, improving the dynamic adjustment and optimization of the model through squeezing and excitation one-dimensional convolution modules and channel-by-channel linear layers.

[0073] Among them, the dual-stream neural network is a neural network architecture that processes multiple input data at the same time. In remote sensing data processing, the dual-stream neural network is usually used to process optical data and synthetic aperture radar (SAR) data at the same time. Each stream (or branch) is dedicated to feature extraction for one data type and the extracted features are fused. Specifically, the optical data stream processes image data from optical sensors, which usually contains visible light and near-infrared information; the SAR data stream processes image data from synthetic aperture radars and provides spatial information based on radar signal reflections.

[0074] Specifically, a two-stream neural network structure is used to extract and process optical and SAR data respectively. At the same time, when extracting high-dimensional features of the data, a squeeze and excitation one-dimensional convolution module (SE-Conv) and a channel-by-channel linear layer are used to improve the model's understanding of local and abstract information.

[0075] In one embodiment, generating a training data set and a test data set based on a regular distribution of missing optical data in combination with low-correlation input features and high-importance input features includes: obtaining a missing rate of optical remote sensing data based on historical optical remote sensing data of crops; constructing an artificial mask based on the optical data missing rate; the artificial mask is used to mark areas where optical data is missing and remove areas where available optical data is available; and constructing a training data set and a test data set based on the correspondence between different types of optical data missing areas and low-correlation input features and high-importance input features under a regular distribution of missing optical data.

[0076] Specifically, the optical remote sensing data set of historical crops is obtained, and the missing rate of each pixel is calculated by analyzing the observations at each time point and each spatial position. That is, for each time point, statistics are made on whether the observation value of the point is valid, so as to obtain a missing rate matrix, which reflects the missing situation of optical data at different time and spatial positions. The value of the missing rate matrix can be obtained by statistical analysis, and the proportion of missing data is the missing rate of optical data at that position. Based on the missing rate of optical data, an artificial mask is further constructed, which is used to mark the missing areas of optical data and remove the valid optical data in these areas. In the process of constructing the mask, for the optical data at each moment, the missing area is marked as 1 and the non-missing area is marked as 0. In the subsequent data processing, it is ensured that the model training will not be disturbed by the missing data, and the data quality and accuracy during the training process are ensured.

[0077] Furthermore, for different degrees of optical data missing and different crops, the corresponding low-correlation input features and high-importance input features are obtained. Among them, low-correlation input features refer to features with a low correlation with any optical data missing situation, and high-importance input features refer to features that have a greater impact on the optical data reconstruction results. These features will perform differently in different missing areas. Therefore, by analyzing the correspondence between the optical data missing law and these features, guidance can be provided for constructing training data sets and test data sets. On this basis, these features are used to construct training data sets and test data sets that conform to the distribution of optical data missing laws, ensuring that the data sets contain enough representative samples, while avoiding the impact of data imbalance or inconsistency on the training effect of the model.

[0078] In this embodiment, the diversity and accuracy of the training data set and the test data set are effectively enhanced, so that the model can better adapt to the situation of missing optical data. During the training process, the model can fully learn the characteristics and laws of the missing area, thereby improving the accuracy and robustness of reconstruction, and can more effectively handle the problem of missing optical data, thereby improving the performance and generalization ability of the remote sensing data reconstruction model.

[0079] In one embodiment, obtaining optical remote sensing data of crops includes: obtaining spatial coordinates of crops according to a land classification data set; and retrieving optical remote sensing data at the spatial coordinates of crops according to the spatial coordinates of crops.

[0080] Among them, the land classification dataset is a land cover classification dataset generated based on remote sensing images and geographic information system (GIS) technology, which provides detailed information on land use and land cover type, and is used to determine the spatial location of crops. The crop spatial coordinates are obtained by identifying the crop distribution area using the land classification dataset, and are used to locate the specific spatial location of crops.

[0081] Specifically, the crop spatial coordinates of seven representative food crops, namely corn, cotton, rice, sorghum, soybean, spring wheat and winter wheat, are obtained using the Cropland Data Layer (CDL), and the satellite optical remote sensing data and SAR data under the crop spatial coordinates are obtained.

[0082] In this embodiment, by obtaining the spatial coordinates of crops based on the land classification data set and retrieving the corresponding optical remote sensing data and SAR data, the crop distribution area can be accurately located to ensure that the extracted remote sensing data is consistent with the actual distribution of crops. This method makes full use of the rich information of land classification data and reduces the analysis errors caused by data mismatch.

[0083] In one embodiment, Figure 4 and Figure 5 As shown, a remote sensing data reconstruction method is provided, comprising the following steps:

[0084] Step 502, obtaining crop spatial coordinates according to the land classification data set.

[0085] Step 504, retrieve the optical remote sensing data and SAR data under the crop spatial coordinates according to the crop spatial coordinates.

[0086] Step 506: Generate an NDVI time series repair dataset based on the optical data missing rule.

[0087] Step 508, acquiring historical optical remote sensing data of crops, and obtaining optical data and SAR data in the historical optical remote sensing data of crops.

[0088] Step 510 , screen the input features of the optical data and the SAR data according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features.

[0089] Step 512, obtaining the missing rate of optical remote sensing data based on historical optical remote sensing data of crops.

[0090] Step 514: construct an artificial mask according to the optical data missing rate; the artificial mask is used to mark the optical data missing area and remove the available optical data area.

[0091] Step 516: construct a training data set and a test data set according to the correspondence between different types of optical data missing regions and low-correlation input features and high-importance input features under the optical data missing regular distribution.

[0092] Step 518: construct an artificial mask according to the optical data missing rate; the artificial mask is used to mark the optical data missing area and remove the available optical data area.

[0093] Step 520 , constructing a training data set and a test data set according to the correspondence between different types of optical data missing regions and low-correlation input features and high-importance input features under the optical data missing regular distribution.

[0094] In step 522, feature extraction is performed on the optical data and SAR data simultaneously based on the dual-stream neural network; during the extraction process, the model is dynamically adjusted and optimized by squeezing and exciting one-dimensional convolution modules and channel-by-channel linear layers. In step 524, a periodic interpolation generative adversarial network using a total loss function is generated based on the training data set and the test data set.

[0095] Step 526, calling the periodic interpolation generative adversarial network to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series.

[0096] Step 528, the high temporal resolution NDVI time series reconstructed by the periodic interpolation generative adversarial network is judged through the void convolution; the authenticity of the local area of ​​the fixed interval in the fixed interval optical remote sensing data is judged, and the data missing area in the regular distribution of the optical data missing is ignored.

[0097] Step 530 , performing image data visualization processing on the high temporal resolution NDVI time series to obtain a reconstructed NDVI image time series.

[0098] In this embodiment, optical remote sensing data of crop areas are collected by remote sensing equipment (such as satellites), and the optical data include NDVI values ​​(vegetation index) at various time points during the growth of crops. Due to weather factors or equipment problems, some optical data may be missing. After acquiring the data, a preliminary NDVI time series repair data set is generated according to the law of data missing (such as missing patterns of temporal and spatial distribution). The NDVI time series repair data set is reconstructed using a trained periodic interpolation generative adversarial network to predict and fill in missing data to obtain a high-temporal resolution NDVI time series that can more accurately describe the growth status of crops at different time points. The repaired high-temporal resolution NDVI time series is converted into an image form to display a dynamic image of the NDVI change process or a static chart of the time series data, so that users can intuitively observe the growth change trend of crops and the vegetation conditions at each time node.

[0099] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are shown in sequence, these steps are not necessarily executed in sequence in the order indicated. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the steps or stages in other steps.

[0100] Based on the same inventive concept, the embodiment of the present application also provides a remote sensing data reconstruction device for implementing the remote sensing data reconstruction method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more remote sensing data reconstruction device embodiments provided below can refer to the limitations of the remote sensing data reconstruction method above, and will not be repeated here.

[0101] In one embodiment, Figure 6 As shown, a remote sensing data reconstruction device is provided, including: a data set module 602, a reconstruction module 604 and a visualization module 606, wherein:

[0102] The data set module 602 is used to obtain optical remote sensing data of crops and generate an NDVI time series repair data set according to the distribution pattern of missing optical data.

[0103] The reconstruction module 604 is used to call the periodic interpolation generative adversarial network to reconstruct the NDVI time series repair data set to obtain a high temporal resolution NDVI time series.

[0104] The visualization module 606 is used to perform image data visualization processing on the NDVI time series with high temporal resolution to obtain a reconstructed NDVI image time series.

[0105] In one embodiment, the apparatus further comprises:

[0106] The acquisition module is used to acquire the optical remote sensing data of historical crops, and obtain the optical data and SAR data in the optical remote sensing data of historical crops.

[0107] The feature screening module is used to screen the input features of optical data and SAR data according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features.

[0108] The training set module is used to generate training data sets and test data sets according to the distribution of optical data missing patterns and combining low-correlation input features and high-importance input features.

[0109] A training module is used to generate a periodic interpolation generative adversarial network using a total loss function based on a training dataset and a test dataset.

[0110] In one embodiment, it further includes:

[0111] The generation module is used to simultaneously extract features from optical data and SAR data based on a two-stream neural network; during the extraction process, the dynamic adjustment and optimization of the model is improved through squeezing and excitation of one-dimensional convolution modules and channel-by-channel linear layers.

[0112] In one embodiment, the training set module includes:

[0113] The missing module is used to obtain the missing rate of optical remote sensing data based on historical optical remote sensing data of crops.

[0114] The mask module is used to construct an artificial mask according to the optical data missing rate; the artificial mask is used to mark the optical data missing area and remove the available optical data area.

[0115] The construction module is used to construct a training data set and a test data set according to the correspondence between different types of optical data missing areas and low-correlation input features and high-importance input features under the optical data missing law distribution.

[0116] In one embodiment, it further includes:

[0117] The discrimination module is used to discriminate the high-temporal-resolution NDVI time series reconstructed by the periodic interpolation generative adversarial network through void convolution; judge the authenticity of the local areas of fixed intervals in the fixed-interval optical remote sensing data, and ignore the data missing areas in the regular distribution of optical data missing.

[0118] In one embodiment, the data set module 402 further includes:

[0119] The coordinate module is used to obtain the spatial coordinates of crops based on the land classification dataset.

[0120] The calling function is used to retrieve the optical remote sensing data under the crop space coordinates according to the crop space coordinates.

[0121] In one embodiment, it further includes:

[0122] The loss module is used for the sum of the basic loss function, the adversarial loss function and the consistent loss function, which is used to constrain the periodic interpolation generative adversarial network.

[0123] Each module in the remote sensing data reconstruction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0124] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a remote sensing data reconstruction method is implemented.

[0125] Those skilled in the art will understand that Figure 7The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0126] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0127] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0128] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0129] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0130] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0131] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A remote sensing data reconstruction method, characterized in that: The method comprises: Obtain optical remote sensing data of crops and generate NDVI time series repair dataset based on the missing rules of optical data; Calling a periodic interpolation generative adversarial network to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series; Image data visualization processing is performed on the high temporal resolution NDVI time series to obtain a reconstructed NDVI image time series.

2. The method according to claim 1, characterized in that The periodic interpolation generative adversarial network includes: Acquire optical remote sensing data of historical crops, and obtain optical data and SAR data in the optical remote sensing data of historical crops; The input features of the optical data and the SAR data are screened according to the Pearson coefficient method and the random forest method to obtain low-correlation input features and high-importance input features; Generate a training data set and a test data set based on the distribution of optical data missing rules and combining the low-correlation input features and the high-importance input features; A periodic interpolation generative adversarial network using a total loss function is generated according to the training data set and the test data set.

3. The method according to claim 2, characterized in that The input features of the optical data and the SAR data are screened according to the Pearson coefficient method and the random forest method, and the low-correlation input features and high-importance input features are obtained, including: Based on a dual-stream neural network, feature extraction is performed simultaneously on the optical data and the SAR data; during the extraction process, the dynamic adjustment and optimization of the model is improved by squeezing and exciting one-dimensional convolution modules and channel-by-channel linear layers.

4. The method according to claim 2, characterized in that: The step of generating a training data set and a test data set based on the optical data missing regular distribution and combining the low-correlation input features and the high-importance input features comprises: Obtaining a missing rate of optical remote sensing data according to the optical remote sensing data of the historical crops; According to the optical data missing rate, an artificial mask is constructed; the artificial mask is used to mark the optical data missing area and remove the available optical data area; According to the correspondence between different types of optical data missing regions under the optical data missing regular distribution and the low-correlation input features and the high-importance input features, a training data set and a test data set are constructed.

5. The method according to claim 1, characterized in that The method further comprises: The high temporal resolution NDVI time series reconstructed by the periodic interpolation generative adversarial network is discriminated through void convolution; the authenticity of the local areas of fixed intervals in the fixed-interval optical remote sensing data is judged, and the data missing areas in the regular distribution of optical data missing are ignored.

6. The method according to claim 1, characterized in that The obtaining of optical remote sensing data of crops comprises: Obtain crop spatial coordinates based on land classification dataset; The optical remote sensing data under the crop spatial coordinates is retrieved according to the crop spatial coordinates.

7. The method according to claim 2, characterized in that The total loss function includes: The sum of the basic loss function, the adversarial loss function and the consistent loss function is used to constrain the periodic interpolation generative adversarial network.

8. A remote sensing data reconstruction device, characterized in that: The device comprises: The dataset module is used to obtain optical remote sensing data of crops and generate NDVI time series repair datasets based on the distribution of missing optical data; A reconstruction module is used to call a periodic interpolation generative adversarial network to reconstruct the NDVI time series repair dataset to obtain a high temporal resolution NDVI time series; The visualization module is used to perform image data visualization processing on the high temporal resolution NDVI time series to obtain a reconstructed NDVI image time series.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A 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 method according to any one of claims 1 to 7 are implemented.

Images