Hetao irrigation area autumn irrigation water remote sensing monitoring method based on space-time fusion algorithm

Abstract

The invention relates to the technical field of agricultural water monitoring, in particular to a Hetao irrigation area autumn irrigation water remote sensing monitoring method based on a space-time fusion algorithm, and the method comprises the steps: obtaining remote sensing data of a to-be-classified Hetao irrigation area in an autumn irrigation period; the remote sensing data of the to-be-classified Hetao irrigation district is input to a preset autumn irrigation classification model, an irrigation farmland image data set of the to-be-classified Hetao irrigation district is obtained, and the preset autumn irrigation classification model is obtained through training of a high-temporal-spatial-resolution image data set generated by original remote sensing data of a target Hetao irrigation district in the autumn irrigation period. Therefore, the problems that time and labor are consumed, comprehensive and accurate data cannot be provided and timeliness is poor in the process of monitoring autumn watering in the Hetao irrigation area through a traditional ground survey method are solved, and accurate and timely monitoring of the autumn watering range, area and progress is achieved through the space-time fusion algorithm and the remote sensing technology based on deep learning.

Description

Technical Field

[0001] This application relates to the technical field of agricultural water monitoring, and particularly relates to a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm. Background Art

[0002] The Inner Mongolia Hetao Irrigation District is located in Bayannur City, Inner Mongolia Autonomous Region. It is a large irrigation district in the upper reaches of the Yellow River and an important national high-quality commercial grain and oil production base. With the implementation of relevant regulations, the rigid constraints on water resources have been continuously strengthened. The available water volume of the Yellow River in the Hetao Irrigation District has decreased, and in-depth water conservation is imminent.

[0003] Autumn irrigation is a non-growing season irrigation system summarized by the people in the Hetao area through long-term production practice for the purpose of leaching salts and preserving soil moisture. The irrigation time is from September to November every year, which is an important guarantee for the sustainable development of the Hetao Irrigation District. However, the water consumption for autumn irrigation is large, accounting for about 1 / 3 of the total annual water consumption. There are also rough water use methods such as flood irrigation and unreasonable repeated irrigation. In recent years, the average gross water consumption per mu for autumn irrigation is greater than the experimental recommended value of about 120 m 3 / mu, and there is a large water-saving space. Monitoring the autumn irrigation range, area and progress in the irrigation district is an important measure to boost in-depth water conservation in the Hetao Irrigation District. Under the background of the continuous strengthening of the rigid constraints on water resources in the irrigation district, it is necessary to accurately identify the autumn irrigation process in the Hetao Irrigation District in order to solve the problems of "flood irrigation" and water-saving and efficiency-increasing, clarify the unnecessary irrigation water consumption in the Hetao Irrigation District, and realize the optimization of the autumn irrigation water quota and the spatio-temporal allocation pattern. However, the current traditional ground survey methods are time-consuming and laborious, cannot provide comprehensive and accurate data, and also have deficiencies in timeliness, which need to be solved urgently. Summary of the Invention

[0004] This application provides a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm to solve the problems of time-consuming, laborious, inability to provide comprehensive and accurate data, and poor timeliness in the traditional ground survey method for monitoring the autumn irrigation process in the Hetao Irrigation District. Through the use of a spatio-temporal fusion algorithm and remote sensing technology based on deep learning, accurate and timely monitoring of the autumn irrigation range, area and progress has been achieved.

[0005] The first aspect embodiment of this application provides a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm, including the following steps:

[0006] Obtain remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period;

[0007] Input the remote sensing data of the Hetao Irrigation District to be classified into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation District to be classified, where the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period.

[0008] According to an embodiment of the present application, before inputting the remote sensing data of the Hetao Irrigation Area to be classified into the preset autumn irrigation classification model, it further includes:

[0009] Obtain the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period, and preprocess the original remote sensing data of the target Hetao Irrigation Area to obtain preprocessed remote sensing data;

[0010] Extract the preprocessed remote sensing data based on a preset extraction strategy to obtain a preliminary extraction result of the autumn irrigation area;

[0011] Fuse the original remote sensing data based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm to obtain the high spatio-temporal resolution image dataset;

[0012] Train a preset neural network based on the high spatio-temporal resolution image dataset to obtain an initial autumn irrigation classification model, verify the initial autumn irrigation classification model based on multi-source field verification data, and when the initial autumn irrigation classification model meets the preset verification conditions, use the initial autumn irrigation classification model as the preset autumn irrigation classification model.

[0013] According to an embodiment of the present application, the extracting the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area includes:

[0014] Calculate the preprocessed remote sensing data using a preset multi-band water body index calculation formula to obtain a water body index calculation result;

[0015] Extract the water body areas where the water body index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result;

[0016] Perform a spatial overlay analysis on the remote sensing image extraction result to obtain the preliminary extraction result of the autumn irrigation area.

[0017] According to an embodiment of the present application, the multi-source field verification data includes at least one of field investigation data, field sampling data, and unmanned aerial vehicle aerial photography data.

[0018] According to an embodiment of the present application, the preset verification conditions include:

[0019] The accuracy rate of the initial autumn irrigation classification model is greater than a preset accuracy rate, the precision of the initial autumn irrigation classification model is greater than a preset precision, the recall rate of the initial autumn irrigation classification model is greater than a preset recall rate, and the F1 score of the initial autumn irrigation classification model is greater than a preset score.

[0020] The remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to the embodiments of the present application inputs the remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified. Thus, the problems of the traditional ground survey method in monitoring the autumn irrigation process in the Hetao Irrigation Area, such as time-consuming, laborious, inability to provide comprehensive and accurate data, and poor timeliness, are solved, and the accurate and timely monitoring of the autumn irrigation range, area and progress is realized by using the spatio-temporal fusion algorithm and the remote sensing technology based on deep learning.

[0021] The second aspect of the embodiments of the present application provides a remote sensing monitoring device for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm, including:

[0022] An acquisition module, configured to acquire remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period;

[0023] An autumn irrigation classification module, configured to input the remote sensing data of the Hetao Irrigation Area to be classified into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified, wherein the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period.

[0024] According to an embodiment of the present application, before inputting the remote sensing data of the Hetao Irrigation Area to be classified into the preset autumn irrigation classification model, the autumn irrigation classification module is further configured to:

[0025] Acquire the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period, and perform preprocessing on the original remote sensing data of the target Hetao Irrigation Area to obtain preprocessed remote sensing data;

[0026] Based on a preset extraction strategy, extract the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area;

[0027] Fuse the original remote sensing data based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm to obtain the high spatio-temporal resolution image dataset;

[0028] Train a preset neural network based on the high spatio-temporal resolution image dataset to obtain an initial autumn irrigation classification model, verify the initial autumn irrigation classification model based on multi-source field verification data, and when the initial autumn irrigation classification model meets the preset verification conditions, use the initial autumn irrigation classification model as the preset autumn irrigation classification model.

[0029] According to an embodiment of the present application, the autumn irrigation classification model is configured to:

[0030] Calculate the preprocessed remote sensing data using a preset multi - band water body index calculation formula to obtain a water body index calculation result;

[0031] Extract the water body areas where the water body index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result;

[0032] Perform a spatial overlay analysis on the remote sensing image extraction result to obtain a preliminary extraction result of the autumn irrigation area.

[0033] According to an embodiment of the present application, the multi - source field verification data includes at least one of field investigation data, field sampling data, and unmanned aerial vehicle aerial photography data.

[0034] According to an embodiment of the present application, the preset verification conditions include:

[0035] The accuracy rate of the initial autumn irrigation classification model is greater than a preset accuracy rate, the precision of the initial autumn irrigation classification model is greater than a preset precision, the recall rate of the initial autumn irrigation classification model is greater than a preset recall rate, and the F1 score of the initial autumn irrigation classification model is greater than a preset score.

[0036] According to the remote sensing monitoring device for autumn irrigation in the Hetao Irrigation Area based on the spatio - temporal fusion algorithm in the embodiments of the present application, input the remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified. Thus, the problems of the traditional ground survey method in monitoring the autumn irrigation process in the Hetao Irrigation Area, such as time - consuming, laborious, unable to provide comprehensive and accurate data, and poor timeliness, are solved. The accurate and timely monitoring of the autumn irrigation scope, area, and progress is realized by adopting the spatio - temporal fusion algorithm and remote sensing technology based on deep learning.

[0037] An embodiment of the third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio - temporal fusion algorithm as described in the above embodiments.

[0038] An embodiment of the fourth aspect of the present application provides a computer - readable storage medium, on which a computer program is stored. The program is executed by a processor to be used to implement the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio - temporal fusion algorithm as described in the above embodiments.

[0039] The additional aspects and advantages of the present application will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present application. Description of the Drawings

[0040] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the accompanying drawings, where:

[0041] Figure 1 FIG. 4 is a flowchart of a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm according to an embodiment of the present application;

[0042] Figure 2 FIG. 8 is a schematic diagram of the investigation situation in the Hetao Irrigation District according to an embodiment of the present application;

[0043] Figure 3 FIG. 12 is a schematic diagram of the soil texture distribution map in the Hetao Irrigation District according to an embodiment of the present application;

[0044] Figure 4 FIG. 16 is a distribution map of the main crop planting structure in the Hetao Irrigation District from 2019 to 2022 according to an embodiment of the present application;

[0045] Figure 5 FIG. 20 is a grading map of the soil salinization degree in the Hetao Irrigation District in 2020 according to an embodiment of the present application;

[0046] Figure 6 FIG. 24 is a technical route map for extracting the autumn irrigation area according to an embodiment of the present application;

[0047] Figure 7 FIG. 28 is a map of the autumn irrigation range of the MODIS image in 2023 according to an embodiment of the present application;

[0048] Figure 8 FIG. 32 is a map of the autumn irrigation range of the GF-1 image in 2023 according to an embodiment of the present application;

[0049] Figure 9 FIG. 36 is a map of the autumn irrigation range of the Sentinel-2 image in 2024 according to an embodiment of the present application;

[0050] Figure 10 FIG. 40 is the remote sensing extraction result of the autumn irrigation area by different methods according to an embodiment of the present application;

[0051] Figure 11 FIG. 44 is a technical route map of the spatio-temporal fusion algorithm according to an embodiment of the present application;

[0052] Figure 12 FIG. 48 is a spatio-temporal fusion data set every 5 days from October 10, 2023 to November 15, 2024 according to an embodiment of the present application;

[0053] Figure 13 FIG. 52 is a spatio-temporal fusion data set every 5 days from October 10, 2024 to November 15, 2024 according to an embodiment of the present application;

[0054] Figure 14 Schematic diagram of the structure of the MLP model according to an embodiment of the present application;

[0055] Figure 15 Schematic diagram of the automatic extraction process of autumn irrigation information of remote sensing images based on deep learning according to an embodiment of the present application;

[0056] Figure 16 Schematic diagram of the progress of the autumn irrigation area in the Hetao Irrigation District every 5 days according to an embodiment of the present application;

[0057] Figure 17 Schematic diagram of the survey points during the autumn irrigation period in 2024 according to an embodiment of the present application;

[0058] Figure 18 Schematic diagram of the receiver operating characteristic curve according to an embodiment of the present application;

[0059] Figure 19 Schematic diagram of the precision-recall curve according to an embodiment of the present application;

[0060] Figure 20 Schematic diagram of the confusion matrix according to an embodiment of the present application;

[0061] Figure 21 Autumn irrigation area map of the Hetao Irrigation District in 2024 according to an embodiment of the present application;

[0062] Figure 22 Schematic block diagram of the remote sensing monitoring device for autumn irrigation water in the Hetao Irrigation District based on the spatio-temporal fusion algorithm according to an embodiment of the present application;

[0063] Figure 23 Schematic diagram of the structure of the electronic device according to an embodiment of the present application. Detailed implementation manners

[0064] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present application, and should not be construed as a limitation to the present application.

[0065] The following describes a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm with reference to the accompanying drawings. Aiming at the problems of time-consuming and laborious in monitoring the autumn irrigation process in the Hetao Irrigation District, inability to provide comprehensive and accurate data, and poor timeliness in the above-mentioned background technology, remote sensing technology can monitor farmland on a large scale. Combining multi-source remote sensing, ground observations, and the latest remote sensing technology can achieve fine observations of the autumn irrigation process. Therefore, this application provides a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm. This method uses remote sensing technology to achieve real-time monitoring of the scope and progress of autumn irrigation water in the Hetao Irrigation District, providing basic support for reducing unnecessary autumn irrigation water in the irrigation district and achieving in-depth water conservation and control.

[0066] Specifically, Figure 1 It is a schematic flowchart of a remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm provided by an embodiment of this application.

[0067] As Figure 1 shown, the remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm includes the following steps:

[0068] In step S101, remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period is obtained.

[0069] Among them, the remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period refers to multi-source satellite remote sensing observation data of the Hetao Irrigation District to be classified during the period from September to November every year. For example, the remote sensing data of the embodiments of this application can include image data from Landsat, Sentinel, MODIS, and high-resolution satellite series, which are not specifically limited here.

[0070] Specifically, in order to understand the current situation of the Hetao Irrigation District, before obtaining the remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period, the embodiments of this application can collect basic data of the Hetao Irrigation District to be classified during the autumn irrigation period through means such as on-site investigations, data collection, field experiments, and analysis. The basic data can include the autumn irrigation time, autumn irrigation water consumption, and autumn irrigation area of the Hetao Irrigation District over the years, the vector distribution map of the irrigation district, the canal system distribution of the irrigation district, the soil texture of the Hetao Irrigation District, the change of the planting structure, the degree of soil salinization, and meteorological data, etc.

[0071] Exemplarily, taking the Hetao Irrigation District to be classified in Bayannur City, Inner Mongolia Autonomous Region as an example, the investigation situation collected through on-site investigation is as Figure 2 shown, the statistical table of autumn irrigation time of the Hetao Irrigation District and different irrigation areas obtained through data collection and analysis is shown in Table 1, and the distribution of soil texture in the Hetao Irrigation District is as Figure 3 shown, and the distribution of the main crop planting structure in the Hetao Irrigation District from 2019 to 2022 is as Figure 4As shown in the figure and the classification of soil salinization degree in Hetao Irrigation Area in 2020 is as follows Figure 5 as shown

[0072] Table 1

[0073]

[0074] Furthermore, through the investigation of basic data, the time characteristics and spatial characteristics of autumn irrigation in the to-be-classified Hetao Irrigation Area are mastered, and the remote sensing data of the autumn irrigation period in the to-be-classified Hetao Irrigation Area are obtained through a public platform

[0075] Exemplarily, the remote sensing data of the to-be-classified Hetao Irrigation Area during the autumn irrigation period in the embodiments of the present application can be the remote sensing data of the autumn irrigation periods in 2023 and 2024. The remote sensing data of the autumn irrigation period respectively include all remote sensing images from October to November in 2023 (as shown in Table 2) and all remote sensing images from October to November in 2024 (as shown in Table 3).

[0076] Table 2

[0077]

[0078]

[0079] Table 3

[0080]

[0081] Furthermore, in order to understand the effect of salt leaching and soil moisture conservation during autumn irrigation and to confirm whether there is autumn irrigation, soil sampling surveys were carried out before and after autumn irrigation and before and after spring sowing in 2023 and 2024. The survey locations were determined based on the characteristics of autumn irrigation water in the Hetao Irrigation Area and the characteristics of the underlying surface (soil type, salinization degree, spectral characteristics, etc.). The progress of autumn irrigation was mainly investigated along the main canals in the Hetao Irrigation Area, and the start and end times of autumn irrigation and irrigation water volume of different plots were recorded synchronously. At the same time, household visits were carried out for autumn irrigation to provide support for model verification. During the survey, based on Google Earth remote sensing images and consulting the local irrigation area management agency about the water body distribution in the Hetao Irrigation Area, the main canals, drainage ditches and lakes in the Hetao Irrigation Area were selected as target water bodies, and the length, width and area of the water bodies were measured using the GPS toolbox. Positioning, sampling and recording were carried out synchronously to provide a working basis for distinguishing water bodies and autumn irrigation plots

[0082] Further, in some embodiments, before inputting the remote sensing data of the Hetao Irrigation District to be classified into a preset autumn irrigation classification model, it further includes: obtaining the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period, and preprocessing the original remote sensing data of the target Hetao Irrigation District to obtain preprocessed remote sensing data; extracting the preprocessed remote sensing data based on a preset extraction strategy to obtain a preliminary extraction result of the autumn irrigation area; fusing the original remote sensing data based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm to obtain a high spatio-temporal resolution image dataset; training a preset neural network based on the high spatio-temporal resolution image dataset to obtain an initial autumn irrigation classification model, and verifying the initial autumn irrigation classification model based on multi-source field verification data, and when the initial autumn irrigation classification model meets the preset verification conditions, taking the initial autumn irrigation classification model as the preset autumn irrigation classification model.

[0083] Further, in some embodiments, extracting the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area includes: calculating the preprocessed remote sensing data using a preset multi-band water index calculation formula to obtain a water index calculation result; extracting the water body area where the water index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result; performing spatial overlay analysis on the remote sensing image extraction result to obtain a preliminary extraction result of the autumn irrigation area.

[0084] Wherein, the preset threshold can be a threshold preset by those skilled in the art, and those skilled in the art can also adjust the preset threshold up and down according to the obvious area of the lake, and no specific limitation is made here.

[0085] Specifically, the embodiments of the present application can obtain the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period through a public platform, and preprocess the original remote sensing data, such as smoothing and denoising, to obtain preprocessed remote sensing data.

[0086] Further, in order to determine the most suitable autumn irrigation area, it is necessary to screen representative water indices from the existing water body remote sensing indices, determine the appropriate water body thresholds for each index respectively for the screened water indices, and determine the autumn irrigation range by statistically analyzing the area data. Evaluate the response of common water indices to autumn irrigation based on measured points, and the evaluation criteria can be that the index increases significantly after autumn irrigation and the index is less sensitive to ice. Based on this, the embodiments of the present application initially select the multi-band water index MBWI as the water index. Among them, the calculation formula of the multi-band water index MBWI is:

[0087] MBWI = 2ρ Green - ρ Red - ρ NIR - ρ SWIR1 - ρ SWIR2

[0088] Among them, ρ Green is the reflectance in the green band, ρ Red is the reflectance in the red band, ρ NIR is the reflectance in the near-infrared band, ρ SWIR1 is the reflectance in short-wave infrared 1, ρ SWIR2 is the reflectance in short-wave infrared 2.

[0089] Furthermore, as Figure 6 shown, for each date's remotely sensed image after preprocessing, calculate the MBWI image, and extract the water body areas where MBWI is greater than 0, including the irrigated farmland in the irrigation area and all water bodies. That is, the extraction results include all water bodies and moist soils in the Hetao Irrigation Area, including non-cultivated areas such as lakes, swamps, and saline-alkali lands with high water content, etc., and use the cultivated land information in the land use data interpreted from remotely sensed images of the irrigation area for masking extraction to obtain the information of irrigated farmland. Then, conduct a spatial overlay analysis on the extraction results of remotely sensed images during the autumn irrigation period (October to November) to obtain the information of irrigated farmland and water bodies during the autumn irrigation period of each year.

[0090] Furthermore, as Figure 7 shown, taking the target Hetao Irrigation Area in Bayannur City, Inner Mongolia Autonomous Region as an example, select the difference-type water body index MBWI to conduct a preliminary analysis on the five-day progress of autumn irrigation in the Hetao Irrigation Area. Use the daily remotely sensed image MODIS MOD09GA remote sensing image data with a spatial resolution of 500m to obtain remotely sensed images every five days starting from October 15, conduct an analysis of the extraction of the autumn irrigation area, and process to obtain eight autumn irrigation progress maps of the Hetao Irrigation Area on different dates, with a final area reaching 6.5741 million mu. Among them, Figure 7 (a) is the autumn irrigation progress map of the Hetao Irrigation Area on October 15, 2023, Figure 7 (b) is the autumn irrigation progress map of the Hetao Irrigation Area on October 20, 2023, Figure 7 (c) is the autumn irrigation progress map of the Hetao Irrigation Area on October 25, 2023, Figure 7 (d) is the autumn irrigation progress map of the Hetao Irrigation Area on October 30, 2023, Figure 7 (e) is the autumn irrigation progress map of the Hetao Irrigation Area on November 5, 2023, Figure 7 (f) is the autumn irrigation progress map of the Hetao Irrigation Area on November 10, 2023, Figure 7 (g) is the autumn irrigation progress map of the Hetao Irrigation Area on November 15, 2023, Figure 7 (h) is the autumn irrigation progress map of the Hetao Irrigation Area on November 20, 2023.

[0091] Similarly, select GF-1 satellite remote sensing data passing through the Hetao Irrigation District from early October to the end of November, ensuring that the cloud cover is less than 20%. Perform radiometric calibration, atmospheric correction, ortho-correction, and mosaicking synchronously. After mask extraction, obtain the high-resolution No. 1 image of the Hetao Irrigation District that can completely cover it. Subsequently, use the NDWI water index to extract the autumn irrigation area, and further process it to obtain eight autumn irrigation progress maps of the Hetao Irrigation District on different dates, with a final area of 4.8896 million mu, as Figure 8 shown, where Figure 8 (a) is the autumn irrigation progress map of the Hetao Irrigation District on October 11, 2023, Figure 8 (b) is the autumn irrigation progress map of the Hetao Irrigation District on October 15, 2023, Figure 8 (c) is the autumn irrigation progress map of the Hetao Irrigation District on October 18, 2023, Figure 8 (d) is the autumn irrigation progress map of the Hetao Irrigation District on October 23, 2023, Figure 8 (e) is the autumn irrigation progress map of the Hetao Irrigation District on October 27, 2023, Figure 8 (f) is the autumn irrigation progress map of the Hetao Irrigation District on October 31, 2023, Figure 8 (g) is the autumn irrigation progress map of the Hetao Irrigation District on November 13, 2023, Figure 8 (h) is the autumn irrigation progress map of the Hetao Irrigation District on November 16, 2023.

[0092] Furthermore, select Sentinel-2 satellite remote sensing data passing through the Hetao Irrigation District from October 20 to November 25, 2024, ensuring that the cloud cover is less than 20%. Use the MBWI water index to extract the autumn irrigation area, and further process it to obtain autumn irrigation progress maps of the Hetao Irrigation District on different dates. The autumn irrigation area is 4.212 million mu, as Figure 9 shown, where Figure 9 (a) is the autumn irrigation progress map of the Hetao Irrigation District on October 22, 2024, Figure 9 (b) is the autumn irrigation progress map of the Hetao Irrigation District on October 27, 2024, Figure 9 (c) is the autumn irrigation progress map of the Hetao Irrigation District on November 11, 2024, Figure 9 (d) is the autumn irrigation progress map of the Hetao Irrigation District on November 16, 2024, Figure 9 (e) is the autumn irrigation progress map of the Hetao Irrigation District on November 21, 2024.

[0093] Furthermore, use different methods to verify the results of remote sensing extraction of the autumn irrigation area. For the 991 mu area of Lansuo Branch Canal, the area identified by the drone for autumn irrigation is 774 mu, and the area identified by Landsat is 826 mu, with a difference of 52 mu, as Figure 10 shown, where Figure 10 (a) is the processing result of the drone image, Figure 10(b) shows the processing results of Landsat-8 images. The accuracy statistics of the autumn irrigation area identified by the drone and Landsat-8 are shown in Table 4 as follows:

[0094] Table 4

[0095]

[0096] It can be seen from Table 4 that by using the actual field monitoring results to judge whether the autumn irrigation is carried out at the points identified by the drone and Landsat-8, the overall accuracies of the drone identification and Landsat-8 identification are 93.3% and 90.0% respectively. The accuracy of identifying whether it is autumn irrigation by single-day remote sensing images is relatively reliable, but it cannot meet the accuracy verification requirements for every 5 days. In the follow-up, it is necessary to fuse multi-source remote sensing images and train a deep learning model for autumn irrigation classification.

[0097] Furthermore, the MODIS, Landsat 8-9, and Sentinel-2 remote sensing data collected during the autumn irrigation period in the Hetao Irrigation District are the basis for forming a high spatio-temporal resolution image dataset. In the autumn irrigation period of 2023, 45 MODIS images were obtained, with a spatial resolution of 500m and a temporal resolution of 1 day, and 1 image can cover the entire irrigation district. 28 Landsat 8-9 images were obtained, and the observation period for the same point is about 16 days, and about 2 images are needed to cover the entire irrigation district. A total of 121 Sentinel-2 images were collected, and the observation period for the same point of the image is about 5 days, and about 6 images are needed to completely cover the entire irrigation district. In the autumn irrigation period of 2024, 42 MODIS images, 28 Landsat 8-9 images, and 120 Sentinel-2 images were obtained.

[0098] Furthermore, aiming at the problem of large differences in the fusion effects of different spatio-temporal fusion algorithms, a variety of classic spatio-temporal fusion algorithms (such as ESTARFM, STRUM, FSDA, OL-STARFM, Fit-FC, OPDL, etc.) are used to fuse different spatio-temporal resolution data such as MODIS (500m), Landsat (30m), and Sentinel-2 (10m), and then the optimal spatio-temporal fusion algorithm suitable for autumn irrigation information extraction in the Hetao Irrigation District is selected, and a high spatio-temporal resolution image dataset of the autumn irrigation period in the Hetao Irrigation District is formed based on this algorithm. The specific technical route of the spatio-temporal fusion algorithm is as Figure 11As shown in the figure, first, data preprocessing is carried out. The surface reflectance data of MODIS, Landsat, and Sentinel are smoothed and denoised, and various algorithms such as ESTARFM are used for spatio-temporal data fusion. Further, parameters are set, including the sliding window size (w) and the number of similar pixels (N), and the accuracy evaluation of the sliding window size (w) and the number of similar pixels (N) is carried out. Further, fusion schemes are generated, including the RI fusion scheme and the IR fusion scheme, and the RI fusion scheme and the IR fusion scheme are compared and evaluated, including quantitative evaluation and qualitative evaluation.

[0099] The following is a detailed introduction to several spatio-temporal fusion algorithms.

[0100] (1) ESTARFM algorithm flow

[0101] STARFM assumes that the surface reflectance of Landsat images and MODIS images on the same day has a consistent correlation, and uses the spectral information of Landsat images and MODIS images and their weighting functions to predict the surface reflectance FP(t2). Its main steps are as follows:

[0102] 1) Sample the MODIS image to the Landsat image resolution;

[0103] 2) Apply the moving window w to the Landsat image to identify similar pixels;

[0104] 3) Assign weights W to each similar pixel according to the spectral difference, time difference, and distance difference of the images ijk ;

[0105] 4) Calculate the reflectance of the central pixel of the moving window. The calculation formula is:

[0106]

[0107] where i and j are the index positions of Landsat pixels in the moving window; n is the number of similar pixels determined within the moving window; L(x (w / 2) ,y (w / 2) ,t2) is the value of the central pixel of the moving window in FP (t2) ; W ijk is the weight of the similar pixel; C(x i ,y i ,t2) is the value of the MODIS pixel at time t2, L(x i ,y i ,t1) is the value of the Landsat pixel at time t1, and C(xi, y i ,t1) is the value of the MODIS pixel at time t1.

[0108] (2) STRUM algorithm flow

[0109] STRUM predicts the surface reflectance of the image based on the idea of mixed pixel decomposition, and uses Bayes' theorem to constrain the inaccurate endmember spectra that may occur in the spectral decomposition during the mixed pixel decomposition process. The main steps are as follows:

[0110] 1) Perform K-means unsupervised classification on L(t1).

[0111] 2) Apply the moving window w to step 1) to identify the number of endmembers and abundances within the window, and merge the endmembers with abundances less than 0.1 into spectrally similar classes.

[0112] 3) Calculate the change value ΔC of the reflectance between C(t1) and C(t2) within the moving window, and apply Bayes' theory for mixed pixel decomposition to minimize the residual ε caused by uncertain spectral endmembers in spectral unmixing.

[0113] 4) Calculate the change value ΔF(k) of each endmember from time t1 to time t2, and integrate it into L(t1) to predict FP(t2). The main calculation formula is as follows:

[0114] ΔC = A k ×ΔF k +ε;

[0115] FP(t2) = L(t1)+ΔF k ;

[0116] where A k is the abundance value of the k-th endmember.

[0117] (3) OL-STARFM algorithm

[0118] STARFM is the first spatio-temporal fusion model that combines the change information of spectrally similar neighboring pixels and can be expressed as:

[0119]

[0120] The coarse image is resampled to the fine resolution by nearest neighbor interpolation. After the segmentation process of the auxiliary fine image, the principle of the OL-processed version of STARFM (OL-STARFM) can be expressed as:

[0121]

[0122] Among them, M[·] represents taking the median of all pixel values in the object. OL-STARFM uses the median instead of the expected value to eliminate the influence of poor-quality pixels on the entire segmented object. However, this strategy will introduce uncertainties, especially when fusing high-quality observation data. Therefore, the embodiment of this application adds a residual compensation step to enhance the preliminary fusion result:

[0123]

[0124] Among them, the residual R is the difference in the temporal changes between the coarse image and the fine image. The bilinear interpolation method is used to remove the block artifacts in the residual. The above formula performs pixel-level calculations but does not combine adjacent similar information, so it takes less time.

[0125] Preferably, the OL-STARFM algorithm is determined as the spatio-temporal fusion algorithm for the autumn irrigation period in the Hetao Irrigation District in the embodiment of this application.

[0126] Furthermore, the OL-STARFM spatio-temporal fusion algorithm is used to fuse the Landsat-8 and MODIS remote sensing data in 2023 and 2024. The true color data results of the fused data are as Figure 12 and Figure 13 shown, among which, Figure 12 is the image dataset of the Hetao Irrigation District from October 10, 2023 to November 20, 2023, Figure 12 (a) is the image data of the Hetao Irrigation District on October 10, 2023, Figure 12 (b) is the image data of the Hetao Irrigation District on October 15, 2023, Figure 12 (c) is the image data of the Hetao Irrigation District on October 20, 2023, Figure 12 (d) is the image data of the Hetao Irrigation District on October 25, 2023, Figure 12 (e) is the image data of the Hetao Irrigation District on October 30, 2023, Figure 12 (f) is the image data of the Hetao Irrigation District on November 05, 2023, Figure 12 (g) is the image data of the Hetao Irrigation District on November 10, 2023, Figure 12 (h) is the image data of the Hetao Irrigation District on November 15, 2023. Figure 13 is the image dataset of the Hetao Irrigation District from October 10, 2024 to November 15, 2024, Figure 13 (a) is the image data of the Hetao Irrigation District on October 10, 2024, Figure 13 (b) is the image data of the Hetao Irrigation District on October 15, 2024, Figure 13 (c) is the image data of the Hetao Irrigation District on October 20, 2024, Figure 13(d) is the image data of the Hetao Irrigation District on October 25, 2024, Figure 13 (e) is the image data of the Hetao Irrigation District on October 30, 2024, Figure 13 (f) is the image data of the Hetao Irrigation District on November 05, 2024, Figure 13 (g) is the image data of the Hetao Irrigation District on November 10, 2024, Figure 13 (h) is the image of the Hetao Irrigation District on November 15, 2024. Among them, the spatial resolution of the data fusion image is 30m, and the temporal resolution is 5 days. The fused data set can provide reliable data support for the construction of the precise identification model of autumn irrigation farmland and the extraction and analysis of the autumn irrigation process in the later stage.

[0127] Furthermore, the autumn irrigation monitoring of high spatio-temporal resolution remote sensing images based on deep learning aims to utilize the high-frequency data of remote sensing images and the powerful feature extraction and classification capabilities of deep learning models to accurately monitor agricultural activities, especially the autumn irrigation link. However, single remote sensing image data usually cannot provide sufficient spatio-temporal resolution to capture the high-frequency changes of agricultural activities such as autumn irrigation because the temporal resolution of remote sensing images is low, while the monitoring of autumn irrigation in the Hetao Irrigation District requires a higher temporal frequency (monitoring every 5 days). Therefore, to achieve high spatio-temporal resolution monitoring, it is necessary to rely on multi-source remote sensing data to extract spatio-temporal features related to autumn irrigation. At the same time, deep learning has powerful feature extraction and modeling capabilities and can effectively process complex spatial and temporal information in remote sensing images. These models can automatically extract features related to autumn irrigation from the image data.

[0128] Furthermore, the MLP algorithm is selected in the embodiment of this application for crop classification. It has strong fault tolerance for samples and high training efficiency and has been widely used in crop remote sensing classification. As Figure 14 shown, the multi-layer perceptron (MLP) is a feed-forward artificial neural network composed of an input layer, a hidden layer, and an output layer. The hidden layer is located between the input layer and the output layer, and the number of layers can be more than one. The layers are fully connected, that is, any neuron in the upper layer has a connection relationship with all neurons in the lower layer.

[0129] Furthermore, select the activation function for the model. Commonly used activation functions include Sigmoid, Tanh, ReLU, etc. The role of the activation function is to introduce non-linearity into the neural network so that it can learn complex non-linear relationships. The activation function needs to have the following properties: First, it is a continuous and differentiable (allowing non-differentiability at a few points) non-linear function. A differentiable activation function can directly use numerical optimization methods to learn network parameters; second, the activation function and its derivative should be as simple as possible, which is beneficial to improving the network calculation efficiency; finally, the value range of the derivative of the activation function needs to be within a suitable interval, avoiding being too large or too small, otherwise it will affect the training efficiency and stability.

[0130] In a neural network, the ReLU function serves as the activation function of neurons, which is the non-linear output result after the linear transformation of neurons. In other words, for the input vector x from the previous layer of the neural network that enters the neuron, the neuron using the ReLU function will output: to the next layer of neurons or as the output of the entire neural network (depending on the position of the current neuron in the network structure).

[0131] Compared with traditional neural network activation functions, such as the logistic function (Logistic sigmoid) and hyperbolic functions such as tanh, the ReLU function has the following advantages: The first aspect is based on the principle of biomimetics: using linear correction and regularization to adjust the activity (i.e., the output is positive) of neurons in the machine neural network; in contrast, the logistic function reaches 0.5 when the input is 0, that is, it is already in a semi-saturated stable state, which does not meet the expectations of actual biology for simulated neural networks. It should be noted that generally, about 50% of the neurons in a neural network using ReLU are in the activated state. The second aspect is more efficient gradient descent and backpropagation, avoiding the problems of gradient explosion and gradient disappearance; the third aspect is to simplify the calculation process, without the influence of functions such as exponential functions in other complex activation functions; at the same time, the dispersion of activity reduces the overall calculation cost of the neural network. Therefore, after comprehensive comparison, this study selects to use the ReLU function.

[0132] The methods of deep learning interpretation technology mainly include image preprocessing, sample library establishment, autumn irrigation classification model construction, and autumn irrigation classification and verification. The automatic extraction process of autumn irrigation information from remote sensing images based on deep learning is as Figure 15As shown in the figure, first, remote sensing images are processed through spatio-temporal fusion technology to generate a sample database, and data augmentation is performed on it to expand the sample database. Then, the expanded sample database is divided into two parts: training samples and test samples. The training samples are used to train the artificial neural network multi-layer perceptron (MLP) model, and then the test samples are used to evaluate the accuracy of the model. After verification, the multi-temporal fused remote sensing images are input into the trained autumn irrigation classification model to generate an autumn irrigation spatial distribution map for tracking the autumn irrigation process, scope, etc. in the Hetao Irrigation Area. Finally, the generated autumn irrigation spatial distribution map is further verified through unmanned aerial vehicle survey data to ensure the accuracy of the results.

[0133] Specifically, image segmentation and sample library establishment are carried out first. Image segmentation is to segment the synthesized high-temporal and high-spatial resolution images. Sample library establishment is to perform sample annotation based on the segmented images to establish a local sample library. At the same time, in order to expand the volume of this sample set and increase the proportion of common features between sample data, samples are augmented by flipping, rotating, scaling, etc., and the existing samples are expanded to make the sample patterns more diverse.

[0134] Furthermore, an autumn irrigation classification model is constructed. The autumn irrigation classification model construction stage is mainly based on the artificial neural network multi-layer perceptron (MLP) model, using training samples to optimize the deep learning model and construct an autumn irrigation classification model. Autumn irrigation classification is mainly to extract the autumn irrigation area using the optimized model, and use unmanned aerial vehicle and on-site actual survey data to train, test and verify the autumn irrigation area inverted by the autumn irrigation classification model. Among them, the multi-source field verification data includes at least one of field research data, field sampling data, and unmanned aerial vehicle aerial photography data.

[0135] Furthermore, the performance metrics of the deep learning model are the basis for evaluating the autumn irrigation classification model and are also important reference data for designing the model. Performance measurement runs through the training and testing stages of the model. In the embodiments of this application, the performance of the autumn irrigation classification model is evaluated using the evaluation metrics commonly used in classification tasks: accuracy, precision, recall, and F1 score. Most of these metrics are based on the confusion matrix. As shown in Table 5, the confusion matrix describes the matching situation between the prediction results and the true values of the autumn irrigation classification model. The calculation process of the performance metrics of the autumn irrigation classification model is shown in Table 6.

[0136] Table 5

[0137]

[0138] Table 6

[0139]

[0140] Furthermore, the progress of the autumn irrigation area in the Hetao Irrigation District every five days is obtained as follows Figure 16 shown in which Figure 16 is the image dataset of the Hetao Irrigation District from October 10, 2023 to November 20, 2023, Figure 16 (a) is the progress of the autumn irrigation area in the Hetao Irrigation District on October 10, 2023, Figure 16 (b) is the progress of the autumn irrigation area in the Hetao Irrigation District on October 15, 2023, Figure 16 (c) is the progress of the autumn irrigation area in the Hetao Irrigation District on October 20, 2023, Figure 16 (d) is the progress of the autumn irrigation area in the Hetao Irrigation District on October 25, 2023, Figure 16 (e) is the progress of the autumn irrigation area in the Hetao Irrigation District on October 30, 2023, Figure 16 (f) is the progress of the autumn irrigation area in the Hetao Irrigation District on November 05, 2023, Figure 16 (g) is the progress of the autumn irrigation area in the Hetao Irrigation District on November 10, 2023, Figure 16 (h) is the progress of the autumn irrigation area in the Hetao Irrigation District on November 15, 2023.

[0141] From Figure 16 it can be seen that from October 10, 2023 to November 20, 2023, the autumn irrigation area in the Hetao Irrigation District gradually increased and reached 4.377 million mu on November 20, 2023. At the same time, the remotely sensed inverted autumn irrigation area was verified using the actual monitoring results, and the statistical results are shown in Table 7. There are a total of 30 monitoring points in the Hetao Irrigation District (13 for autumn irrigation and 17 for non-autumn irrigation). Through remote sensing, 11 autumn irrigation fields and 15 non-autumn irrigation fields were identified, with relatively high recognition accuracies of 84.6% and 88.2% respectively; there are a total of 30 regional monitoring points for unmanned aerial vehicles (24 for autumn irrigation and 6 for non-autumn irrigation). Through remote sensing, 22 autumn irrigation fields and 5 non-autumn irrigation fields were identified, with recognition accuracies of 91.7% and 83.3% respectively; the overall autumn irrigation accuracy is 88.3%.

[0142] Table 7

[0143]

[0144] Furthermore, in some embodiments, preset verification conditions are included: the accuracy of the initial autumn irrigation classification model is greater than the preset accuracy, and the precision of the initial autumn irrigation classification model is greater than the preset precision, and the recall rate of the initial autumn irrigation classification model is greater than the preset recall rate, and the F1 score of the initial autumn irrigation classification model is greater than the preset score.

[0145] Furthermore, the accuracy of the autumn irrigation classification model in 2024 is evaluated as follows Figure 17As shown, 1,465 survey points during the autumn irrigation period in 2024 were imported into the model. The accuracy of the overall model data test set was 0.9181, and the classification accuracy of autumn irrigation was relatively reliable. The Receiver Operating Characteristic curve (ROC) is as shown in Figure 18 and the Precision-Recall Curve is as shown in Figure 19 and the confusion matrix is as shown in Figure 20 .

[0146] Among them, the ROC curve is a curve drawn through the outputs of classifiers at different thresholds. Its horizontal axis is the False Positive Rate (FPR), and its vertical axis is the True Positive Rate (TPR). TPR is also known as Recall, while FPR is 1 minus the True Positive Rate (Specificity). The ROC curve can show the performance of the classifier at all possible thresholds. The True Positive Rate (TPR), also known as sensitivity, is the ratio of the number of positive samples correctly predicted by the model to the total number of actual positive samples. The False Positive Rate (FPR), also known as the false alarm rate, is the ratio of the number of negative samples mispredicted as positive by the model to the total number of actual negative samples. The AUC value is the area under the ROC curve, ranging from 0.5 to 1.0, and is used to measure the overall performance of the model. The higher the AUC value, the better the classification performance of the model. When AUC is 0.5, it means the performance of the model is the same as random guessing; when AUC is 1, it means the model has perfect classification ability.

[0147] The ROC curve is widely used because it has some unique advantages. First, the ROC curve can provide a comprehensive perspective to observe the performance of the classifier at all classification thresholds. Second, when the distribution of positive and negative samples in the test set changes, the ROC curve has good stability. This is particularly important in real-world applications because in many cases, the problem of class imbalance will be encountered, that is, the number of samples in one class is much larger than that in the other class.

[0148] The Precision-Recall Curve is a tool for evaluating the performance of classification models, suitable for dealing with binary classification problems with imbalanced data sets. It shows the performance of the model by plotting the precision and recall at different probability thresholds.

[0149] Precision is the ratio of the number of positive class samples correctly predicted by the model to the total number of samples predicted as positive class, and the formula is as follows: Precision = True Positive / (True Positive + False Positive)

[0150] Recall is the ratio of the number of positive class samples correctly predicted by the model to the total number of actual positive class samples. The formula is as follows: Recall = True Positives / (True Positives + False Negatives).

[0151] As can be seen, precision focuses on the quality of positive class predictions, while recall focuses on the coverage of positive class instances.

[0152] Furthermore, the data precision classification report of the autumn irrigation classification model in the embodiments of the present application is shown in Table 8.

[0153] Table 8

[0154]

[0155] As can be seen from Table 8, the overall F1 Score of the autumn irrigation classification model is 0.9337, indicating good model operation results.

[0156] In step S102, the remote sensing data of the Hetao Irrigation Area to be classified is input into the preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified. Among them, the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period.

[0157] Exemplarily, the remote sensing data of the Hetao Irrigation Area to be classified is input into the preset autumn irrigation classification model to generate a 5-day-by-5-day image dataset of irrigated farmland during the autumn irrigation periods of 2023 and 2024 in the Hetao Irrigation Area.

[0158] Figure 21 It is the autumn irrigation process distribution map of the Hetao Irrigation Area in 2024. Among them, Figure 21 (a) shows the autumn irrigation area of the Hetao Irrigation Area on October 25, 2024, Figure 21 (b) shows the autumn irrigation area of the Hetao Irrigation Area on October 30, 2024, Figure 21 (c) shows the autumn irrigation area of the Hetao Irrigation Area on November 10, 2024, Figure 21 (d) shows the autumn irrigation area of the Hetao Irrigation Area on November 15, 2024, Figure 21 (e) shows the autumn irrigation area of the Hetao Irrigation Area on November 20, 2024, Figure 21 (f) shows the autumn irrigation area of the Hetao Irrigation Area on November 25, 2024.

[0159] From Figure 21 it can be obtained that the autumn irrigation area was in a period of rapid change from October 25 to October 31, 2024; it was in a relatively stable period from October 31 to November 15, 2024; it was generally stable from November 15 to November 25, 2024; and by November 25, 2024, the autumn irrigation area changed to 4.345 million mu.

[0160] According to the remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to the embodiments of the present application, the remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period is input into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified. Thus, the problems of the traditional ground survey method in monitoring the autumn irrigation process in the Hetao Irrigation Area, such as time-consuming, laborious, inability to provide comprehensive and accurate data, and poor timeliness, are solved. The accurate and timely monitoring of the autumn irrigation range, area, and progress is realized by adopting the spatio-temporal fusion algorithm and remote sensing technology based on deep learning.

[0161] Next, a remote sensing monitoring device for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to the embodiments of the present application is described with reference to the accompanying drawings.

[0162] Figure 22 It is a block schematic diagram of a remote sensing monitoring device for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to the embodiments of the present application.

[0163] As Figure 22 shown, the remote sensing monitoring device 10 for autumn irrigation water in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm includes: an acquisition module 100 and an autumn irrigation classification module 200.

[0164] Among them, the acquisition module 100 is used to acquire the remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period; the autumn irrigation classification module 200 is used to input the remote sensing data of the Hetao Irrigation Area to be classified into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified, where the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period.

[0165] Further, in some embodiments, before inputting the remote sensing data of the Hetao Irrigation Area to be classified into the preset autumn irrigation classification model, the autumn irrigation classification module 200 is further used to: acquire the original remote sensing data of the target Hetao Irrigation Area during the autumn irrigation period, and perform preprocessing on the original remote sensing data of the target Hetao Irrigation Area to obtain preprocessed remote sensing data; extract the preprocessed remote sensing data based on a preset extraction strategy to obtain a preliminary extraction result of the autumn irrigation area; fuse the original remote sensing data based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm to obtain a high spatio-temporal resolution image dataset; train a preset neural network based on the high spatio-temporal resolution image dataset to obtain an initial autumn irrigation classification model, and verify the initial autumn irrigation classification model based on multi-source field verification data, and when the initial autumn irrigation classification model meets the preset verification conditions, use the initial autumn irrigation classification model as the preset autumn irrigation classification model.

[0166] Further, in some embodiments, the autumn irrigation classification model 200 is configured to: calculate the preprocessed remote sensing data using a preset multi-band water body index calculation formula to obtain a water body index calculation result; extract water body areas where the water body index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result; perform a spatial overlay analysis on the remote sensing image extraction result to obtain a preliminary extraction result of the autumn irrigation area.

[0167] Further, in some embodiments, the multi-source field verification data includes at least one of field research data, field sampling data, and UAV aerial photography data.

[0168] Further, in some embodiments, the preset verification conditions include: the accuracy rate of the initial autumn irrigation classification model is greater than a preset accuracy rate, the precision of the initial autumn irrigation classification model is greater than a preset precision, the recall rate of the initial autumn irrigation classification model is greater than a preset recall rate, and the F1 score of the initial autumn irrigation classification model is greater than a preset score.

[0169] It should be noted that the foregoing explanation of the embodiments of the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm also applies to the remote sensing monitoring device for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm of this embodiment, and will not be elaborated here.

[0170] According to the remote sensing monitoring device for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm of the embodiments of the present application, the remote sensing data of the Hetao Irrigation Area to be classified during the autumn irrigation period is input into a preset autumn irrigation classification model to obtain an image dataset of irrigated farmland in the Hetao Irrigation Area to be classified. Thus, the problems of the traditional ground survey method being time-consuming and laborious, unable to provide comprehensive and accurate data, and having poor timeliness in monitoring the autumn irrigation process in the Hetao Irrigation Area are solved, and the precise and timely monitoring of the autumn irrigation scope, area, and progress is realized by adopting the spatio-temporal fusion algorithm and remote sensing technology based on deep learning.

[0171] Figure 23 The following is a schematic structural diagram of an electronic device provided by the embodiments of the present application. The electronic device may include:

[0172] A memory 2301, a processor 2302, and a computer program stored on the memory 2301 and executable on the processor 2302.

[0173] When the processor 2302 executes the program, it implements the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm provided in the above embodiments.

[0174] Further, the electronic device further includes:

[0175] A communication interface 2303 for communication between the memory 2301 and the processor 2302.

[0176] A memory 2301 for storing a computer program that can run on a processor 2302.

[0177] The memory 2301 may include a high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory.

[0178] If the memory 2301, the processor 2302, and the communication interface 2303 are implemented independently, the communication interface 2303, the memory 2301, and the processor 2302 can be interconnected via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 23 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0179] Optionally, in a specific implementation, if the memory 2301, the processor 2302, and the communication interface 2303 are integrated on a single chip, the memory 2301, the processor 2302, and the communication interface 2303 can communicate with each other through an internal interface.

[0180] The processor 2302 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0181] The embodiments of the present application also provide a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation Area based on a spatio-temporal fusion algorithm is implemented.

[0182] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or N embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0183] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of this application, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0184] Although the embodiments of this application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting this application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A remote sensing monitoring method for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm, characterized in that, It includes the following steps: Obtain the remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period; Input the remote sensing data of the Hetao Irrigation District to be classified into a preset autumn irrigation classification model to obtain the irrigated farmland image dataset of the Hetao Irrigation District to be classified. Among them, the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period.

2. The method according to claim 1, wherein Before inputting the remote sensing data of the Hetao Irrigation District to be classified into the preset autumn irrigation classification model, it further includes: Obtain the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period, and preprocess the original remote sensing data of the target Hetao Irrigation District to obtain the preprocessed remote sensing data; Based on a preset extraction strategy, extract the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area; Fuse the original remote sensing data based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm to obtain the high spatio-temporal resolution image dataset; Based on the high spatio-temporal resolution image dataset, train a preset neural network to obtain an initial autumn irrigation classification model, verify the initial autumn irrigation classification model based on multi-source field verification data, and when the initial autumn irrigation classification model meets the preset verification conditions, use the initial autumn irrigation classification model as the preset autumn irrigation classification model.

3. The method according to claim 2, wherein The extraction of the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area includes: Calculate the preprocessed remote sensing data using a preset multi-band water body index calculation formula to obtain a water body index calculation result; Extract the water body areas where the water body index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result; Perform a spatial overlay analysis on the remote sensing image extraction result to obtain the preliminary extraction result of the autumn irrigation area.

4. The method according to claim 2, characterized in that The multi-source field verification data includes at least one of field investigation data, field sampling data, and UAV aerial photography data.

5. The method according to claim 2, wherein The preset verification conditions include: The accuracy rate of the initial autumn irrigation classification model is greater than a preset accuracy rate, the precision of the initial autumn irrigation classification model is greater than a preset precision, the recall rate of the initial autumn irrigation classification model is greater than a preset recall rate, and the F1 score of the initial autumn irrigation classification model is greater than a preset score.

6. A remote sensing monitoring device for autumn irrigation water in the Hetao Irrigation District based on a spatio-temporal fusion algorithm, characterized in that, It includes: An acquisition module for obtaining the remote sensing data of the Hetao Irrigation District to be classified during the autumn irrigation period; An autumn irrigation classification module for inputting the remote sensing data of the Hetao Irrigation District to be classified into a preset autumn irrigation classification model to obtain the irrigated farmland image dataset of the Hetao Irrigation District to be classified. Among them, the preset autumn irrigation classification model is trained by a high spatio-temporal resolution image dataset generated from the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period.

7. The device according to claim 6, characterized in that, Before inputting the remote sensing data of the Hetao Irrigation District to be classified into the preset autumn irrigation classification model, the autumn irrigation classification module is further used for: Obtain the original remote sensing data of the target Hetao Irrigation District during the autumn irrigation period, and preprocess the original remote sensing data of the target Hetao Irrigation District to obtain the preprocessed remote sensing data; Based on a preset extraction strategy, extract the preprocessed remote sensing data to obtain a preliminary extraction result of the autumn irrigation area; Based on the preliminary extraction result of the autumn irrigation area and a preset spatio-temporal fusion algorithm, fuse the original remote sensing data to obtain the high spatio-temporal resolution image dataset; Based on the high spatio-temporal resolution image dataset, train a preset neural network to obtain an initial autumn irrigation classification model, and verify the initial autumn irrigation classification model based on multi-source field verification data. When the initial autumn irrigation classification model meets the preset verification conditions, use the initial autumn irrigation classification model as the preset autumn irrigation classification model.

8. The device according to claim 7, characterized in that, The autumn irrigation classification model is used for: Use a preset multi-band water body index calculation formula to calculate the preprocessed remote sensing data to obtain a water body index calculation result; Extract the water body areas where the water body index calculation result is greater than a preset threshold to obtain a remote sensing image extraction result; Perform a spatial overlay analysis on the remote sensing image extraction result to obtain the preliminary extraction result of the autumn irrigation area.

9. An electronic device, characterized in that, Including: A memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to any one of claims 1-5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to be used for implementing the remote sensing monitoring method for autumn irrigation in the Hetao Irrigation Area based on the spatio-temporal fusion algorithm according to any one of claims 1-5.

Images