Radiometric Calibration Field Site Selection Method, Device, Electronic Equipment and Storage Medium

Abstract

The present invention provides a method, apparatus, electronic device and storage medium for selecting a radiation calibration field site. The method includes: obtaining measurement data of multi-source characteristic parameters of a to-be-measured address; performing normalization processing on the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain a normalization result of each type of characteristic parameter; inputting the normalization result of each type of characteristic parameter into a radiation calibration field site selection evaluation model to obtain an evaluation result of the to-be-measured address output by the radiation calibration field site selection evaluation model; the radiation calibration field site selection evaluation model is trained according to multi-source characteristic parameter data samples of addresses and corresponding class labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters and social environment characteristic parameters. The present invention can achieve intelligent site selection of the radiation calibration field, and the automated calculation process can greatly reduce the influence of manual intervention and subjective factors in the site selection process, and improve the effect of radiation calibration field site selection.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite remote sensing radiation calibration, and in particular to a radiation calibration site selection method, device, electronic equipment and storage medium. Background Art

[0002] The selection of the radiation calibration site address will directly affect the accuracy and frequency of satellite payload radiation calibration. Therefore, choosing a suitable radiation calibration site address is an important basis for satellite remote sensing radiation calibration technology.

[0003] At present, there are some methods that can be used to predict the site selection of radiation calibration sites. However, the existing site selection methods are highly dependent on manual experience and require a combination of expert experience and manual adjustment of the measurement index thresholds. They are inefficient and essentially rely on manual intervention and scoring. It is difficult to ensure that multiple measurement indicators of the selected address achieve optimal results at the same time, resulting in poor results in the site selection of radiation calibration sites.

[0004] Therefore, how to better realize the site selection of radiation calibration sites has become a technical problem that needs to be solved urgently in the industry. Summary of the invention

[0005] The present invention provides a radiation calibration field site selection method, device, electronic equipment and storage medium, which are used to better realize the radiation calibration field site selection.

[0006] The present invention provides a radiation calibration field site selection method, comprising:

[0007] Acquire measurement data of multi-source characteristic parameters of the address to be measured;

[0008] Normalizing the measurement data of each type of characteristic parameters in the multi-source characteristic parameters to obtain a normalized result of each type of characteristic parameters;

[0009] Inputting the normalized results of each type of characteristic parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model;

[0010] The radiation calibration site selection assessment model is obtained by training based on address multi-source characteristic parameter data samples and corresponding category labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters and social environment characteristic parameters.

[0011] According to a radiation calibration field site selection method provided by the present invention, the measurement data of each type of characteristic parameters in the multi-source characteristic parameters are normalized to obtain the normalized results of each type of the characteristic parameters, including:

[0012] Obtain the thresholds of the atmospheric characteristic parameters, the thresholds of the surface characteristic parameters, and the thresholds of the social environment characteristic parameters respectively;

[0013] Based on the measurement data and threshold of the atmospheric characteristic parameters, perform normalization processing on the measurement data of the atmospheric characteristic parameters to obtain the normalized result of the atmospheric characteristic parameters;

[0014] Based on the measurement data and threshold of the surface characteristic parameters, perform normalization processing on the measurement data of the surface characteristic parameters to obtain the normalized result of the surface characteristic parameters;

[0015] Based on the measurement data and threshold of the social environment characteristic parameters, perform normalization processing on the measurement data of the social environment characteristic parameters to obtain the normalized result of the social environment characteristic parameters.

[0016] According to a method for selecting a radiation calibration field provided by the present invention, the atmospheric characteristic parameters include at least three of the following: surface albedo, aerosol optical depth, cloud cover, normalized difference vegetation index, surface humidity, and surface temperature.

[0017] According to a method for selecting a radiation calibration field provided by the present invention, the surface characteristic parameters include at least three of the following: rainfall, surface type, distance to the main stem of the surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information, and flood discharge area information.

[0018] According to a method for selecting a radiation calibration field provided by the present invention, the social environment characteristic parameters include at least one of the following: traffic accessibility and population density.

[0019] According to a method for selecting a radiation calibration field provided by the present invention, the radiation calibration field site selection evaluation model includes an improved artificial neural network model, a spatial feature extraction model, a temporal feature extraction model, a fully connected layer, and a classification model; the improved artificial neural network model includes an input layer and a hidden layer connected in sequence;

[0020] The step of inputting the normalized results of each type of the characteristic parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-tested address output by the radiation calibration field site selection evaluation model includes:

[0021] Input the normalized results of each type of the characteristic parameters into the improved artificial neural network model to obtain the integrated feature vector groups corresponding to each type of the characteristic parameters output by the hidden layer of the improved artificial neural network model;

[0022] Input the integrated feature vector group corresponding to each type of the feature parameters into the spatial feature extraction model, and obtain the spatial feature vector group corresponding to each type of the feature parameters output by the spatial feature extraction model;

[0023] Input the spatial feature vector group corresponding to each type of the feature parameters into the temporal feature extraction model, and obtain the temporal feature vector group corresponding to each type of the feature parameters output by the temporal feature extraction model;

[0024] Input the temporal feature vector group corresponding to each type of the feature parameters into the fully connected layer, and obtain the spatio-temporal feature vector corresponding to each type of the feature parameters output by the fully connected layer;

[0025] Input the spatio-temporal feature vector corresponding to each type of the feature parameters into the classification model, and obtain the evaluation result of the to-be-tested address output by the classification model.

[0026] According to a radiation calibration field site selection method provided by the present invention, the radiation calibration field site selection evaluation model adopts an artificial neural network model; before inputting the normalization result of each type of the feature parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-tested address output by the radiation calibration field site selection evaluation model, the method further includes: performing a normalization process on the address multi-source feature parameter data sample to obtain the normalization result of the address multi-source feature parameter data sample;

[0027] Use the normalization result of the address multi-source feature parameter data sample and the category label corresponding to the address multi-source feature parameter data sample as a group of training samples, and obtain multiple groups of training samples;

[0028] Initialize the connection weights of the artificial neural network model;

[0029] Use the multiple groups of training samples to train the artificial neural network model, iteratively optimize the connection weights of the artificial neural network model until the number of training times reaches the maximum number of training times, and obtain the trained artificial neural network model;

[0030] According to the trained artificial neural network model, obtain the radiation calibration field site selection evaluation model.

[0031] The present invention also provides a radiation calibration field site selection device, including:

[0032] An acquisition module, configured to acquire measurement data of multi-source feature parameters of a to-be-tested address;

[0033] A processing module, configured to perform a normalization process on the measurement data of each type of feature parameters in the multi-source feature parameters to obtain the normalization result of each type of the feature parameters;

[0034] An evaluation module, configured to input the normalization result of each type of the feature parameters into a radiometric calibration field site selection evaluation model, and obtain an evaluation result of the to-be-tested address output by the radiometric calibration field site selection evaluation model;

[0035] The radiometric calibration field site selection evaluation model is trained according to address multi-source feature parameter data samples and corresponding class labels; the multi-source feature parameters include atmospheric feature parameters, surface feature parameters, and social environment feature parameters.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the radiometric calibration field site selection method as described in any one of the above is implemented.

[0037] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the radiometric calibration field site selection method as described in any one of the above is implemented.

[0038] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the radiometric calibration field site selection method as described in any one of the above is implemented.

[0039] The radiometric calibration field site selection method, device, electronic device, and storage medium provided by the present invention, by considering the influence of the calibration field regional environmental characteristics on the satellite remote sensing radiometric calibration field site selection, fully excavate the multi-source features of the radiometric calibration field from three aspects of the atmospheric environment, surface environment, and social environment, and through a deep learning algorithm, use address multi-source feature parameter data samples and corresponding class labels to train a radiometric calibration field site selection evaluation model, enabling the model to learn the prior knowledge and rules of site selection. Then, for the measurement data of the multi-source feature parameters of the to-be-tested address, its normalization result is input into the radiometric calibration field site selection evaluation model, and the model can automatically predict and output an evaluation result of whether the to-be-tested address is a radiometric calibration field, which can realize the intelligent site selection of the radiometric calibration field. At the same time, the automated calculation process can greatly reduce the influence of manual intervention and subjective factors in the site selection process, is conducive to changing the dependence on expert qualitative selection in the traditional mode, and improves the effect of radiometric calibration field site selection. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1It is a schematic flowchart of the method for selecting a radiation calibration field provided by the present invention.

[0042] Figure 2 It is a schematic structural diagram of the device for selecting a radiation calibration field provided by the present invention.

[0043] Figure 3 It is a schematic physical structure diagram of the electronic device provided by the present invention. Detailed implementation manners

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] The following Figures 1 - 3 describes the method, device, electronic device, and storage medium for selecting a radiation calibration field of the present invention.

[0046] Figure 1 It is a schematic flowchart of the method for selecting a radiation calibration field provided by the present invention. As Figure 1 shown, the method includes the following:

[0047] Step 110: Obtain measurement data of multi-source characteristic parameters of the address to be measured;

[0048] Step 120: Perform normalization processing on the measurement data of the multi-source characteristic parameters to obtain a normalization result of the multi-source characteristic parameters; Step 130: Input the normalization result of the multi-source characteristic parameters into the radiation calibration field site selection evaluation model to obtain an evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model;

[0049] The radiation calibration field site selection evaluation model is trained according to the address multi-source characteristic parameter data samples and the corresponding category labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters.

[0050] It should be noted that due to the complexity of the remote sensing image imaging process, the electromagnetic wave energy received by the sensor is inconsistent with the energy radiated by the target itself. The energy output by the sensor contains various distortions caused by factors such as the sun position, atmospheric conditions, terrain effects, and the performance of the sensor itself. These distortions are not the radiation of the ground target. Therefore, they will affect the use and understanding of the image and must be corrected and eliminated. The basic methods for correction and elimination are radiation calibration and radiation calibration.

[0051] Therefore, it can be understood that radiometric calibration is mainly used to calibrate the radiance of remote sensing images to achieve quantitative remote sensing. Radiometric calibration can generally also be called calibration, and its main purpose is to ensure the accuracy of the remote sensing data obtained by the sensor.

[0052] Specifically, the multi-source feature parameters described in the embodiments of the present invention refer to the regional environmental feature data associated with radiometric calibration, which may specifically include atmospheric feature parameters, surface feature parameters, and social environmental feature parameters.

[0053] Among them, the atmospheric feature parameters are parameters used to characterize the regional atmospheric environment features, such as aerosol optical thickness, cloud cover, and other parameters; the surface feature parameters are parameters used to characterize the regional surface environment features, such as rainfall, surface type, and other parameters; the social environmental feature parameters are parameters used to characterize the regional human activity environment features, such as population density, and other parameters.

[0054] The address to be measured described in the embodiments of the present invention refers to the address area where the regional multi-source feature parameters meet specific threshold conditions, which is used to measure whether it meets the requirements of the radiometric calibration field. For example, for different feature parameters, an address with an aerosol optical thickness less than 0.5, a cloud cover less than 0.4, a rainfall less than 300 mm, a flat and non-rough terrain surface type, and a sparse population distribution is selected as the address to be measured.

[0055] The radiometric calibration field site selection evaluation model described in the embodiments of the present invention is obtained by training a deep neural network model based on the address multi-source feature parameter data samples and the corresponding class labels. It is used to learn the prior knowledge and rules of radiometric calibration field site selection, identify and predict the input multi-source feature parameter data, and determine whether the address to be measured is a suitable radiometric calibration field address.

[0056] It can be understood that the address multi-source feature parameter data samples may include atmospheric feature parameter measurement data samples, surface feature parameter measurement data samples, and social environmental feature parameter measurement data samples.

[0057] Among them, the deep neural network can adopt an artificial neural network (ANN), such as a backpropagation neural network (BP), a radial basis function neural network (RBF), etc., and can also adopt a model jointly composed of an ANN model and a convolutional neural network (CNN) model, a long short-term memory (LSTM) artificial neural network model, and an extreme gradient boosting (XGBoost) model. It can also be other neural networks used for radiometric calibration field site selection evaluation, which is not specifically limited in the present invention.

[0058] Among them, the model training samples are composed of multiple groups of address multi-source feature parameter data samples carrying class labels.

[0059] The class labels described in the embodiments of the present invention belong to numerical labels with values of 1 or 0. Among them, "1" represents the label data of a suitable radiometric calibration field address, and "0" represents the label data of an unsuitable radiometric calibration field address. It can be obtained in advance in the form of manual annotation.

[0060] The class labels described in the embodiments of the present invention are determined in advance according to the address multi-source feature parameter data samples and are in one-to-one correspondence with the address multi-source feature parameter data samples. That is to say, each address multi-source feature parameter data sample in the training sample is preset to carry a corresponding class label.

[0061] In the embodiment of the present invention, in step 110, after determining the address to be measured, the measurement data of the multi-source feature parameters of the address to be measured can be obtained through a remote sensing satellite data platform, including the measurement data of atmospheric feature parameters, surface feature parameters, and social environment feature parameters.

[0062] Furthermore, in order to improve the model processing efficiency, in the embodiment of the present invention, in step 120, a numerical normalization processing method is adopted to normalize the measurement data of the multi-source feature parameters to obtain the normalization result of the multi-source feature parameters.

[0063] Further, in the embodiments of the present invention, in step 130, the normalization result of the multi-source feature parameters obtained in step 120 is input into the radiation calibration field site selection evaluation model. Through data recognition and prediction of the radiation calibration field site selection evaluation model, the evaluation result of the to-be-tested address can be accurately obtained to determine whether the to-be-tested address is a suitable radiation calibration field address. Among them, the evaluation result can be output in the form of a score. Furthermore, according to the radiation calibration fields with the top three scores, a suitable area can be selected according to the weather conditions for the calibration task.

[0064] The radiation calibration field site selection method according to the embodiments of the present invention, by considering the influence of the environmental characteristics of the calibration field area on the site selection of satellite remote sensing radiation calibration fields, fully excavates the multi-source features of the radiation calibration field from three aspects: the atmospheric environment, the surface environment, and the social environment, and through a deep learning algorithm, uses the multi-source feature parameter data samples of the address and the corresponding category labels to train the radiation calibration field site selection evaluation model, enabling the model to learn the prior knowledge and rules of site selection. Furthermore, for the measurement data of the multi-source feature parameters of the to-be-tested address, its normalization result is input into the radiation calibration field site selection evaluation model, and the model can automatically predict and output the evaluation result of whether the to-be-tested address is a radiation calibration field, which can realize the intelligent site selection of the radiation calibration field. At the same time, the automated calculation process can greatly reduce the influence of manual intervention and subjective factors in the site selection process, which is conducive to changing the dependence on expert qualitative selection in the traditional mode and improving the effect of radiation calibration field site selection.

[0065] Compared with the prior art, the method according to the embodiments of the present invention, by considering the radiation calibration field task scenario and diverse site selection objectives, designs quantitative evaluation indicators, enabling the evaluation of the radiation calibration field to be transformed from manual qualitative evaluation to quantitative evaluation, which is conducive to carrying out automated site selection prediction and simplifying the landing point selection process. At the same time, in the embodiments of the present invention, a more comprehensive index parameter system is adopted, which can improve the work efficiency of carrying out site selection analysis on a large scale.

[0066] Based on the content of the above embodiments, as an optional embodiment, the atmospheric characteristic parameters include at least three of the following: surface albedo, aerosol optical depth, cloud cover, normalized difference vegetation index, surface humidity, and surface temperature.

[0067] Specifically, in the embodiments of the present invention, the selected atmospheric characteristic parameters include surface albedo, aerosol optical depth, cloud cover, normalized difference vegetation index, surface humidity, and surface temperature.

[0068] Among them, the surface albedo refers to the ratio of the solar radiation reflected by the earth's surface to the solar radiation reaching the ground, which quantifies the radiation interaction between the atmosphere and the earth's surface. This ratio can reflect the absorption ability of the earth's surface to solar radiation. The magnitude of the surface albedo is affected by various factors, including vegetation coverage, soil moisture, solar altitude angle, and weather conditions, etc. For example, the more vegetation cover and the higher the soil moisture, the lower the surface albedo usually is, because vegetation and water can absorb more solar radiation. The surface albedo has an important impact on the effect of radiometric calibration.

[0069] Aerosol optical depth is a physical quantity that describes the attenuation effect of aerosols in the atmosphere on light, and it is determined by measuring the attenuation degree of light by aerosol components in the atmosphere. It is found that in addition to the energy attenuation of the atmospheric transmittance during the remote sensing imaging process, more importantly, due to the degradation effect of scattering on the imaging quality, the increase in the aerosol optical depth of the atmosphere leads to the enhancement of the aerosol scattering intensity. Thus, aerosols have a degradation effect on remote sensing imaging in the spatial domain, resulting in a reduction in the quality of remote sensing images.

[0070] Cloud cover is a measure of cloud amount, which reflects the coverage of clouds on the earth's surface. When the cloud cover reaches a certain level, it will directly affect the imaging quality of remote sensing images and the clarity and continuity of the observed surface features.

[0071] The Normalized Difference Vegetation Index (NDVI) is one of the important parameters reflecting the growth and nutritional information of surface crops. It can eliminate most of the irradiance changes related to instrument calibration, solar angle, terrain, cloud shadow, and atmospheric conditions, and enhance the response ability of remote sensing images to vegetation.

[0072] Surface humidity refers to the dryness or wetness of the ground, which reflects the content and distribution of surface moisture. Surface temperature, also known as ground temperature, is the temperature condition at the interface between the atmosphere and the earth's surface. Both have an important impact on the effect of radiometric calibration.

[0073] In the embodiments of the present invention, the atmospheric characteristic parameters include at least three of the following: surface albedo, aerosol optical depth, cloud cover, Normalized Difference Vegetation Index, surface humidity, and surface temperature. That is to say, any three parameters, or more parameters among the above six parameters, can be selected for processing to participate in the subsequent radiometric calibration field site selection evaluation calculation.

[0074] The method of the embodiment of the present invention takes into account the influence of the atmospheric environmental characteristics of the calibration field area on the selection of the satellite remote sensing radiation calibration field, fully excavates the relevant atmospheric characteristic parameters affecting the radiation calibration effect in the area, selects the surface albedo, aerosol optical thickness, cloud cover, normalized difference vegetation index, surface humidity and surface temperature, improves the richness of the research parameters, and is conducive to improving the effect of the radiation calibration field selection.

[0075] Based on the content of the above embodiment, as an optional embodiment, the surface characteristic parameters include at least three of the following: rainfall, surface type, distance to the main trunk of the surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information and flood discharge area information.

[0076] Specifically, in the embodiment of the present invention, the selected surface characteristic parameters include rainfall, surface type, distance to the main trunk of the surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information and flood discharge area information.

[0077] Among them, rainfall refers to the total amount of precipitation on the surface within a certain period of time.

[0078] The surface type refers to the topographic features to which the surface structure of the address to be measured belongs, including flat and non-rough terrain, uneven and rough terrain, flat and rough terrain, and uneven and non-rough terrain.

[0079] The distance to the main trunk of the surrounding river refers to the distance between the address to be measured and the main river channel of the surrounding river.

[0080] The distance to the tributary of the surrounding river refers to the distance between the address to be measured and the branch river channel of the surrounding river.

[0081] The distance to the surrounding dam refers to the distance between the address to be measured and the branch river channel of the existing dam in the surrounding area.

[0082] The land use information refers to the data and information about the land use status of the address to be measured, indicating whether it is a farmland area, whether there are residents, etc.

[0083] The flood discharge area information refers to the information indicating whether the area is a flood discharge area and whether there are flood disasters.

[0084] It has been found through research that the above-mentioned several types of surface characteristic parameters will all affect the effect of the radiation calibration field selection.

[0085] It should be noted that a suitable radiation calibration field usually selects some flat places with little rainfall, high and uniform surface reflectivity.

[0086] In an embodiment of the present invention, the surface feature parameters include at least three of the following: rainfall, surface type, distance to the main surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information, and flood discharge area information. That is to say, any three of the above eight parameters, or more parameters, can be selected for processing to participate in the subsequent radiation calibration field site selection evaluation calculation.

[0087] The method of the embodiment of the present invention fully excavates the relevant surface feature parameters affecting the radiation calibration effect in the region by the influence of the surface environment characteristics of the calibration field area on the satellite remote sensing radiation calibration field site selection, and selects rainfall, surface type, distance to the main surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information, and flood discharge area information, further improving the richness of the research parameters and being beneficial to further improving the effect of the radiation calibration field site selection.

[0088] Based on the content of the above embodiment, as an optional embodiment, the social environment feature parameters include at least one of the following: traffic accessibility and population density.

[0089] Specifically, in the embodiment of the present invention, the selected social environment feature parameters include traffic accessibility and population density. Among them, traffic accessibility is used to reflect the traffic conditions of the address to be measured and can be divided into easy to reach and not easy to reach, and its specific value can be determined according to the actual traffic conditions of the address to be measured. Population density is used to reflect the population distribution of the address to be measured.

[0090] In the embodiment of the present invention, the traffic accessibility and population density of the regional address can also be considered as the parameter conditions for the radiation calibration field site selection.

[0091] In the embodiment of the present invention, the social environment feature parameters include at least one of the following: traffic accessibility and population density. That is to say, any one of the above two parameters, or both parameters, can be selected for processing to participate in the subsequent radiation calibration field site selection evaluation calculation.

[0092] The method of the embodiment of the present invention selects traffic accessibility and population density for research by considering the influence of the social environment characteristics of the calibration field area on the satellite remote sensing radiation calibration field site selection, fully excavates the relevant social environment feature parameters affecting the radiation calibration effect in the region, and is beneficial to further improving the effect of the radiation calibration field site selection.

[0093] Based on the content of the above embodiment, as an optional embodiment, normalization processing is performed on the measurement data of each type of feature parameter in the multi-source feature parameters to obtain the normalization result of each type of feature parameter, including:

[0094] Obtain the thresholds of atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters respectively;

[0095] Based on the measured data and threshold of the atmospheric characteristic parameters, normalize the measured data of the atmospheric characteristic parameters to obtain the normalized result of the atmospheric characteristic parameters;

[0096] Based on the measured data and threshold of the surface characteristic parameters, normalize the measured data of the surface characteristic parameters to obtain the normalized result of the surface characteristic parameters;

[0097] Based on the measured data and threshold of the social environment characteristic parameters, normalize the measured data of the social environment characteristic parameters to obtain the normalized result of the social environment characteristic parameters.

[0098] Specifically, the thresholds of the atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters described in the embodiments of the present invention are all used to screen the addresses to be measured for calibration field site selection prediction, and they can all be determined according to the actual radiation calibration field statistical data.

[0099] In a specific embodiment of the present invention, for the atmospheric characteristic parameters, the threshold of the surface albedo can be set to 0.2, the threshold of the aerosol optical depth can be set to 0.5, the threshold of the cloud cover can be set to 0.4, the threshold of the NDVI value can be set to 0.2, the threshold of the surface temperature can be set to 285K, and the threshold of the surface humidity can be set to 0.53. That is to say, in this embodiment, the measured data of the atmospheric characteristic parameters of the address to be measured need to meet the following conditions, namely: the surface albedo is greater than 0.2, the aerosol optical depth is less than 0.5, the cloud cover is less than 0.4, the NDVI value is less than 0.2, the surface temperature is greater than 285K, and the surface humidity is less than 0.53.

[0100] For the surface characteristic parameters, the threshold of the rainfall can be set to 300mm, the threshold of the surface type can be set to the threshold for characterizing flat and non-rough terrain, the threshold of the distance to the main stem of the surrounding river can be set to 1000 meters, the threshold of the distance to the tributary of the surrounding river can be set to 300 meters, the threshold of the distance to the surrounding dam can be set to 3000 meters, the threshold of the slope can be set to 4%, the threshold of the land use information can be set to the threshold for characterizing no farmland and residents, and the threshold of the flood discharge area information can be set to the threshold for characterizing no flood disasters. That is to say, in this embodiment, the measured data of the surface characteristic parameters of the address to be measured need to meet the following conditions, namely: the rainfall is less than 300mm, the surface type is flat and non-rough terrain, the distance to the main stem of the surrounding river is greater than 1000 meters, the distance to the tributary of the surrounding river is greater than 300 meters, the distance to the surrounding dam is greater than 3000 meters, the slope is less than 4%, the land use information is no farmland and residents, and the flood discharge area information is no flood disasters.

[0101] For the social environment characteristic parameters, the threshold of traffic accessibility can be set as the threshold for characterizing easy access, and the threshold of population density can be set as the threshold for characterizing less population distribution. That is to say, in this embodiment, the measurement data of the social environment characteristic parameters of the address to be measured need to meet the following conditions, namely: the traffic accessibility is easy to reach, and the population density is less population distribution.

[0102] Furthermore, in the embodiment of the present invention, the thresholds of the atmospheric characteristic parameters, the surface characteristic parameters, and the social environment characteristic parameters are respectively obtained, and the measurement data of the atmospheric characteristic parameters are normalized by using the measurement data and the threshold of the atmospheric characteristic parameters to obtain the normalized result of the atmospheric characteristic parameters.

[0103] Specifically, for the atmospheric characteristic parameters, assuming that the threshold of the surface albedo is 0.2, the threshold of the aerosol optical depth is 0.5, the threshold of the cloud cover is 0.4, the threshold of the NDVI value is 0.2, the threshold of the surface temperature is 285K, and the threshold of the surface humidity is 0.53, the value ranges of the parameters in the atmospheric characteristic parameters are determined, that is: the surface albedo is greater than 0.2, the aerosol optical depth is less than 0.5, the cloud cover is less than 0.4, the NDVI value is less than 0.2, the surface temperature is greater than 285K, and the surface humidity is less than 0.53. Furthermore, the maximum-minimum normalization method can be used to normalize various measurement data of the atmospheric characteristic parameters, so that the values of all remote sensing measurement data are reduced to the range of [0, 1], and thus the normalized result of the atmospheric characteristic parameters can be obtained.

[0104] Among them, the specific calculation formula of the maximum-minimum normalization is as follows:

[0105]

[0106] In the formula, X represents the observed value of a certain remote sensing data, X max represents the maximum observed value of this remote sensing data, X min represents the minimum observed value of this remote sensing data, and Y represents the normalized result corresponding to this remote sensing data.

[0107] Similarly, for the surface characteristic parameters and the social environment characteristic parameters, according to the above normalization method, the measurement data of the surface characteristic parameters can be normalized by using the measurement data and the threshold of the surface characteristic parameters to obtain the normalized result of the surface characteristic parameters; the measurement data of the social environment characteristic parameters can be normalized by using the measurement data and the threshold of the social environment characteristic parameters to obtain the normalized result of the social environment characteristic parameters.

[0108] The method of the embodiment of the present invention performs data normalization processing on three types of characteristic data, namely atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters of the address to be measured in advance, so that the numerical ranges between different characteristics can be unified to the same scale, eliminating the influence of dimensions between different characteristics, which is beneficial to improving the training effect and prediction accuracy of the model.

[0109] Based on the content of the above embodiment, as an optional embodiment, the radiation calibration field site selection evaluation model includes an improved ANN model, a spatial feature extraction model, a temporal feature extraction model, a fully connected (FC) layer, and a classification model; the improved ANN model includes an input layer and a hidden layer connected in sequence;

[0110] Input the normalization results of each type of characteristic parameter into the radiation calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model, including:

[0111] Input the normalization results of each type of characteristic parameter into the improved artificial neural network model to obtain the integrated feature vector group corresponding to each type of characteristic parameter output by the hidden layer of the improved artificial neural network model;

[0112] Input the integrated feature vector group corresponding to each type of characteristic parameter into the spatial feature extraction model to obtain the spatial feature vector group corresponding to each type of characteristic parameter output by the spatial feature extraction model;

[0113] Input the spatial feature vector group corresponding to each type of characteristic parameter into the temporal feature extraction model to obtain the temporal feature vector group corresponding to each type of characteristic parameter output by the temporal feature extraction model;

[0114] Input the temporal feature vector group corresponding to each type of characteristic parameter into the FC layer to obtain the spatio-temporal feature vector corresponding to each type of characteristic parameter output by the FC layer;

[0115] Input the spatio-temporal feature vector corresponding to each type of characteristic parameter into the classification model to obtain the evaluation result of the address to be measured output by the classification model.

[0116] Specifically, the integrated feature vector group described in the embodiment of the present invention refers to the result obtained by processing the normalization results of each type of characteristic parameter by using the spatial integration characteristics of the hidden layer neurons in the improved ANN model. In this embodiment, the improved ANN model is obtained by removing the output layer of the conventional ANN model and is composed of only an input layer and a hidden layer.

[0117] The spatial feature extraction model described in the embodiment of the present invention is used to extract the spatial features of the input data, and specifically, a CNN model or other deep neural networks that can extract the spatial features of data can be used. The present invention does not make specific limitations on this.

[0118] The spatial feature vector group described in the embodiments of the present invention refers to the data results obtained by processing the integrated feature vector group corresponding to each type of feature parameter through a spatial feature extraction model.

[0119] The time feature extraction model described in the embodiments of the present invention is used to extract the time features of the input data. Specifically, it can adopt an LSTM artificial neural network model, or other deep neural networks that can extract the time features of data. The present invention does not make specific limitations in this regard.

[0120] The time feature vector group described in the embodiments of the present invention refers to the data results obtained by processing the spatial feature vector group corresponding to each type of feature parameter through a time feature extraction model.

[0121] The classification model described in the embodiments of the present invention can specifically adopt an XGBoost model, or other boosting tree models or neural network models that can be used for classification recognition. The present invention does not make specific limitations in this regard.

[0122] In the embodiments of the present invention, after obtaining the normalization results of each type of feature parameter of the address to be measured, the normalization results of each type of feature parameter can be further input into an improved artificial neural network model. After passing through its input layer and hidden layer in sequence, an integrated feature vector group corresponding to each type of feature parameter is output, including an integrated feature vector group corresponding to atmospheric feature parameters, an integrated feature vector group corresponding to surface feature parameters, and an integrated feature vector group corresponding to social environment feature parameters.

[0123] Furthermore, in the embodiments of the present invention, the integrated feature vector group corresponding to each type of feature parameter is input into a spatial feature extraction model. Through the feature extraction operation of the spatial feature extraction model, a spatial feature vector group corresponding to each type of feature parameter can be output, including a spatial feature vector group corresponding to atmospheric feature parameters, a spatial feature vector group corresponding to surface feature parameters, and a spatial feature vector group corresponding to social environment feature parameters.

[0124] In the embodiments of the present invention, then the spatial feature vector group corresponding to each type of feature parameter is input into a time feature extraction model again. Through the feature extraction operation of the time feature extraction model, a time feature vector group corresponding to each type of feature parameter can be output, including a time feature vector group corresponding to atmospheric feature parameters, a time feature vector group corresponding to surface feature parameters, and a time feature vector group corresponding to social environment feature parameters.

[0125] Further, in the embodiments of the present invention, the time feature vector groups corresponding to each type of feature parameter are input into the FC layer for feature integration and data dimensionality reduction, and the spatio-temporal feature vectors corresponding to each type of feature parameter are output. It can be understood that the spatio-temporal feature vectors contain the time features and spatial features of the feature parameters.

[0126] Finally, the spatio-temporal feature vectors corresponding to each type of feature parameter output by the above FC layer are input into the classification model, and the evaluation result of the address to be measured can be obtained.

[0127] Optionally, in the embodiments of the present invention, the radiometric calibration field site selection evaluation model can directly adopt a conventional ANN model, where the ANN model includes an input layer, a hidden layer, and an output layer. By using the multi-source feature parameter data samples of the address and the corresponding class labels to train the ANN model, the radiometric calibration field site selection evaluation model is obtained.

[0128] Optionally, in the embodiments of the present invention, the radiometric calibration field site selection evaluation model includes the improved ANN model, CNN model, LSTM model, FC layer, and XGBoost model connected in sequence, where the ANN model includes an input layer, a hidden layer, and an output layer. By using the multi-source feature parameter data samples of the address and the corresponding class labels to train the ANN model, the radiometric calibration field site selection evaluation model is obtained.

[0129] The method of the embodiments of the present invention constructs a radiometric calibration field site selection evaluation model through an improved ANN model, a spatial feature extraction model, a time feature extraction model, an FC layer, and a classification model, further integrating the spatio-temporal variation features of various feature parameters of the address to be measured. Compared with the method that only uses the ANN model, the classification and recognition accuracy of the model can be further improved, and the accuracy of the radiometric calibration field site selection evaluation can be improved.

[0130] Based on the content of the above embodiments, as an optional embodiment, the radiometric calibration field site selection evaluation model adopts an ANN model; before inputting the normalization result of each type of feature parameter into the radiometric calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiometric calibration field site selection evaluation model, the method further includes:

[0131] Perform normalization processing on the multi-source feature parameter data samples of the address to obtain the normalization result of the multi-source feature parameter data samples of the address;

[0132] Take the normalization result of the multi-source feature parameter data samples of the address and the class labels corresponding to the multi-source feature parameter data samples of the address as a set of training samples, and obtain multiple sets of training samples;

[0133] Initialize the connection weights of the ANN model;

[0134] The ANN model is trained using multiple groups of training samples, and the connection weights of the ANN model are iteratively optimized until the number of training times reaches the maximum number of training times, obtaining a trained ANN model;

[0135] According to the trained ANN model, a radiometric calibration site selection evaluation model is obtained.

[0136] Specifically, in the embodiment of the present invention, before inputting the normalization result of the multi-source feature parameters into the radiometric calibration site selection evaluation model to obtain the evaluation result of the to-be-tested address output by the radiometric calibration site selection evaluation model, the radiometric calibration site selection evaluation model also needs to be trained. Among them, the radiometric calibration site selection evaluation model uses an ANN model.

[0137] In the embodiment of the present invention, by using the pre-collected address multi-source feature parameter data samples and the category labels corresponding to each address multi-source feature parameter data sample, the ANN model is trained, and the specific implementation manner of the model training can be described as follows.

[0138] In the embodiment of the present invention, first, the thresholds of each feature parameter are obtained, and the address multi-source feature parameter data samples are normalized to obtain the normalization results of the address multi-source feature parameter data samples. Furthermore, the normalization results of the address multi-source feature parameter data samples and the category labels corresponding to the address multi-source feature parameter data samples can be used as a group of training samples, and multiple groups of training samples can be obtained according to different multi-source feature parameter data samples. Optionally, 80% of the samples can be randomly selected from the training samples as the training set, and the remaining 20% as the test set.

[0139] Furthermore, in the embodiment of the present invention, the ANN model is trained using the sample data in the training set. First, the connection weights of the ANN model are initialized. Specifically, for the atmospheric feature parameters, the initial connection weight of the surface albedo can be set to 0.03, the initial connection weight of the aerosol optical depth can be set to 0.04, the initial connection weight of the cloud cover can be set to 0.03, the initial connection weight of the NDVI value can be set to 0.09, the initial connection weight of the surface temperature value can be set to 0.04, and the initial connection weight of the surface humidity can be set to 0.07.

[0140] For the surface feature parameters, the initial connection weight of rainfall can be set to 0.04, the initial connection weight of surface type can be set to 0.06, the initial connection weight of the distance to the main stem of the surrounding river can be set to 0.06, the initial connection weight of the distance to the tributary of the surrounding river can be set to 0.06, the initial connection weight of the distance to the surrounding dam can be set to 0.06, the initial connection weight of slope can be set to 0.1, the initial connection weight of land use information can be set to 0.12, and the initial connection weight of flood discharge area information can be set to 0.04.

[0141] For the social environment feature parameters, the initial connection weight of traffic accessibility can be set to 0.07, and the initial connection weight of population density can be set to 0.03.

[0142] Further, after initializing the connection weights of the ANN model, the obtained multiple groups of training samples can be used to train the ANN model, iteratively optimize the connection weights of the ANN model until the number of training times reaches the maximum number of training times, and obtain the trained ANN model, so as to finally train the radiation calibration field site selection evaluation model.

[0143] In the embodiment of the present invention, the trained radiation calibration field site selection evaluation model can be used to perform site selection prediction on the entire research area. For each input sample, the site selection prediction result is obtained, and engineering operations are performed by judging whether the basic engineering indexes of the predicted site selection result meet the conditions, and the final site selection judgment is made.

[0144] The method of the embodiment of the present invention can improve the model accuracy and the accuracy of predicting the radiation calibration field address by using the normalization results of the address multi-source feature parameter data samples and the category labels corresponding to the address multi-source feature parameter data samples as a group of training samples to train the ANN model to obtain the radiation calibration field site selection evaluation model.

[0145] Next, the radiation calibration field site selection device provided by the present invention will be described. The radiation calibration field site selection device described below can be mutually referred to the radiation calibration field site selection method described above.

[0146] Figure 2 It is a schematic structural diagram of the radiation calibration field site selection device provided by the present invention, as Figure 2 shown, including: an acquisition module 210, a processing module 220, and an evaluation module 230 that are connected in sequence.

[0147] Among them, the acquisition module 210 is used to acquire the measurement data of the multi-source feature parameters of the address to be measured;

[0148] A processing module 220 is configured to perform normalization processing on the measurement data of each type of feature parameter in the multi-source feature parameters to obtain the normalization result of each type of feature parameter;

[0149] An evaluation module 230 is configured to input the normalization result of each type of feature parameter into a radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-tested address output by the radiation calibration field site selection evaluation model;

[0150] The radiation calibration field site selection evaluation model is trained according to the address multi-source feature parameter data samples and the corresponding class labels; the multi-source feature parameters include atmospheric feature parameters, surface feature parameters, and social environment feature parameters.

[0151] The radiation calibration field site selection device described in this embodiment can be used to execute the above-mentioned radiation calibration field site selection method embodiment, and its principle and technical effect are similar, which will not be elaborated here.

[0152] The radiation calibration field site selection device of the embodiment of the present invention, by considering the influence of the calibration field regional environmental characteristics on the satellite remote sensing radiation calibration field site selection, fully excavates the multi-source features of the radiation calibration field from three aspects of the atmospheric environment, the surface environment, and the social environment, and through a deep learning algorithm, uses the address multi-source feature parameter data samples and the corresponding class labels to train the radiation calibration field site selection evaluation model, enabling the model to learn the prior knowledge and rules of site selection. Then, for the measurement data of the multi-source feature parameters of the to-be-tested address, its normalization result is input into the radiation calibration field site selection evaluation model, and the model can automatically predict and output the evaluation result of whether the to-be-tested address is a radiation calibration field, which can realize the intelligent site selection of the radiation calibration field. At the same time, the automated calculation process can greatly reduce the influence of manual intervention and subjective factors in the site selection process, which is beneficial to changing the dependence on expert qualitative selection in the traditional mode and improving the effect of radiation calibration field site selection.

[0153] Figure 3 FIG. illustrates a schematic physical structure diagram of an electronic device, such as Figure 3As shown in the figure, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340. Among them, the processor 310, the communications interface 320, and the memory 330 complete communication with each other through the communication bus 340. The processor 310 may call the logical instructions in the memory 330 to execute the radiation calibration field site selection method, which includes: obtaining measurement data of multi-source characteristic parameters of the address to be measured; normalizing the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain the normalization result of each type of the characteristic parameter; inputting the normalization result of each type of the characteristic parameter into the radiation calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model; the radiation calibration field site selection evaluation model is trained according to the address multi-source characteristic parameter data samples and the corresponding class labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters.

[0154] In addition, when the logical instructions in the above-mentioned memory 330 can be implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0155] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the radiation calibration field site selection method provided by each of the above methods. The method includes: obtaining measurement data of multi-source characteristic parameters of a to-be-measured address; performing normalization processing on the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain a normalization result of each type of the characteristic parameter; inputting the normalization result of each type of the characteristic parameter into a radiation calibration field site selection evaluation model to obtain an evaluation result of the to-be-measured address output by the radiation calibration field site selection evaluation model; the radiation calibration field site selection evaluation model is trained according to address multi-source characteristic parameter data samples and corresponding class labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters.

[0156] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the radiation calibration field site selection method provided by each of the above methods. The method includes: obtaining measurement data of multi-source characteristic parameters of a to-be-measured address; performing normalization processing on the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain a normalization result of each type of the characteristic parameter; inputting the normalization result of each type of the characteristic parameter into a radiation calibration field site selection evaluation model to obtain an evaluation result of the to-be-measured address output by the radiation calibration field site selection evaluation model; the radiation calibration field site selection evaluation model is trained according to address multi-source characteristic parameter data samples and corresponding class labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters.

[0157] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0158] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for selecting a location for a radiometric calibration field, characterized in that, Including: Obtaining measurement data of multi-source characteristic parameters of the address to be measured; Performing normalization processing on the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain the normalization result of each type of the characteristic parameters; Inputting the normalization result of each type of the characteristic parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model; The radiation calibration field site selection evaluation model is trained according to the multi-source characteristic parameter data samples of the address and the corresponding category labels; the multi-source characteristic parameters include atmospheric characteristic parameters, surface characteristic parameters, and social environment characteristic parameters; the atmospheric characteristic parameters include at least three of the following: surface albedo, aerosol optical depth, cloud cover, normalized difference vegetation index, surface humidity, and surface temperature; the surface characteristic parameters include at least three of the following: rainfall, surface type, distance to the main stream of the surrounding river, distance to the tributary of the surrounding river, distance to the surrounding dam, slope, land use information, and flood discharge area information; the social environment characteristic parameters include at least one of the following: traffic accessibility and population density; The performing normalization processing on the measurement data of each type of characteristic parameter in the multi-source characteristic parameters to obtain the normalization result of each type of the characteristic parameters includes: Respectively obtaining the thresholds of the atmospheric characteristic parameters, the thresholds of the surface characteristic parameters, and the thresholds of the social environment characteristic parameters; Based on the measurement data and thresholds of the atmospheric characteristic parameters, performing normalization processing on the measurement data of the atmospheric characteristic parameters to obtain the normalization result of the atmospheric characteristic parameters; Based on the measurement data and thresholds of the surface characteristic parameters, performing normalization processing on the measurement data of the surface characteristic parameters to obtain the normalization result of the surface characteristic parameters; Based on the measurement data and thresholds of the social environment characteristic parameters, performing normalization processing on the measurement data of the social environment characteristic parameters to obtain the normalization result of the social environment characteristic parameters; The radiation calibration field site selection evaluation model includes an improved artificial neural network model, a spatial feature extraction model, a temporal feature extraction model, a fully connected layer, and a classification model; the improved artificial neural network model includes an input layer and a hidden layer connected in sequence; The inputting the normalization result of each type of the characteristic parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the address to be measured output by the radiation calibration field site selection evaluation model includes: Inputting the normalization result of each type of the characteristic parameters into the improved artificial neural network model to obtain the integrated feature vector group corresponding to each type of the characteristic parameters output by the hidden layer of the improved artificial neural network model; Inputting the integrated feature vector group corresponding to each type of the characteristic parameters into the spatial feature extraction model to obtain the spatial feature vector group corresponding to each type of the characteristic parameters output by the spatial feature extraction model; Inputting the spatial feature vector group corresponding to each type of the characteristic parameters into the temporal feature extraction model to obtain the temporal feature vector group corresponding to each type of the characteristic parameters output by the temporal feature extraction model; Input the time feature vector groups corresponding to each type of the feature parameters into the fully connected layer to obtain the spatio-temporal feature vectors corresponding to each type of the feature parameters output by the fully connected layer; Input the spatio-temporal feature vectors corresponding to each type of the feature parameters into the classification model to obtain the evaluation result of the to-be-detected address output by the classification model.

2. The radiometric calibration site selection method according to claim 1, wherein The radiation calibration field site selection evaluation model adopts an artificial neural network model; before inputting the normalization results of each type of the feature parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-detected address output by the radiation calibration field site selection evaluation model, the method further includes: Perform normalization processing on the address multi-source feature parameter data samples to obtain the normalization results of the address multi-source feature parameter data samples; Use the normalization results of the address multi-source feature parameter data samples and the category labels corresponding to the address multi-source feature parameter data samples as a group of training samples, and obtain multiple groups of training samples; Initialize the connection weights of the artificial neural network model; Use the multiple groups of training samples to train the artificial neural network model, iteratively optimize the connection weights of the artificial neural network model until the number of training times reaches the maximum number of training times to obtain a trained artificial neural network model; Obtain the radiation calibration field site selection evaluation model according to the trained artificial neural network model.

3. A radiometric calibration field site selection device, characterized in that It includes: An acquisition module, configured to acquire measurement data of multi-source feature parameters of a to-be-detected address; A processing module, configured to perform normalization processing on the measurement data of each type of feature parameters in the multi-source feature parameters to obtain the normalization results of each type of the feature parameters; An evaluation module, configured to input the normalization results of each type of the feature parameters into a radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-detected address output by the radiation calibration field site selection evaluation model; The radiation calibration field site selection evaluation model is trained according to address multi-source feature parameter data samples and corresponding category labels; the multi-source feature parameters include atmospheric feature parameters, surface feature parameters, and social environment feature parameters; the atmospheric feature parameters include at least three of the following: surface albedo, aerosol optical thickness, cloud cover, normalized difference vegetation index, surface humidity, and surface temperature; the surface feature parameters include at least three of the following: rainfall, surface type, distance to the main surrounding river, distance to the surrounding river tributaries, distance to the surrounding dam, slope, land use information, and flood discharge area information; the social environment feature parameters include at least one of the following: traffic accessibility and population density; When the processing module performs normalization processing on the measurement data of each type of feature parameters in the multi-source feature parameters to obtain the normalization results of each type of the feature parameters, it includes: Respectively obtain the thresholds of the atmospheric feature parameters, the thresholds of the surface feature parameters, and the thresholds of the social environment feature parameters; Based on the measurement data and thresholds of the atmospheric feature parameters, perform normalization processing on the measurement data of the atmospheric feature parameters to obtain the normalization results of the atmospheric feature parameters; Normalize the measurement data of the surface feature parameters based on the measurement data and thresholds of the surface feature parameters to obtain the normalization result of the surface feature parameters; Normalize the measurement data of the social environment feature parameters based on the measurement data and thresholds of the social environment feature parameters to obtain the normalization result of the social environment feature parameters; The radiation calibration field site selection evaluation model includes an improved artificial neural network model, a spatial feature extraction model, a temporal feature extraction model, a fully connected layer, and a classification model; the improved artificial neural network model includes an input layer and a hidden layer connected in sequence; The evaluation module inputs the normalization results of each type of the feature parameters into the radiation calibration field site selection evaluation model to obtain the evaluation result of the to-be-tested address output by the radiation calibration field site selection evaluation model, including: Input the normalization results of each type of the feature parameters into the improved artificial neural network model to obtain an integrated feature vector group corresponding to each type of the feature parameters output by the hidden layer of the improved artificial neural network model; Input the integrated feature vector group corresponding to each type of the feature parameters into the spatial feature extraction model to obtain a spatial feature vector group corresponding to each type of the feature parameters output by the spatial feature extraction model; Input the spatial feature vector group corresponding to each type of the feature parameters into the temporal feature extraction model to obtain a temporal feature vector group corresponding to each type of the feature parameters output by the temporal feature extraction model; Input the temporal feature vector group corresponding to each type of the feature parameters into the fully connected layer to obtain a spatio-temporal feature vector corresponding to each type of the feature parameters output by the fully connected layer; Input the spatio-temporal feature vector corresponding to each type of the feature parameters into the classification model to obtain the evaluation result of the to-be-tested address output by the classification model.

4. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the radiation calibration field site selection method according to any one of claims 1 to 2.

5. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the radiation calibration field site selection method according to any one of claims 1 to 2.

Images