WO2021012898A1 - 基于人工智能的农业保险查勘方法及相关设备 - Google Patents

基于人工智能的农业保险查勘方法及相关设备 Download PDF

Info

Publication number
WO2021012898A1
WO2021012898A1 PCT/CN2020/098926 CN2020098926W WO2021012898A1 WO 2021012898 A1 WO2021012898 A1 WO 2021012898A1 CN 2020098926 W CN2020098926 W CN 2020098926W WO 2021012898 A1 WO2021012898 A1 WO 2021012898A1
Authority
WO
WIPO (PCT)
Prior art keywords
underwriting
subject
spectral
investigated
agricultural insurance
Prior art date
Application number
PCT/CN2020/098926
Other languages
English (en)
French (fr)
Inventor
王小山
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2021012898A1 publication Critical patent/WO2021012898A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/08—Insurance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/188—Vegetation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/58—Extraction of image or video features relating to hyperspectral data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB

Definitions

  • This application relates to the technical field of classification models, in particular to an artificial intelligence-based agricultural insurance survey method and related equipment.
  • Agricultural insurance refers to the insurance activity in which the insurance institution assumes the responsibility of indemnifying insurance money for property losses caused by the agreed natural disasters and other accidents of the insured person in the agricultural production process in accordance with the agricultural insurance contract.
  • the inventor realized that in order to avoid false underwriting, insurance institutions need to send survey personnel to conduct surveys to verify the authenticity of the underwriting targets provided by the insured.
  • the method of sending survey personnel to conduct surveys It is time-consuming and labor-intensive, and requires the cooperation of farmers, which is more difficult.
  • the main purpose of this application is to provide an artificial intelligence-based agricultural insurance survey method and related equipment, which aims to solve the difficult technical problem of sending survey personnel to survey, which is time-consuming and labor-intensive, and requires the cooperation of farmers.
  • this application provides an artificial intelligence-based agricultural insurance survey method, which includes the following steps:
  • control the airborne hyperspectral imaging spectrum system to collect the spectrum data of the underwriting subject to be investigated
  • the spectral characteristics of the underwriting subject to be investigated are input into the trained crop type identification model for analysis, so as to obtain which crop the underwriting subject to be investigated belongs to as a survey result.
  • this application also provides an artificial intelligence-based agricultural insurance survey device.
  • the artificial intelligence-based agricultural insurance survey device includes:
  • An acquisition module configured to control the airborne hyperspectral imaging spectroscopy system to collect the spectrum data of the underwriting subject to be investigated according to the geographic location information;
  • the preprocessing module is used to preprocess the spectrum data of the underwriting subject to be investigated
  • a spectral feature extraction module configured to extract the spectral features of the underwriting subject to be surveyed from the preprocessed spectral data of the underwriting subject to be surveyed;
  • the analysis module is used to input the spectral characteristics of the underwriting subject to be surveyed into the trained crop type identification model for analysis, to obtain which crop the underwriting subject to be surveyed belongs to as the survey result.
  • this application also provides an artificial intelligence-based agricultural insurance survey equipment.
  • the artificial intelligence-based agricultural insurance survey equipment includes a processor, a memory, and stored in the memory and can be used by the The artificial intelligence-based agricultural insurance survey program executed by the processor, wherein when the artificial intelligence-based agricultural insurance survey program is executed by the processor, the steps of the artificial intelligence-based agricultural insurance survey method are implemented.
  • this application also provides a computer-readable storage medium on which is stored an artificial intelligence-based agricultural insurance survey program, wherein the artificial intelligence-based agricultural insurance survey program is When the processor is executed, the steps of the above-mentioned artificial intelligence-based agricultural insurance survey method are realized.
  • This application provides an artificial intelligence-based agricultural insurance survey method and related equipment.
  • the geographic location information of the underwriting subject to be surveyed is first extracted from the agricultural insurance survey task, and then based on the extracted geographic location Information, control the airborne hyperspectral imaging spectroscopy system to collect the spectrum data of the underwriting subject to be investigated, and then preprocess the collected spectral data of the underwritten subject to be investigated, and extract the spectral characteristics of the underwritten subject to be investigated from the preprocessed spectral data, Input the extracted spectral features of the underwriting subject to be surveyed into the trained crop type recognition model for analysis, to obtain which crop the underwriting subject to be surveyed belongs to, as the survey result, realize the treatment based on the trained crop type recognition model
  • the analysis of survey underwriting targets can significantly improve the recognition accuracy of crop types and realize convenient, efficient, fast and accurate agricultural insurance surveys.
  • Figure 1 is a schematic diagram of the hardware structure of the artificial intelligence-based agricultural insurance survey equipment involved in the embodiment of the application;
  • FIG. 2 is a schematic flow chart of the first embodiment of the agricultural insurance survey method based on artificial intelligence in the application;
  • FIG. 3 is a schematic flowchart of a second embodiment of the application for an artificial intelligence-based agricultural insurance survey method
  • FIG. 4 is a schematic diagram of functional modules of the first embodiment of the agricultural insurance survey device based on artificial intelligence in this application.
  • the artificial intelligence-based agricultural insurance survey method involved in the embodiments of this application is mainly applied to artificial intelligence-based agricultural insurance survey equipment.
  • the artificial intelligence-based agricultural insurance survey equipment may be a personal computer (PC), a server, etc. with data Equipment for processing functions.
  • FIG 1 is a schematic diagram of the hardware structure of the artificial intelligence-based agricultural insurance survey equipment involved in the embodiment of the application.
  • the artificial intelligence-based agricultural insurance survey equipment may include a processor 1001 (for example, a central processing unit, a CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005.
  • the communication bus 1002 is used to realize the connection and communication between these components;
  • the user interface 1003 may include a display (Display), an input unit such as a keyboard (Keyboard);
  • the network interface 1004 may optionally include a standard wired interface, a wireless interface (Such as wireless fidelity WIreless-FIdelity, WI-FI interface);
  • the memory 1005 can be a high-speed random access memory (random access memory, RAM), or a stable memory (non-volatile memory), such as disk memory, memory
  • 1005 may also be a storage device independent of the foregoing processor 1001.
  • the hardware structure shown in FIG. 1 does not constitute a limitation to the present application, and may include more or less components than those shown in the figure, or combine certain components, or different component arrangements.
  • the memory 1005 as a computer storage medium in FIG. 1 may include an operating system, a network communication module, and an agricultural insurance survey program based on artificial intelligence.
  • the processor 1001 can call the artificial intelligence-based agricultural insurance survey program stored in the memory 1005, and execute the artificial intelligence-based agricultural insurance survey method provided by each embodiment of the present application.
  • the embodiment of the application provides an artificial intelligence-based agricultural insurance survey method.
  • FIG. 2 is a schematic flowchart of the first embodiment of the agricultural insurance survey method based on artificial intelligence in this application.
  • the artificial intelligence-based agricultural insurance survey method is implemented by an artificial intelligence-based agricultural insurance survey equipment.
  • the artificial intelligence-based agricultural insurance survey equipment may be a personal computer, a server, or other equipment with data processing functions.
  • the artificial intelligence-based agricultural insurance survey method includes the following steps:
  • Step S10 when the agricultural insurance survey task is received, extract the geographic location information of the underwriting subject to be surveyed from the agricultural insurance survey task;
  • remote sensing technology is used to identify the types of crops of the underwriting targets provided by the insured, so as to realize the survey of the underwriting targets to be surveyed.
  • Traditional multi-spectral remote sensing remote sensing sensors have few bands, low spectral resolution, and the effect of vegetation spectrum similarity, which cannot obtain high crop type identification accuracy.
  • Hyperspectral technology can detect subtle changes in crops in many narrow bands. Spectral differences can accurately distinguish the types of crops. Therefore, this embodiment uses a UAV equipped with a hyperspectral imaging spectrometer (the UAV and its equipped hyperspectral imaging spectrometer are defined as the airborne hyperspectral imaging spectroscopy system) to collect the pending survey underwriting Spectral data of the target.
  • a communication connection between the artificial intelligence-based agricultural insurance survey equipment and the airborne hyperspectral imaging spectroscopy system needs to be established in advance, so that the artificial intelligence-based agricultural insurance survey equipment can manage the airborne hyperspectral imaging spectroscopy system.
  • relevant staff of the insurance institution can trigger the corresponding agricultural insurance survey task on the artificial intelligence-based agricultural insurance survey equipment.
  • the agricultural insurance survey task carries the geography of the underwriting subject to be surveyed. Location information.
  • the geographic location information may specifically be geographic coordinates.
  • Step S20 controlling the airborne hyperspectral imaging spectroscopy system to collect the spectrum data of the underwriting subject to be investigated according to the geographic location information;
  • the preset height can be flexibly set according to actual conditions, and the number of spectra recorded in each direction is not limited.
  • Step S30 preprocessing the spectrum data of the underwriting subject to be investigated
  • the server needs to preprocess the spectral data of the underwritten subject to be surveyed, that is, use the Savitzky-Golay filter fitting method to first The spectral data is smoothed and denoised, and then the bands (1350 ⁇ 1400nm, 1800 ⁇ 1950nm, 2400 ⁇ 2500nm) that are strongly affected by water vapor absorption in the spectral data of the underwriting subject to be investigated are eliminated, and finally the spectral data of the underwriting subject to be investigated is averaged , Calculate the average of the 20 spectra that will be collected.
  • Step S40 extracting the spectral characteristics of the underwriting subject to be surveyed from the preprocessed spectral data of the underwriting subject to be surveyed;
  • spectral differentiation method uses the spectral differentiation method and the continuum removal method to analyze some significant reflection and absorption characteristics in the pre-processed spectral data of the underwriting subject to be investigated, so as to determine the spectral feature area from the pre-processed spectral data .
  • the spectral differentiation method mathematically simulates the reflection spectrum and obtains differential values of different orders to quickly determine the spectral bending point and extract the reflection and absorption peak parameters.
  • this embodiment adopts the first-order spectral differential, and its calculation formula is as follows:
  • ⁇ i is the band wavelength
  • ⁇ '( ⁇ i ) is the first derivative of the wavelength ⁇ i .
  • the continuum removal method is to convert the part of the reflection spectrum with strong absorption characteristics, amplify its absorption characteristics, and compare them on a common baseline, which is convenient to analyze and extract the spectral absorption characteristics.
  • the absorption depth (DEP) can be obtained through the spectrum after the removal of the continuum.
  • CR min is the minimum spectral reflectance after removing the continuum in an absorption valley
  • ⁇ a and ⁇ b are the wavelength values of the absorption start and end points, respectively.
  • spectral differentiation method and continuum removal method analyze the reflection and absorption characteristics of the pre-processed spectral data of the underwriting subject to be investigated, and find the green peak, red valley, red edge, and blue chlorophyll absorption bands in the visible to near-infrared range.
  • the chlorophyll absorption band, weak water absorption band, narrow water and oxygen absorption band, and water and carbon dioxide strong absorption band in the red light region are the main spectral characteristic regions for identifying crop types. Extract the green peak amplitude and green peak position from the green peak area, extract the red valley amplitude and red valley position from the red valley area, and extract the red edge amplitude and red edge position from the red edge area to obtain six differential features.
  • the absorption depth (DEP) and absorption area (DEP) and absorption area are extracted from the blue light zone chlorophyll absorption band, the red light zone chlorophyll absorption band, the weak water absorption band, the narrow water and oxygen absorption band, and the water and carbon dioxide strong absorption band (five absorption bands).
  • AREA ten absorption characteristics are obtained.
  • Step S50 Input the spectral characteristics of the underwriting subject to be surveyed into the trained crop type recognition model for analysis, to obtain which crop the underwriting subject to be surveyed belongs to as the survey result.
  • the sixteen extracted spectral features are input into the trained crop type recognition model for analysis.
  • the crop type recognition model is specifically a Back Propagation Neural Network (Back Propagation Neural Network) model.
  • the trained crop type recognition model has good stability and predictive ability, and there is only one prediction result. Input the sixteen extracted spectral features into the trained crop type recognition model to obtain the crop type code output by the crop type recognition model, that is, determine which crop the underwriting subject belongs to as the survey result.
  • This case provides an artificial intelligence-based agricultural insurance survey method.
  • the geographic location information of the underwriting subject to be investigated is first extracted from the agricultural insurance survey task, and then the machine is controlled based on the extracted geographic location information.
  • the hyperspectral imaging spectroscopy system collects the spectral data of the underwriting subject to be investigated, and then preprocesses the collected spectral data of the underwritten subject to be investigated, and extracts the spectral features of the underwritten subject to be investigated from the preprocessed spectral data, and extracts the
  • the spectral features of the survey underwriting targets are input into the trained crop type identification model for analysis to determine which crops to be surveyed underwriting targets belong to.
  • the implementation of the underwriting targets for the survey based on the trained crop type identification model Analysis can significantly improve the recognition accuracy of crop types, and realize convenient, efficient, fast and accurate agricultural insurance surveys.
  • the second embodiment of the agricultural insurance survey method based on artificial intelligence in this application is proposed based on the first embodiment.
  • the difference between the second embodiment of the artificial intelligence-based agricultural insurance survey method and the first embodiment of the artificial intelligence-based agricultural insurance survey method is that, referring to FIG. 3, before the step S10, it may include:
  • Step S60 obtaining spectral data of several crops, and pre-processing the spectral data of each crop respectively;
  • the crop type recognition model needs to be trained in advance. Specifically, first collect a large amount of spectral data of several common crops.
  • the airborne hyperspectral imaging spectroscopy system is used to measure the canopy spectra of six common crops of wheat, rice, corn, potatoes, soybeans, and rape.
  • Preprocess the spectrum of each crop that is, use the Savitzky-Golay filter fitting method to smooth and denoise the spectrum of each crop separately, and then remove the bands that are strongly affected by water vapor absorption in the spectrum of each crop , And finally average the spectrum of each crop, that is, take the average of the spectrum of each crop.
  • Step S70 extracting spectral features from the preprocessed spectral data of each crop respectively as training samples, and constructing a training sample set according to the training samples;
  • the spectral features of the training data are extracted from the spectral feature regions of the pre-processed spectral data of each crop to construct a training sample set, for example:
  • Training sample set ⁇ training sample 1, training sample 2, training sample 3, training sample 4, training sample 5, training sample 6 ⁇
  • Step S80 Train a crop type recognition model according to the training sample set to obtain a trained crop type recognition model.
  • the process of training the BPNN model is as follows:
  • BPNN model Create a BPNN model, set the BPNN model to three layers (input layer + hidden layer + output layer), set the number of input layer nodes (5 layers), the number of hidden layer nodes (optionally 17-25 layers), Number of output layer nodes (1 layer), hidden layer transfer function (tansig function), output layer transfer function (purelin function), number of training iterations (optionally 20000), learning efficiency in BPNN network (optionally 0.05) , The minimum error of training target (0.0001);
  • traincgf uses the traincgf method to iteratively train the constructed BPNN model, that is, input a training sample, find the output of each hidden layer and output layer forward, and then calculate the deviation between the output of the output layer and the correct output, which is the output
  • the error of the layer using the minimum square error to measure the error size
  • the gradient descent method is used to optimize the weight and bias to make the error smaller, which is the back propagation error.
  • the output layer is used as the input to obtain the reverse output of the output layer, and then the reverse output is combined with the connection weight as the reverse input of the hidden layer to obtain the reverse output of the hidden layer, and then based on the hidden layer Calculate the weight gradient connected between the input layer and the hidden layer (the input of the output layer is multiplied by the reverse output of the hidden layer).
  • the weight and bias of the BPNN model can be updated Set
  • the BPNN model after the BPNN model is trained, it is not put into use immediately, but the accuracy of the trained BPNN model is first tested.
  • preset verification data can be obtained
  • the spectral characteristics of the verification data can be extracted from the spectral feature area of the verification data to construct a verification sample set
  • the normalization function premnmx is used to normalize each test sample in the test sample set
  • each sample in the normalized test sample set is sequentially input into the trained BPNN model to obtain the crop type (that is, the prediction result) output by the trained BPNN model.
  • judge whether the trained BPNN model is accurate in identifying each crop according to the actual crop types and prediction results.
  • identification accuracy accurate identification number/corresponding test sample set
  • the embodiment of the application also provides an artificial intelligence-based agricultural insurance survey device.
  • Figure 4 is a schematic diagram of the functional modules of the first embodiment of the agricultural insurance surveying device based on artificial intelligence in this application.
  • the artificial intelligence-based agricultural insurance survey device includes:
  • the geographic location information extraction module 10 is used to extract the geographic location information of the underwriting subject to be investigated from the agricultural insurance investigation task when an agricultural insurance survey task is received;
  • the collection module 20 is configured to control the airborne hyperspectral imaging spectrum system to collect the spectrum data of the underwriting subject to be surveyed according to the geographic location information;
  • the preprocessing module 30 is used to preprocess the spectrum data of the underwriting subject to be investigated
  • the spectral feature extraction module 40 is configured to extract the spectral features of the underwriting subject to be surveyed from the preprocessed spectral data of the underwriting subject to be surveyed;
  • the analysis module 50 is configured to input the spectral characteristics of the underwriting subject to be surveyed into the trained crop type identification model for analysis to obtain which crop the underwriting subject to be surveyed belongs to as a survey result.
  • the virtual function modules of the above-mentioned artificial intelligence-based agricultural insurance surveying device are stored in the memory 1005 of the artificial intelligence-based agricultural insurance surveying equipment shown in FIG. 1, and are used to implement all functions of the artificial intelligence-based agricultural insurance surveying program;
  • each module is executed by the processor 1001, it realizes the analysis based on the trained crop type recognition model for the underwriting target of the survey, which can significantly improve the recognition accuracy of the crop type, and realizes convenient, efficient, fast and accurate agricultural insurance survey.
  • the collection module 20 includes:
  • the recording unit is used to control the probe of the airborne hyperspectral imaging spectroscopy system vertically downwards in the four directions of the geographic coordinates, and measure at the preset height of the canopy of the underwriting subject to be surveyed, in each direction A number of spectra are recorded on the top, which constitute the spectrum data of the underwriting subject to be investigated.
  • the pre-processing module 30 includes:
  • the smoothing and denoising processing unit is used to perform smoothing and denoising processing on the spectrum data of the underwriting subject to be investigated by using a filter fitting method;
  • the elimination unit is used to eliminate the strong water absorption band from the spectral data of the underwriting subject to be investigated after smoothing and denoising processing;
  • the averaging processing unit is used to perform averaging processing on the spectrum data of the underwriting subject to be investigated excluding the water intensity absorption band.
  • the spectral feature extraction module 40 includes:
  • a determining unit configured to analyze the preprocessed spectral data of the undercover subject to be investigated by using the spectral differentiation method and the continuum removal method, so as to determine the spectral characteristic region from the preprocessed spectral data of the underwritten subject under investigation;
  • the extraction unit is used to extract the differential feature and the absorption feature from the spectral feature area to obtain the spectral feature of the underwriting subject to be investigated.
  • the extraction unit includes:
  • a first extraction subunit configured to extract differential features from the green peak area, the red valley area, and the red edge area
  • the second extraction subunit is used to extract the chlorophyll absorption band from the blue light region, the chlorophyll absorption band in the red light region, the weak water absorption band, the water and oxygen narrow absorption band, and the water and carbon dioxide strong absorption band Extract absorption features.
  • the artificial intelligence-based agricultural insurance survey device further includes:
  • the acquisition module is used to acquire the spectral data of several crops, and preprocess the spectral data of each crop respectively;
  • the construction module is used to extract spectral features from the preprocessed spectral data of each crop as a training sample, and construct a training sample set according to the training sample;
  • the training module is used to train a crop type recognition model according to the training sample set to obtain a trained crop type recognition model.
  • the training module includes:
  • a normalization processing unit for performing normalization processing on each training sample in the training sample set
  • the initialization unit is used to create a crop type recognition model based on the back propagation neural network BPNN, and initialize the parameters of the crop type recognition model;
  • An obtaining unit configured to input the normalized training samples into the crop type recognition model to obtain forward output and reverse output;
  • the update unit is configured to use a gradient descent method to update the parameters of the crop type recognition model according to the forward output and the reverse output to obtain a trained crop type recognition model.
  • each module in the artificial intelligence-based agricultural insurance survey device corresponds to the steps in the above-mentioned artificial intelligence-based agricultural insurance survey method embodiment, and its functions and realization processes are not repeated here.
  • the embodiments of the present application also provide a computer-readable storage medium.
  • the computer-readable storage medium may be nonvolatile or volatile.
  • the computer-readable storage medium of this application stores an artificial intelligence-based agricultural insurance survey program, wherein when the artificial intelligence-based agricultural insurance survey program is executed by a processor, the steps of the above-mentioned artificial intelligence-based agricultural insurance survey method are realized .
  • the method implemented when the artificial intelligence-based agricultural insurance survey program is executed can refer to the various embodiments of the artificial intelligence-based agricultural insurance survey method of this application, which will not be repeated here.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Accounting & Taxation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Finance (AREA)
  • Molecular Biology (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Mining & Mineral Resources (AREA)
  • Primary Health Care (AREA)
  • Human Resources & Organizations (AREA)
  • Development Economics (AREA)
  • Technology Law (AREA)
  • Agronomy & Crop Science (AREA)
  • Tourism & Hospitality (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Evolutionary Biology (AREA)
  • Animal Husbandry (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

本申请属于分类模型技术领域，提供一种基于人工智能的农业保险查勘方法及相关设备，该方法包括：当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；对所述待查勘承保标的的光谱数据进行预处理；从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。

Description

基于人工智能的农业保险查勘方法及相关设备
优先权信息
本申请要求于2019年7月23日提交中国专利局、申请号为201910665932.0，发明名称为“基于人工智能的农业保险查勘方法及相关设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及分类模型技术领域，尤其涉及一种基于人工智能的农业保险查勘方法及相关设备。
背景技术
农业保险，是指保险机构根据农业保险合同，对参保人在农业生产过程中的因保险标的遭受约定的自然灾害等事故所造成的财产损失承担赔偿保险金责任的保险活动。目前，在农业保险的承保阶段，发明人意识到，为避免虚假承保，保险机构需要派遣查勘人员进行查勘，以核实参保人提供的承保标的的真实性，然而，派遣查勘人员进行查勘的方式，耗时耗力，还需要农户的配合，难度较大。
发明内容
本申请的主要目的在于提供一种基于人工智能的农业保险查勘方法及相关设备，旨在解决派遣查勘人员进行查勘的方式，耗时耗力，还需要农户的配合，难度较大的技术问题。
为实现上述目的，本申请提供一种基于人工智能的农业保险查勘方法，所述基于人工智能的农业保险查勘方法包括以下步骤：
当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
对所述待查勘承保标的的光谱数据进行预处理；
从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
此外，为实现上述目的，本申请还提供基于人工智能的农业保险查勘装置，所述基于 人工智能的农业保险查勘装置包括：
采集模块，用于根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
预处理模块，用于对所述待查勘承保标的的光谱数据进行预处理；
光谱特征提取模块，用于从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
分析模块，用于将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
此外，为实现上述目的，本申请还提供一种基于人工智能的农业保险查勘设备，所述基于人工智能的农业保险查勘设备包括处理器、存储器、以及存储在所述存储器上并可被所述处理器执行的基于人工智能的农业保险查勘程序，其中所述基于人工智能的农业保险查勘程序被所述处理器执行时，实现如上述的基于人工智能的农业保险查勘方法的步骤。
此外，为实现上述目的，本申请还提供一种计算机可读存储介质，所述计算机可读存储介质上存储有基于人工智能的农业保险查勘程序，其中所述基于人工智能的农业保险查勘程序被处理器执行时，实现如上述的基于人工智能的农业保险查勘方法的步骤。
本申请提供一种基于人工智能的农业保险查勘方法及相关设备，当接收到农业保险查勘任务时，首先从该农业保险查勘任务中提取待查勘承保标的的地理位置信息，然后根据提取的地理位置信息，控制机载高光谱成像光谱***采集待查勘承保标的光谱数据，再对采集的待查勘承保标的光谱数据进行预处理，并从预处理后的光谱数据中提取待查勘承保标的的光谱特征，将提取的待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出待查勘承保标的属于何种农作物，作为查勘结果，实现了基于训练好的农作物类型识别模型对待查勘承保标的进行分析，能显著提高农作物类型的识别精度，实现了便捷、高效、快速、准确地农业保险查勘。
附图说明
图1为本申请实施例方案中涉及的基于人工智能的农业保险查勘设备的硬件结构示意图；
图2为本申请基于人工智能的农业保险查勘方法第一实施例的流程示意图；
图3为本申请基于人工智能的农业保险查勘方法第二实施例的流程示意图；
图4为本申请基于人工智能的农业保险查勘装置第一实施例的功能模块示意图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请实施例涉及的基于人工智能的农业保险查勘方法主要应用于基于人工智能的农业保险查勘设备，该基于人工智能的农业保险查勘设备可以是个人计算机(personal computer，PC)、服务器等具有数据处理功能的设备。
参照图1，图1为本申请实施例方案中涉及的基于人工智能的农业保险查勘设备的硬件结构示意图。本申请实施例中，基于人工智能的农业保险查勘设备可以包括处理器1001(例如中央处理器Central Processing Unit，CPU)，通信总线1002，用户接口1003，网络接口1004，存储器1005。其中，通信总线1002用于实现这些组件之间的连接通信；用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)；网络接口1004可选的可以包括标准的有线接口、无线接口(如无线保真WIreless-FIdelity，WI-FI接口)；存储器1005可以是高速随机存取存储器(random access memory，RAM)，也可以是稳定的存储器(non-volatile memory)，例如磁盘存储器，存储器1005可选的还可以是独立于前述处理器1001的存储装置。本领域技术人员可以理解，图1中示出的硬件结构并不构成对本申请的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
继续参照图1，图1中作为一种计算机存储介质的存储器1005可以包括操作***、网络通信模块以及基于人工智能的农业保险查勘程序。在图1中，处理器1001可以调用存储器1005中存储的基于人工智能的农业保险查勘程序，并执行本申请各实施例提供的基于人工智能的农业保险查勘方法。
本申请实施例提供了一种基于人工智能的农业保险查勘方法。
参照图2，图2为本申请基于人工智能的农业保险查勘方法第一实施例的流程示意图。
本实施例中，该基于人工智能的农业保险查勘方法由基于人工智能的农业保险查勘设备实现，该基于人工智能的农业保险查勘设备可以是个人计算机、服务器等具有数据处理功能的设备，所述基于人工智能的农业保险查勘方法包括以下步骤：
步骤S10，当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
本实施例利用遥感技术对参保人提供的承保标的进行农作物类型识别，实现对待查勘承保标的的查勘。由于传统的多光谱遥感受遥感传感器波段少、光谱分辨率低、植被光谱相似性的影响，无法获得较高的农作物类型识别精度，而高光谱技术则能在众多窄波段范围探测农作物间细微的光谱差异，准确区分农作物类型，由此，本实施例利用无人机搭载高光谱成像光谱仪(将无人机及其搭载的高光谱成像光谱仪定义为机载高光谱成像光谱***)采集待查勘承保标的的光谱数据。具体地，需预先建立基于人工智能的农业保险查勘设备与机载高光谱成像光谱***之间的通信连接，以使基于人工智能的农业保险查勘设备对机载高光谱成像光谱***进行管理。
本实施例中，在农业保险的承保阶段，保险机构的相关工作人员可以在基于人工智能的农业保险查勘设备上触发相应的农业保险查勘任务，该农业保险查勘任务中携带待查勘承保标的的地理位置信息，该地理位置信息具体可以是地理坐标。当基于人工智能的农业保险查勘设备在接收到农业保险查勘任务时，首先从农业保险查勘任务中提取待查勘承保标的的地理坐标。
步骤S20,根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
之后，基于待查勘承保标的的地理坐标，分别在所述地理坐标的四个方向上，控制机载高光谱成像光谱***的探头垂直向下，在待查勘承保标的冠层的预设高度处进行测量，并在每个方向上记录若干条光谱，构成待查勘承保标的的光谱数据，其中，预设高度可以根据实际灵活设置，在每个方向上记录的光谱数量也不作限定。例如，可以分别在该地理坐标的四个方向，控制机载高光谱成像光谱***的探头垂直向下，在距离待查勘承保标的冠层1-2米处进行测量，在每个方向上记录5条光谱，得到20条光谱，作为待查勘承保标的的光谱数据。
步骤S30，对所述待查勘承保标的的光谱数据进行预处理；
由于采集的待查勘承保标的的光谱数据波段多、数据量大、信息冗余严重，服务器需要对待查勘承保标的的光谱数据进行预处理，即，利用Savitzky-Golay滤波拟合法先对待查勘承保标的的光谱数据进行平滑去噪处理，接着剔除待查勘承保标的的光谱数据中受水汽吸收影响强烈的波段(1350～1400nm、1800～1950nm、2400～2500nm),最后对待查勘承保标的的光谱数据做平均处理，即将采集的20条光谱求取平均值。
步骤S40，从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
对采集的光谱数据进行预处理之后，从预处理后的待查勘承保标的的光谱数据中提取光谱特征(包括微分特征和吸收特征)。即，首先利用光谱微分法和连续统去除法，对预处理后的待查勘承保标的的光谱数据中的一些显著的反射和吸收特征进行分析，以从预处理后的光谱数据中确定光谱特征区域。其中，光谱微分法通过数学模拟反射光谱和求取不同阶数微分值，以迅速确定光谱弯曲点，提取反射和吸收峰参数。考虑到实际应用中，一阶微分对噪声敏感值较低，本实施例采用一阶光谱微分，其计算式如下：
Figure PCTCN2020098926-appb-000001
其中，λ i为波段波长，ρ’(λ i)为波长λ i的一阶导数。
连续统去除法是将反射光谱吸收强烈的部分波段进行转换，放大其吸收特征，并在共同基线进行比较，便于分析和提取光谱吸收特征，通过连续统去除后的光谱可求取吸收深度(DEP)、吸收宽度(WID)和吸收面积(AREA)等特征参数，其公式如下：
DEP＝1-CR min
WID＝λ b-λ a
ARER＝DEP·WID
其中，CR min为一个吸收谷内连续统去除后最小光谱反射率，λ a、λ b分别为吸收起点、终点的波长值。
利用光谱微分法和连续统去除法，分析预处理后的待查勘承保标的的光谱数据的反射和吸收特征，发现在可见光至近红外波段的绿峰、红谷、红边、蓝光区叶绿素吸收带、红光区叶绿素吸收带、水弱吸收带、水和氧窄吸收带以及水和二氧化碳强吸收带为识别农作物类型的主要光谱特征区域。从绿峰区域提取绿峰幅值和绿峰位置，从红谷区域提取红谷幅值和红谷位置、从红边区域提取红边幅值和红边位置，得到六种微分特征。从蓝光区叶绿素吸收带、红光区叶绿素吸收带、水弱吸收带、水和氧窄吸收带以及水和二氧化碳强吸收带(五处吸收带)中分别提取吸收深度(DEP)和吸收面积(AREA)，得到十种吸收特征。
步骤S50，将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
在从预处理后的查勘承保标的的光谱数据中提取光谱特征之后，将提取的十六种光谱特征输入至训练好的农作物类型识别模型中进行分析。其中，农作物类型识别模型具体为反向传播神经网络模型(Back Propagation Neural Network)模型，训练好的农作物类型识别模型具有较好的稳定性和预测能力，预测结果只有一个。将提取的十六种光谱特征输入至训练好的农作物类型识别模型中，可得到农作物类型识别模型输出的农作物类型编码，也就是确定了待查勘承保标的属于何种农作物，作为查勘结果。
之后，将查勘结果与参保人提供的承保标的进行比对，判断二者是否一致，若一致，确认参保人提供的承保标的真实。
本案提供一种基于人工智能的农业保险查勘方法，当接收到农业保险查勘任务时，首先从该农业保险查勘任务中提取待查勘承保标的的地理位置信息，然后根据提取的地理位置信息，控制机载高光谱成像光谱***采集待查勘承保标的光谱数据，再对采集的待查勘承保标的光谱数据进行预处理，并从预处理后的光谱数据中提取待查勘承保标的的光谱特征，将提取的待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出待查勘承保标的属于何种农作物，作为查勘结果，实现了基于训练好的农作物类型识别模型对待查勘承保标的进行分析，能显著提高农作物类型的识别精度，实现了便捷、高效、快速、准确地农业保险查勘。
进一步地，基于第一实施例提出本申请基于人工智能的农业保险查勘方法的第二实施例。基于人工智能的农业保险查勘方法的第二实施例与基于人工智能的农业保险查勘方法的第一实施例的区别在于，参照图3，所述步骤S10之前，可以包括：
步骤S60，获取若干种农作物的光谱数据，并分别对每种农作物的光谱数据进行预处 理；
应当理解，在本实施例中，在步骤S10之前，需预先训练农作物类型识别模型。具体地，首先采集若干种常见农作物的大量光谱数据，在本实施例中，利用机载高光谱成像光谱***实测小麦、水稻、玉米、土豆、大豆、油菜六种常见农作物冠层光谱，再分别对每种农作物的光谱进行预处理，即，利用Savitzky-Golay滤波拟合法先对分别对每种农作物的光谱进行平滑去噪处理，接着分别剔除每种农作物的光谱中受水汽吸收影响强烈的波段，最后分别对每种农作物的光谱做平均处理，即分别对每种农作物的光谱求取平均值。
步骤S70，分别从预处理后的每种农作物的光谱数据中，提取光谱特征作为训练样本，并根据所述训练样本构建训练样本集；
进一步地，按照第一实施例中提取光谱特征的方式，分别从预处理后的每种农作物的光谱数据的光谱特征区域提取训练数据的光谱特征构建训练样本集，例如：
训练样本集＝{训练样本1，训练样本2，训练样本3，训练样本4，训练样本5，训练样本6}
＝{小麦的光谱特征，玉米的光谱特征，水稻的光谱特征，玉米的光谱特征，土豆的光谱特征，大豆的光谱特征，油菜的光谱特征}
步骤S80，根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型。
训练BPNN模型的过程如下：
a、采用归一化函数premnmx对训练样本集中的每一训练样本进行归一化处理，将归一化后的每一训练样本作为BPNN模型的输入，其对应的农作物类型编码作为正确输出；
b、创建BPNN模型，将BPNN模型设置为三层(输入层+隐含层+输出层)，设置输入层节点数(5层)、隐含层节点数(可选为17-25层)、输出层节点数(1层)、隐含层传递函数(tansig函数)、输出层传递函数(purelin函数)、训练迭代次数(可选为20000次)、BPNN网络中学习效率(可选为0.05)、训练目标最小误差(0.0001)；
c、初始化BPNN模型的参数，该参数包括权重和偏置(偏置可看作是每个神经元的自身权重)，即，将BPNN模型的权重和偏置初始化为来自于正态分布(0，1)的随机数；
d、采用traincgf方法对构建的BPNN模型进行迭代训练，即，输入一个训练样本，向前求出各个隐含层以及输出层的输出，再计算输出层的输出与正确输出的偏差，也就是输出层的误差(用最小方根差来衡量误差大小)，由于误差越小表示模型的精度越高，因此利用梯度下降法对权重和偏置进行优化使得误差变小，也就是反向传播误差，具体地，将输出层作为输入，得到输出层的反向输出，然后将反向输出与连接权重结合，作为隐含层的反向输入，得到隐含层的反向输出，然后基于隐含层的反向输出计算输入层与隐含层之间所连的权重梯度(输出层的输入乘以隐含层的反向输出)，有了权重梯度梯度之后，就可以更新BPNN模型的权重和偏置；
e、对于每个训练样本，循环d过程，直至其误差小于等于先前设置的训练目标最小误差或者已经达到迭代次数，即可获得训练好的BPNN模型。
在更多的实施中，BPNN模型训练好后，并不立即投入使用，而是先检测训练好的BPNN模型的精度。具体地，可以获取预设的验证数据，从验证数据的光谱特征区域提取验证数据的光谱特征构建验证样本集，采用归一化函数premnmx对测试样本集中的每一测试样本进行归一化处理，然后将归一化后的测试样本集中的每个样本，依次输入至训练好的BPNN模型中，得到训练好的BPNN模型输出的农作物类型(也就是预测结果)。之后，根据实际农作物类型和预测结果判断训练好的BPNN模型对每种农作物的识别是否准确，若准确，则将识别准确数量加1，然后根据公式：识别精度＝识别准确数量/测试样本集对应的农作物类型总数，计算训练好的BPNN模型的识别精度，然后将计算的识别精度与预设阈值(比如80％)进行比对，如果训练好的BPNN模型的识别精度大于预设阈值，则判定训练好的BPNN模型的识别精度满足条件，可以投入使用。
此外，本申请实施例还提供一种基于人工智能的农业保险查勘装置。
参照图4，图4为本申请基于人工智能的农业保险查勘装置第一实施例的功能模块示意图。
本实施例中，所述基于人工智能的农业保险查勘装置包括：
地理位置信息提取模块10，用于当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
采集模块20，用于根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
预处理模块30，用于对所述待查勘承保标的的光谱数据进行预处理；
光谱特征提取模块40，用于从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
分析模块50，用于将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
其中，上述基于人工智能的农业保险查勘装置的各虚拟功能模块存储于图1所示基于人工智能的农业保险查勘设备的存储器1005中，用于实现基于人工智能的农业保险查勘程序的所有功能；各模块被处理器1001执行时，实现了基于训练好的农作物类型识别模型对待查勘承保标的进行分析，能显著提高农作物类型的识别精度，实现了便捷、高效、快速、准确地农业保险查勘。
进一步的，所述采集模块20包括：
记录单元，用于分别在所述地理坐标的四个方向上，控制机载高光谱成像光谱***的探头垂直向下，在待查勘承保标的冠层的预设高度处进行测量，在每个方向上记录若干条光谱，构成所述待查勘承保标的的光谱数据。
进一步的，所述预处理模块30包括：
平滑去噪处理单元，用于采用滤波拟合法对所述待查勘承保标的的光谱数据进行平滑去噪处理；
剔除单元，用于从平滑去噪处理后的所述待查勘承保标的的光谱数据中，剔除水强吸收波段；
平均处理单元，用于对剔除了水强吸收波段的所述待查勘承保标的的光谱数据做平均处理。
进一步的，所述光谱特征提取模块40包括：
确定单元，用于利用光谱微分法和连续统去除法分析预处理后的所述待查勘承保标的的光谱数据，以从预处理后的所述待查勘承保标的的光谱数据中确定光谱特征区域；
提取单元，用于从所述光谱特征区域中提取微分特征和吸收特征，得到待查勘承保标的的光谱特征。
进一步的，所述提取单元包括：
第一提取子单元，用于从所述绿峰区域、所述红谷区域和所述红边区域提取微分特征；
第二提取子单元，用于从所述蓝光区叶绿素吸收带、所述红光区叶绿素吸收带、所述水弱吸收带、所述水和氧窄吸收带以及所述水和二氧化碳强吸收带提取吸收特征。
进一步的，所述基于人工智能的农业保险查勘装置还包括：
获取模块，用于获取若干种农作物的光谱数据，并分别对每种农作物的光谱数据进行预处理；
构建模块，用于分别从预处理后的每种农作物的光谱数据中，提取光谱特征作为训练样本，并根据所述训练样本构建训练样本集；
训练模块，用于根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型。
进一步的，所述训练模块包括：
归一化处理单元，用于对所述训练样本集中的每一训练样本进行归一化处理；
初始化单元，用于创建基于反向传播神经网络BPNN的农作物类型识别模型，并初始化所述农作物类型识别模型的参数；
获取单元，用于将归一化处理后的训练样本输入至所述农作物类型识别模型中，获取前向输出和反向输出；
更新单元，用于采用梯度下降法，根据所述前向输出和反向输出更新所述农作物类型识别模型的参数，得到训练好的农作物类型识别模型。
其中，上述基于人工智能的农业保险查勘装置中各个模块的功能实现与上述基于人工智能的农业保险查勘方法实施例中各步骤相对应，其功能和实现过程在此处不再一一赘述。
此外，本申请实施例还提供一种计算机可读存储介质，所述计算机可读存储介质可以是非易失性，也可以是易失性。
本申请计算机可读存储介质上存储有基于人工智能的农业保险查勘程序，其中所述基于人工智能的农业保险查勘程序被处理器执行时，实现如上述的基于人工智能的农业保险查勘方法的步骤。
其中，基于人工智能的农业保险查勘程序被执行时所实现的方法可参照本申请基于人工智能的农业保险查勘方法的各个实施例，此处不再赘述。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者***不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者***所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者***中还存在另外的相同要素。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (22)

  1. 一种基于人工智能的农业保险查勘方法，其中，所述基于人工智能的农业保险查勘方法包括以下步骤：
    当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
    根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
    对所述待查勘承保标的的光谱数据进行预处理；
    从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
    将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
  2. 如权利要求1所述的基于人工智能的农业保险查勘方法，其中，所述地理位置信息为地理坐标，
    所述根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据的步骤包括:
    分别在所述地理坐标的四个方向上，控制机载高光谱成像光谱***的探头垂直向下，在待查勘承保标的冠层的预设高度处进行测量，在每个方向上记录若干条光谱，构成所述待查勘承保标的的光谱数据。
  3. 如权利要求2所述的基于人工智能的农业保险查勘方法，其中，所述对所述待查勘承保标的的光谱数据进行预处理的步骤包括：
    采用滤波拟合法对所述待查勘承保标的的光谱数据进行平滑去噪处理；
    从平滑去噪处理后的所述待查勘承保标的的光谱数据中，剔除水强吸收波段；
    对剔除了水强吸收波段的所述待查勘承保标的的光谱数据做平均处理。
  4. 如权利要求3所述的基于人工智能的农业保险查勘方法，其中，所述从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征的步骤包括：
    利用光谱微分法和连续统去除法分析预处理后的所述待查勘承保标的的光谱数据，以从预处理后的所述待查勘承保标的的光谱数据中确定光谱特征区域；
    从所述光谱特征区域中提取微分特征和吸收特征，得到待查勘承保标的的光谱特征。
  5. 如权利要求4所述的基于人工智能的农业保险查勘方法，其中，所述光谱特征区域包括绿峰区域、红谷区域、红边区域、蓝光区叶绿素吸收带、红光区叶绿素吸收带、水弱吸收带、水和氧窄吸收带以及水和二氧化碳强吸收带，
    所述从所述光谱特征区域中提取微分特征和吸收特征的步骤包括：
    从所述绿峰区域、所述红谷区域和所述红边区域提取微分特征；
    从所述蓝光区叶绿素吸收带、所述红光区叶绿素吸收带、所述水弱吸收带、所述水和氧窄吸收带以及所述水和二氧化碳强吸收带提取吸收特征。
  6. 如权利要求1所述的基于人工智能的农业保险查勘方法，其中，所述当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息的步骤之前，包括：
    获取若干种农作物的光谱数据，并分别对每种农作物的光谱数据进行预处理；
    分别从预处理后的每种农作物的光谱数据中，提取光谱特征作为训练样本，并根据所述训练样本构建训练样本集；
    根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型。
  7. 如权利要求6所述的基于人工智能的农业保险查勘方法，其中，所述根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型的步骤包括：
    对所述训练样本集进行归一化处理；
    创建基于反向传播神经网络BPNN的农作物类型识别模型，并初始化所述农作物类型识别模型的参数；
    将归一化处理后的训练样本集输入至所述农作物类型识别模型中，获取前向输出和反向输出；
    采用梯度下降法，根据所述前向输出和反向输出更新所述农作物类型识别模型的参数，得到训练好的农作物类型识别模型。
  8. 一种基于人工智能的农业保险查勘装置，其中，所述基于人工智能的农业保险查勘装置包括：
    地理位置信息提取模块，用于当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
    采集模块，用于根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
    预处理模块，用于对所述待查勘承保标的的光谱数据进行预处理；
    光谱特征提取模块，用于从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
    分析模块，用于将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
  9. 一种基于人工智能的农业保险查勘设备，其中，所述基于人工智能的农业保险查勘设备包括处理器、存储器、以及存储在所述存储器上并可被所述处理器执行的基于人工智能的农业保险查勘程序，其中所述基于人工智能的农业保险查勘程序被所述处理器执行时，实现如下步骤：
    当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
    根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
    对所述待查勘承保标的的光谱数据进行预处理；
    从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
    将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
  10. 如权利要求9所述的基于人工智能的农业保险查勘设备，其中，所述地理位置信息为地理坐标，
    所述根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据的步骤包括：
    分别在所述地理坐标的四个方向上，控制机载高光谱成像光谱***的探头垂直向下，在待查勘承保标的冠层的预设高度处进行测量，在每个方向上记录若干条光谱，构成所述待查勘承保标的的光谱数据。
  11. 如权利要求10所述的基于人工智能的农业保险查勘设备，其中，所述对所述待查勘承保标的的光谱数据进行预处理的步骤包括：
    采用滤波拟合法对所述待查勘承保标的的光谱数据进行平滑去噪处理；
    从平滑去噪处理后的所述待查勘承保标的的光谱数据中，剔除水强吸收波段；
    对剔除了水强吸收波段的所述待查勘承保标的的光谱数据做平均处理。
  12. 如权利要求11所述的基于人工智能的农业保险查勘设备，其中，所述从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征的步骤包括：
    利用光谱微分法和连续统去除法分析预处理后的所述待查勘承保标的的光谱数据，以从预处理后的所述待查勘承保标的的光谱数据中确定光谱特征区域；
    从所述光谱特征区域中提取微分特征和吸收特征，得到待查勘承保标的的光谱特征。
  13. 如权利要求12所述的基于人工智能的农业保险查勘设备，其中，所述光谱特征区域包括绿峰区域、红谷区域、红边区域、蓝光区叶绿素吸收带、红光区叶绿素吸收带、水弱吸收带、水和氧窄吸收带以及水和二氧化碳强吸收带，
    所述从所述光谱特征区域中提取微分特征和吸收特征的步骤包括：
    从所述绿峰区域、所述红谷区域和所述红边区域提取微分特征；
    从所述蓝光区叶绿素吸收带、所述红光区叶绿素吸收带、所述水弱吸收带、所述水和氧窄吸收带以及所述水和二氧化碳强吸收带提取吸收特征。
  14. 如权利要求9所述的基于人工智能的农业保险查勘设备，其中，所述当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息的步骤之前，包括：
    获取若干种农作物的光谱数据，并分别对每种农作物的光谱数据进行预处理；
    分别从预处理后的每种农作物的光谱数据中，提取光谱特征作为训练样本，并根据所述训练样本构建训练样本集；
    根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型。
  15. 如权利要求14所述的基于人工智能的农业保险查勘设备，其中，所述根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型的步骤包括：
    对所述训练样本集进行归一化处理；
    创建基于反向传播神经网络BPNN的农作物类型识别模型，并初始化所述农作物类型识别模型的参数；
    将归一化处理后的训练样本集输入至所述农作物类型识别模型中，获取前向输出和反向输出；
    采用梯度下降法，根据所述前向输出和反向输出更新所述农作物类型识别模型的参数，得到训练好的农作物类型识别模型。
  16. 一种计算机可读存储介质，其中，所述计算机可读存储介质上存储有基于人工智能的农业保险查勘程序，其中所述基于人工智能的农业保险查勘程序被处理器执行时，实现如下步骤：
    当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息；
    根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据；
    对所述待查勘承保标的的光谱数据进行预处理；
    从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征；
    将所述待查勘承保标的的光谱特征输入至训练好的农作物类型识别模型中进行分析，以得出所述待查勘承保标的属于何种农作物，作为查勘结果。
  17. 如权利要求16所述的计算机可读存储介质，其中，所述地理位置信息为地理坐标，
    所述根据所述地理位置信息，控制机载高光谱成像光谱***采集所述待查勘承保标的的光谱数据的步骤包括：
    分别在所述地理坐标的四个方向上，控制机载高光谱成像光谱***的探头垂直向下，在待查勘承保标的冠层的预设高度处进行测量，在每个方向上记录若干条光谱，构成所述待查勘承保标的的光谱数据。
  18. 如权利要求17所述的计算机可读存储介质，其中，所述对所述待查勘承保标的的光谱数据进行预处理的步骤包括：
    采用滤波拟合法对所述待查勘承保标的的光谱数据进行平滑去噪处理；
    从平滑去噪处理后的所述待查勘承保标的的光谱数据中，剔除水强吸收波段；
    对剔除了水强吸收波段的所述待查勘承保标的的光谱数据做平均处理。
  19. 如权利要求18所述的计算机可读存储介质，其中，所述从预处理后的所述待查勘承保标的的光谱数据中，提取所述待查勘承保标的的光谱特征的步骤包括：
    利用光谱微分法和连续统去除法分析预处理后的所述待查勘承保标的的光谱数据，以从预处理后的所述待查勘承保标的的光谱数据中确定光谱特征区域；
    从所述光谱特征区域中提取微分特征和吸收特征，得到待查勘承保标的的光谱特征。
  20. 如权利要求19所述的计算机可读存储介质，其中，所述光谱特征区域包括绿峰区域、红谷区域、红边区域、蓝光区叶绿素吸收带、红光区叶绿素吸收带、水弱吸收带、水和氧窄吸收带以及水和二氧化碳强吸收带，
    所述从所述光谱特征区域中提取微分特征和吸收特征的步骤包括：
    从所述绿峰区域、所述红谷区域和所述红边区域提取微分特征；
    从所述蓝光区叶绿素吸收带、所述红光区叶绿素吸收带、所述水弱吸收带、所述水和氧窄吸收带以及所述水和二氧化碳强吸收带提取吸收特征。
  21. 如权利要求16所述的计算机可读存储介质，其中，所述当接收到农业保险查勘任务时，从所述农业保险查勘任务中提取待查勘承保标的的地理位置信息的步骤之前，包括：
    获取若干种农作物的光谱数据，并分别对每种农作物的光谱数据进行预处理；
    分别从预处理后的每种农作物的光谱数据中，提取光谱特征作为训练样本，并根据所述训练样本构建训练样本集；
    根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型。
  22. 如权利要求21所述的计算机可读存储介质，其中，所述根据所述训练样本集训练农作物类型识别模型，得到训练好的农作物类型识别模型的步骤包括：
    对所述训练样本集进行归一化处理；
    创建基于反向传播神经网络BPNN的农作物类型识别模型，并初始化所述农作物类型识别模型的参数；
    将归一化处理后的训练样本集输入至所述农作物类型识别模型中，获取前向输出和反向输出；
    采用梯度下降法，根据所述前向输出和反向输出更新所述农作物类型识别模型的参数，得到训练好的农作物类型识别模型。
PCT/CN2020/098926 2019-07-23 2020-06-29 基于人工智能的农业保险查勘方法及相关设备 WO2021012898A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910665932.0A CN110502997A (zh) 2019-07-23 2019-07-23 基于人工智能的农业保险查勘方法及相关设备
CN201910665932.0 2019-07-23

Publications (1)

Publication Number Publication Date
WO2021012898A1 true WO2021012898A1 (zh) 2021-01-28

Family

ID=68586657

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/098926 WO2021012898A1 (zh) 2019-07-23 2020-06-29 基于人工智能的农业保险查勘方法及相关设备

Country Status (2)

Country Link
CN (1) CN110502997A (zh)
WO (1) WO2021012898A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222039A (zh) * 2021-05-24 2021-08-06 中国平安财产保险股份有限公司 遥感影像标的查勘模型的训练方法、装置、设备及介质
CN114550108A (zh) * 2022-04-26 2022-05-27 广东省农业科学院植物保护研究所 一种草地贪夜蛾的识别预警方法及***
CN114694043A (zh) * 2022-03-30 2022-07-01 中国人民解放军空军军医大学 复杂场景下机载多光谱多域优选特征的地面伤员识别方法、装置及介质
CN117523433A (zh) * 2023-11-27 2024-02-06 广东华茂高科种业有限公司 一种水稻不同生长周期的叶绿素含量监测方法
CN117689481A (zh) * 2024-02-04 2024-03-12 国任财产保险股份有限公司 一种基于无人机视频数据自然灾害保险处理方法及***

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110502997A (zh) * 2019-07-23 2019-11-26 平安科技（深圳）有限公司 基于人工智能的农业保险查勘方法及相关设备
CN117538658A (zh) * 2023-11-16 2024-02-09 深圳市美信检测技术股份有限公司 基于红外光谱及热成像的人工智能故障定位方法及装置
CN117315492B (zh) * 2023-11-29 2024-04-02 中国平安财产保险股份有限公司 基于无人机技术的种植险预警方法、***、设备及介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090164256A1 (en) * 2007-12-20 2009-06-25 International Business Machines Device, system, and method of collaborative insurance
CN107169018A (zh) * 2017-04-06 2017-09-15 河南云保遥感科技有限公司 一种农业保险查勘、定损***及其实现方法
CN107909492A (zh) * 2017-11-30 2018-04-13 河南云保遥感科技有限公司 一种利用机器学习技术的农业保险查勘定损方法
CN108053325A (zh) * 2017-11-30 2018-05-18 河南云保遥感科技有限公司 一种基于农作物遥感技术的农业保险定损方法
CN109063660A (zh) * 2018-08-09 2018-12-21 成都天地量子科技有限公司 一种基于多光谱卫星影像的作物识别方法
CN110502997A (zh) * 2019-07-23 2019-11-26 平安科技（深圳）有限公司 基于人工智能的农业保险查勘方法及相关设备

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10664702B2 (en) * 2016-12-30 2020-05-26 International Business Machines Corporation Method and system for crop recognition and boundary delineation

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090164256A1 (en) * 2007-12-20 2009-06-25 International Business Machines Device, system, and method of collaborative insurance
CN107169018A (zh) * 2017-04-06 2017-09-15 河南云保遥感科技有限公司 一种农业保险查勘、定损***及其实现方法
CN107909492A (zh) * 2017-11-30 2018-04-13 河南云保遥感科技有限公司 一种利用机器学习技术的农业保险查勘定损方法
CN108053325A (zh) * 2017-11-30 2018-05-18 河南云保遥感科技有限公司 一种基于农作物遥感技术的农业保险定损方法
CN109063660A (zh) * 2018-08-09 2018-12-21 成都天地量子科技有限公司 一种基于多光谱卫星影像的作物识别方法
CN110502997A (zh) * 2019-07-23 2019-11-26 平安科技（深圳）有限公司 基于人工智能的农业保险查勘方法及相关设备

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222039A (zh) * 2021-05-24 2021-08-06 中国平安财产保险股份有限公司 遥感影像标的查勘模型的训练方法、装置、设备及介质
CN113222039B (zh) * 2021-05-24 2023-09-05 中国平安财产保险股份有限公司 遥感影像标的查勘模型的训练方法、装置、设备及介质
CN114694043A (zh) * 2022-03-30 2022-07-01 中国人民解放军空军军医大学 复杂场景下机载多光谱多域优选特征的地面伤员识别方法、装置及介质
CN114550108A (zh) * 2022-04-26 2022-05-27 广东省农业科学院植物保护研究所 一种草地贪夜蛾的识别预警方法及***
CN114550108B (zh) * 2022-04-26 2022-07-08 广东省农业科学院植物保护研究所 一种草地贪夜蛾的识别预警方法及***
CN117523433A (zh) * 2023-11-27 2024-02-06 广东华茂高科种业有限公司 一种水稻不同生长周期的叶绿素含量监测方法
CN117523433B (zh) * 2023-11-27 2024-04-19 广东华茂高科种业有限公司 一种水稻不同生长周期的叶绿素含量监测方法
CN117689481A (zh) * 2024-02-04 2024-03-12 国任财产保险股份有限公司 一种基于无人机视频数据自然灾害保险处理方法及***
CN117689481B (zh) * 2024-02-04 2024-04-19 国任财产保险股份有限公司 一种基于无人机视频数据自然灾害保险处理方法及***

Also Published As

Publication number Publication date
CN110502997A (zh) 2019-11-26

Similar Documents

Publication Publication Date Title
WO2021012898A1 (zh) 基于人工智能的农业保险查勘方法及相关设备
Bock et al. From visual estimates to fully automated sensor-based measurements of plant disease severity: status and challenges for improving accuracy
Qiao et al. UAV-based chlorophyll content estimation by evaluating vegetation index responses under different crop coverages
Chen et al. Mapping croplands, cropping patterns, and crop types using MODIS time-series data
Lowe et al. Hyperspectral image analysis techniques for the detection and classification of the early onset of plant disease and stress
Delalieux et al. Hyperspectral indices to diagnose leaf biotic stress of apple plants, considering leaf phenology
Santoso et al. Classifying the severity of basal stem rot disease in oil palm plantations using WorldView-3 imagery and machine learning algorithms
CN111091052A (zh) 一种基于最大似然法的玉米倒伏面积提取***及方法
Jiang et al. Assessing mangrove leaf traits under different pest and disease severity with hyperspectral imaging spectroscopy
Sibanda et al. Testing the capabilities of the new WorldView-3 space-borne sensor’s red-edge spectral band in discriminating and mapping complex grassland management treatments
Chen et al. Mapping wheat rust based on high spatial resolution satellite imagery
Huang et al. Identification of fusarium head blight in winter wheat ears based on fisher’s linear discriminant analysis and a support vector machine
Shu et al. Improving the estimation accuracy of SPAD values for maize leaves by removing UAV hyperspectral image backgrounds
Adjorlolo et al. Predicting C3 and C4 grass nutrient variability using in situ canopy reflectance and partial least squares regression
Zhang et al. Development, evaluation and application of a spectral knowledge base to detect yellow rust in winter wheat
Sarrazin et al. Fusing small-footprint waveform LiDAR and hyperspectral data for canopy-level species classification and herbaceous biomass modeling in savanna ecosystems
Huang et al. Detection of scab in wheat ears using in situ hyperspectral data and support vector machine optimized by genetic algorithm
Colorado et al. A novel NIR-image segmentation method for the precise estimation of above-ground biomass in rice crops
Zhou et al. Wheat phenology detection with the methodology of classification based on the time-series UAV images
Guan et al. An improved approach to estimating crop lodging percentage with Sentinel-2 imagery using machine learning
Hu et al. Fractional order modeling and recognition of nitrogen content level of rubber tree foliage
CN114199800A (zh) 一种水稻纹枯病识别方法、***、设备及介质
Qiu et al. Detection of the 3D temperature characteristics of maize under water stress using thermal and RGB-D cameras
Lu et al. Inversion of chlorophyll content under the stress of leaf mite for jujube based on model PSO-ELM method
Meivel et al. Monitoring of potato crops based on multispectral image feature extraction with vegetation indices

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20843933

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20843933

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20843933

Country of ref document: EP

Kind code of ref document: A1