Crop remote sensing classification method and system
The hyperspectral crop images of farmland are obtained through satellites and pre-processed and dimensionality reduction. Combined with the Support Vector Machine (SVM) classification model, the problems of low remote sensing classification accuracy and data redundancy in the existing technology are solved, and efficient and accurate crop classification is achieved.
Patent Information
- Application Number
- CN202411938765.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-26
- Publication Date
- 2025-04-18
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art has problems in crop remote sensing classification with low spatial resolution, short band range and high data acquisition cost, resulting in low classification accuracy and data redundancy.
Satellites are used to obtain hyperspectral farm crop images and pre-process them through de-cloud processing, radiation correction, image fusion and image stitching. Then, the hyperspectral data is reduced in dimension, redundant data is removed, and a support vector machine (SVM) classification model is built for training to achieve fine classification of crops.
It improves the accuracy and efficiency of crop classification, reduces the amount of data calculation, improves the computing efficiency, and can more accurately identify crop species.
Smart Images
Figure CN119851154A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of agricultural remote sensing, and in particular relates to a crop remote sensing classification method and system. Background Art
[0002] Obtaining detailed information on crops helps the country understand the growth of crops, estimate crop yields, and obtain crop disaster data in a timely manner. The main way to obtain detailed information on crops is through field surveys by local agricultural personnel. This type of survey is time-consuming and prone to omissions, making it difficult to obtain accurate data. With the development of space technology, detailed information on ground crops can be obtained through space satellite technology.
[0003] Most general space satellite remote sensing technologies use multispectral images for fine classification. However, multispectral images have many disadvantages such as low spatial resolution, short band range and small number of bands. At this time, there are great limitations in classifying crops in small areas based on multispectral images. If high spatial resolution remote sensing images are used, the data acquisition cost will be high. Therefore, hyperspectral remote sensing images with a wide wavelength range, a large number of bands and complete spectral curves can be introduced as image data for fine classification of crops, thereby providing strong support for fine management of crops. Summary of the invention
[0004] The purpose of the present invention is to provide a crop remote sensing classification method and system to solve the problems raised in the above background technology.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] A crop remote sensing classification method comprises the following steps:
[0007] S1. Data acquisition: using satellites to obtain high-resolution and hyperspectral remote sensing image data of farmland crops, and using drones to collect image data of different types of farmland crops in the farmland area;
[0008] S2, preprocessing and correction, preprocessing and correction of the acquired farmland crop remote sensing image data;
[0009] S3, data dimensionality reduction processing, dimensionality reduction processing of hyperspectral data, and removal of redundant data;
[0010] S4. Build a crop classification learning model for training, retain the optimal training model, and use it for crop identification and classification of satellite remote sensing images.
[0011] Preferably, in S1, a plurality of marking points are selected by a drone, labels corresponding to the types of farmland crops are produced, and a data set is established, and the data set is divided into a training set and a test set in a ratio of 8:2.
[0012] Preferably, in S2, the preprocessing of farmland crop remote sensing image data includes cloud removal, radiation correction, image fusion and image stitching.
[0013] Preferably, in S2, the central wavelength and half-width of the wavelength of the imaging spectrometer are used in a radiation calibration system designed in a MATLAB environment to complete the radiation correction from the image DN value to the surface reflectivity.
[0014] Preferably, in S2, the hyperspectral data captured by the satellite imager are divided into hyperspectral panchromatic images and hyperspectral cube images, the hyperspectral panchromatic images and the hyperspectral cube images are fused and processed by Pansharpen algorithm to obtain sub-band images extracted by IDL programming, Photoscan is used to process the sub-band images and hyperspectral panchromatic images extracted by IDL programming, the sub-band images are merged by IDL programming, the images are aligned in combination with the measured GPS points, vector data are introduced for splicing and then fused, and the hyperspectral data of the experimental area is output.
[0015] Preferably, in S3, the minimum noise separation MNF variation is selected to perform dimensionality reduction processing on the hyperspectral image data, and the first 20 MNF principal components are selected as the spectral features after the data dimensionality reduction.
[0016] Preferably, in S4, a support vector machine SVM is selected as the crop classification learning model, a radial basis kernel function is selected as the kernel function, and a training set is introduced to train the crop classification learning model.
[0017] Preferably, in S4, the test set is introduced into the crop classification learning model for testing, the overall classification accuracy and Kappa coefficient indicators are used to evaluate the crop classification learning model, an evaluation threshold is set, and the crop classification learning model training is repeated until the overall classification accuracy and Kappa coefficient indicators are higher than the evaluation threshold, thereby completing the construction of the crop classification learning model.
[0018] A crop remote sensing classification system, using a crop remote sensing classification method, specifically includes:
[0019] A data acquisition module, wherein the data acquisition processing module is used to acquire satellite farmland crop remote sensing image data and drone farmland crop image data;
[0020] A data processing module, which is used to process the acquired farmland crop remote sensing image data and perform dimensionality reduction processing on the hyperspectral data;
[0021] A classification model is used to classify and identify crops based on the collected data.
[0022] Compared with the prior art, the present invention has the following beneficial effects:
[0023] The present invention uses satellites to acquire hyperspectral images of farmland crops, and then uses cloud removal, radiation correction, image fusion and image stitching to process the hyperspectral images of farmland crops. The processed hyperspectral data is then subjected to dimensionality reduction processing to remove redundant information, thereby reducing the amount of data calculation and improving computing efficiency. The crop classification learning model uses an SVM classifier with high classification accuracy and strong generalization ability. Compared with general multispectral crop classification models, the classification model combined with dimensionality reduction processing of hyperspectral image data can greatly improve the classification accuracy and accurately identify the types of crops. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is a frame diagram of the crop remote sensing classification method of the present invention;
[0025] Figure 2 This is a block diagram of the crop remote sensing classification system of the present invention. DETAILED DESCRIPTION
[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0027] Example:
[0028] See also Figure 1 - Figure 2 As shown, a crop remote sensing classification method comprises the following steps:
[0029] S1. Data acquisition: Use satellites to acquire high-resolution and hyperspectral remote sensing image data of farmland crops. Use drones to collect image data of different types of farmland crops in the farmland area. When using satellites to acquire high-resolution and hyperspectral remote sensing image data of farmland crops, choose weather with few clouds and little wind as much as possible to obtain clear remote sensing images. Use drones to select multiple marking points, create labels corresponding to the types of farmland crops, and establish a data set. The data set is divided into a training set and a test set in a ratio of 8:2.
[0030] S2. Preprocessing and correction: preprocessing and correction of the acquired farmland crop remote sensing image data. The preprocessing of farmland crop remote sensing image data includes cloud removal, radiation correction, image fusion and image stitching. The central wavelength and half-width of the imaging spectrometer are used in the radiation calibration system designed in the MATLAB environment to complete the radiation correction from the image DN value to the surface reflectivity.
[0031] The hyperspectral data captured by the satellite imager are divided into hyperspectral panchromatic images and hyperspectral cube images. The Pansharpen algorithm is used to fuse the hyperspectral panchromatic images and hyperspectral cube images to obtain the sub-band images extracted by IDL programming. Photoscan is used to process the sub-band images extracted by IDL programming and hyperspectral panchromatic images. The sub-band images are merged through IDL programming, and the registration is combined with the measured GPS points. The vector data is introduced for splicing and fusion of the images to output the hyperspectral data of the experimental area.
[0032] S3, data dimensionality reduction processing, dimensionality reduction processing of hyperspectral data, remove redundant data, select minimum noise separation MNF change to reduce the dimensionality of hyperspectral image data, the band before the inflection point of the eigenvalue is rich in feature information, while the band component information after the inflection point is basically zero, so the 20 MNF principal components before the inflection point are selected as the spectral features after data dimensionality reduction;
[0033] S4. Construct a crop classification learning model for training, retain the optimal training model, and use it for crop identification and classification of satellite remote sensing images. Select support vector machine SVM as the crop classification learning model. Compared with traditional classifiers, SVM has higher accuracy and stronger generalization ability. In order to avoid the classification results being affected by other parameters, the radial basis kernel function is selected as the kernel function, and the training set is introduced to train the crop classification learning model.
[0034] In S4, the test set is introduced into the crop classification learning model for testing, the overall classification accuracy and Kappa coefficient indicators are used to evaluate the crop classification learning model, an evaluation threshold is set, and the crop classification learning model training is repeated until the overall classification accuracy and Kappa coefficient indicators are higher than the evaluation threshold, thereby completing the construction of the crop classification learning model.
[0035] Embodiment 2:
[0036] See also Figure 1 - Figure 2 As shown, a crop remote sensing classification method comprises the following steps:
[0037] S1. Data acquisition: Use satellites to acquire high-resolution and hyperspectral remote sensing image data of farmland crops. Use drones to collect image data of different types of farmland crops in the farmland area. When using satellites to acquire high-resolution and hyperspectral remote sensing image data of farmland crops, choose weather with few clouds and little wind as much as possible to obtain clear remote sensing images. Use drones to select multiple marking points, create labels corresponding to the types of farmland crops, and establish a data set. The data set is divided into a training set and a test set in a ratio of 8:2.
[0038] S2. Preprocessing and correction: preprocessing and correction of the acquired farmland crop remote sensing image data. The preprocessing of farmland crop remote sensing image data includes cloud removal, radiation correction, image fusion and image stitching. The central wavelength and half-width of the imaging spectrometer are used in the radiation calibration system designed in the MATLAB environment to complete the radiation correction from the image DN value to the surface reflectivity.
[0039] The hyperspectral data captured by the satellite imager are divided into hyperspectral panchromatic images and hyperspectral cube images. The Pansharpen algorithm is used to fuse the hyperspectral panchromatic images and hyperspectral cube images to obtain the sub-band images extracted by IDL programming. Photoscan is used to process the sub-band images extracted by IDL programming and hyperspectral panchromatic images. The sub-band images are merged through IDL programming, and the registration is combined with the measured GPS points. The vector data is introduced for splicing and fusion of the images to output the hyperspectral data of the experimental area.
[0040] S3, data dimensionality reduction processing, dimensionality reduction processing of hyperspectral data, remove redundant data, select minimum noise separation MNF change to reduce the dimensionality of hyperspectral image data, the band before the inflection point of the eigenvalue is rich in feature information, while the band component information after the inflection point is basically zero, so the 20 MNF principal components before the inflection point are selected as the spectral features after data dimensionality reduction;
[0041] S4. Construct a crop classification learning model for training, retain the optimal training model, and use it for crop identification and classification of satellite remote sensing images. Select support vector machine SVM as the crop classification learning model. Compared with traditional classifiers, SVM has higher accuracy and stronger generalization ability. In order to avoid the classification results being affected by other parameters, the radial basis kernel function is selected as the kernel function, and the training set is introduced to train the crop classification learning model.
[0042] In S4, the test set is introduced into the crop classification learning model for testing, the overall classification accuracy and Kappa coefficient indicators are used to evaluate the crop classification learning model, an evaluation threshold is set, and the crop classification learning model training is repeated until the overall classification accuracy and Kappa coefficient indicators are higher than the evaluation threshold, thereby completing the construction of the crop classification learning model.
[0043] A crop remote sensing classification system, using a crop remote sensing classification method, specifically includes:
[0044] A data acquisition module, wherein the data acquisition processing module is used to acquire satellite farmland crop remote sensing image data and drone farmland crop image data;
[0045] A data processing module, which is used to process the acquired farmland crop remote sensing image data and perform dimensionality reduction processing on the hyperspectral data;
[0046] A classification model is used to classify and identify crops based on the collected data.
[0047] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.
[0048] In the drawings of the embodiments disclosed in the present invention, only the structures related to the embodiments disclosed in the present invention are involved, and other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of the present invention can be combined with each other.
Claims
1. A crop remote sensing classification method, characterized in that: The following steps are involved: S1. Data acquisition: using satellites to obtain high-resolution and hyperspectral remote sensing image data of farmland crops, and using drones to collect image data of different types of farmland crops in the farmland area; S2, preprocessing and correction, preprocessing and correction of the acquired farmland crop remote sensing image data; S3, data dimensionality reduction processing, dimensionality reduction processing of hyperspectral data, and removal of redundant data; S4. Build a crop classification learning model for training, retain the optimal training model, and use it for crop identification and classification of satellite remote sensing images.
2. A crop remote sensing classification method according to claim 1, characterized in that: In S1, a plurality of marking points are selected by a drone, labels corresponding to the types of farmland crops are produced, and a data set is established, and the data set is divided into a training set and a test set in a ratio of 8:
2.
3. A crop remote sensing classification method according to claim 2, characterized in that: In S2, the preprocessing of farmland crop remote sensing image data includes cloud removal, radiation correction, image fusion and image stitching.
4. A crop remote sensing classification method according to claim 3, characterized in that: In S2, the central wavelength and half-width of the imaging spectrometer are used in a radiation calibration system designed in a MATLAB environment to complete the radiation correction from the image DN value to the surface reflectivity.
5. A crop remote sensing classification method according to claim 4, characterized in that: In the S2, the hyperspectral data captured by the satellite imager are divided into hyperspectral panchromatic images and hyperspectral cube images, and the hyperspectral panchromatic images and hyperspectral cube images are fused and processed by Pansharpen algorithm to obtain sub-band images extracted by IDL programming, and Photoscan is used to process the sub-band images extracted by IDL programming and hyperspectral panchromatic images, and the sub-band images are merged through IDL programming, and the actual GPS points are combined for registration, and vector data are introduced for splicing and fusion of the images to output the hyperspectral data of the experimental area.
6. A crop remote sensing classification method according to claim 5, characterized in that: In S3, the minimum noise separation MNF variation is selected to perform dimensionality reduction processing on the hyperspectral image data, and the first 20 MNF principal components are selected as the spectral features after the data dimensionality reduction.
7. A crop remote sensing classification method according to claim 6, characterized in that: In S4, a support vector machine SVM is selected as a crop classification learning model, a radial basis kernel function is selected as a kernel function, and a training set is introduced to train the crop classification learning model.
8. A crop remote sensing classification method according to claim 7, characterized in that: In S4, the test set is introduced into the crop classification learning model for testing, the overall classification accuracy and Kappa coefficient indicators are used to evaluate the crop classification learning model, an evaluation threshold is set, and the crop classification learning model training is repeated until the overall classification accuracy and Kappa coefficient indicators are higher than the evaluation threshold, thereby completing the construction of the crop classification learning model.
9. A crop remote sensing classification system, using a crop remote sensing classification method according to any one of claims 1 to 8, characterized in that: Specifically include: A data acquisition module, wherein the data acquisition processing module is used to acquire satellite farmland crop remote sensing image data and drone farmland crop image data; A data processing module, which is used to process the acquired farmland crop remote sensing image data and perform dimensionality reduction processing on the hyperspectral data; A classification model is used to classify and identify crops based on the collected data.