CN-115512218-B - Tea garden identification method, device, equipment and storage medium based on multi-temporal remote sensing image

CN115512218BCN 115512218 BCN115512218 BCN 115512218BCN-115512218-B

Abstract

The application discloses a tea garden identification method, device, equipment and storage medium based on multi-temporal remote sensing images, and relates to the technical field of agricultural remote sensing, comprising the following steps of determining the physical period and the agricultural operation period of tea leaves, and collecting multi-temporal remote sensing image data corresponding to an area to be identified; the method comprises the steps of calculating a normalized vegetation index value and a pre-constructed normalized target crop difference index value according to remote sensing image data, obtaining an altitude value and a spectrum characteristic value of each band, calculating weights of the normalized vegetation index, the normalized target crop difference index, an altitude characteristic and the spectrum characteristic according to the obtained characteristic values, and classifying according to the weights of the characteristic values to obtain a recognition result of a region to be recognized. According to the application, the growth period and important agronomic operation time characteristics of tea are fully considered, a new characteristic index is constructed according to the remote sensing spectrum characteristics of the tea garden in the period, and the tea garden information can be accurately identified by matching with a classification algorithm.

Inventors

CAO LI
GU HUIBO
YE HONGBO
YUE XIAOLAN

Assignees

浙江甲骨文超级码科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20220913

Claims (6)

1. A tea garden identification method based on multi-temporal remote sensing images is characterized by comprising the following steps: Determining a physical weather period and an agricultural operation period of tea, and collecting multi-temporal remote sensing image data of an area to be identified in the physical weather period and the agricultural operation period; Filtering images containing coastal aerosols, water vapor and cloud bands from the remote sensing image data, and carrying out radiation correction and atmosphere correction on the filtered remote sensing image data; calculating a normalized vegetation index value and a pre-constructed normalized target crop difference index value according to the remote sensing image data, and acquiring an altitude value of the area to be identified and a spectral characteristic value of each wave band in the remote sensing image data; Respectively calculating the weights of the normalized vegetation index, the normalized target crop difference index, the elevation characteristic and the spectrum characteristic according to the obtained characteristic values, wherein the weights comprise: Respectively constructing a plurality of second classifiers taking tea gardens and non-tea gardens as classification targets by taking the normalized vegetation index, the normalized target crop difference index, the elevation characteristic and the spectral characteristic as classification characteristics; Inputting the obtained characteristic values into each second classifier for training, and taking the sum of the classification accuracy of different classes under the same characteristic as the multiclass weight of the corresponding characteristic; Classifying according to the weight of each characteristic value to obtain the identification result of the area to be identified, wherein the method comprises the following steps: Combining points repeatedly appearing in different categories and points of unknown categories in the training result into a new data set, voting by taking multi-category weights of each feature as division basis, and obtaining all recognition results of the region to be recognized; The construction method of the normalized target crop difference index comprises the following steps: Collecting a training sample, wherein the training sample comprises a tea garden sample and each control group sample; comparing the tea garden sample with each comparison group sample, and respectively extracting remote sensing image wave band characteristics corresponding to the tea garden and each comparison group; constructing a normalized target crop difference index according to the wave band characteristics, wherein the normalized target crop difference index is as follows: NDCI=(B max -B min )/(B max +B min ) Wherein B max represents band information in which the target crop band is at the highest point in all categories at the same time, and B min represents band information in which the target crop band is at the lowest point in all categories at the same time.
2. The tea garden identification method based on multi-temporal remote sensing image according to claim 1, wherein the remote sensing image data is provided by a sentinel number 2 satellite.
3. The tea garden identification method based on multi-temporal remote sensing image according to claim 1, further comprising performing time-phase differential processing on the remote sensing image data to obtain a time-sequence differential data source, wherein the time-sequence differential data source is used for representing time-varying characteristics of the remote sensing image.
4. Tea garden recognition device based on multi-temporal remote sensing image, its characterized in that includes: the acquisition module is used for determining the physical weather period and the agronomic operation period of the tea, and acquiring multi-temporal remote sensing image data of the area to be identified in the physical weather period and the agronomic operation period; The extraction module is used for calculating a normalized vegetation index value and a pre-constructed normalized target crop difference index value according to the remote sensing image data, and acquiring an altitude value of the area to be identified and a spectral characteristic value of each wave band in the remote sensing image data; the division module is used for respectively calculating the weights of the normalized vegetation index, the normalized target crop difference index, the elevation characteristic and the spectrum characteristic according to the obtained characteristic values, and comprises the following steps: Respectively constructing a plurality of second classifiers taking tea gardens and non-tea gardens as classification targets by taking the normalized vegetation index, the normalized target crop difference index, the elevation characteristic and the spectral characteristic as classification characteristics; Inputting the obtained characteristic values into each second classifier for training, and taking the sum of the classification accuracy of different classes under the same characteristic as the multiclass weight of the corresponding characteristic; The identification module is used for classifying according to the weight of each characteristic value to obtain an identification result of the area to be identified, and comprises the following steps: Combining points repeatedly appearing in different categories and points of unknown categories in the training result into a new data set, voting by taking multi-category weights of each feature as division basis, and obtaining all recognition results of the region to be recognized; The construction method of the normalized target crop difference index comprises the following steps: Collecting a training sample, wherein the training sample comprises a tea garden sample and each control group sample; comparing the tea garden sample with each comparison group sample, and respectively extracting remote sensing image wave band characteristics corresponding to the tea garden and each comparison group; constructing a normalized target crop difference index according to the wave band characteristics, wherein the normalized target crop difference index is as follows: NDCI=(B max -B min )/(B max +B min ) Wherein, B max represents the wave band information of the highest point of the wave bands of the target crops in all categories at the same time, and B min represents the wave band information of the lowest point of the wave bands of the target crops in all categories at the same time; The method further comprises the following steps of after acquiring multi-temporal remote sensing image data of the area to be identified in the physical period and the agronomic operation period: And filtering images containing coastal aerosol, water vapor and a rolling wave band in the remote sensing image data, and carrying out radiation correction and atmosphere correction on the filtered remote sensing image data.
5. An electronic device comprising a memory and a processor, the memory configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement a multi-temporal remote sensing image-based tea garden identification method as claimed in any one of claims 1-3.
6. A computer readable storage medium storing a computer program, wherein the computer program when executed causes a computer to implement a tea garden identification method based on multi-temporal remote sensing image as claimed in any one of claims 1 to 3.

Description

Tea garden identification method, device, equipment and storage medium based on multi-temporal remote sensing image Technical Field The application relates to the technical field of agricultural remote sensing, in particular to a tea garden identification method, device, equipment and storage medium based on multi-temporal remote sensing images. Background With the development of satellite remote sensing images, the data acquisition of high-resolution satellite remote sensing images and multispectral and hyperspectral satellite remote sensing images is more and more convenient, so that the technology for identifying tea garden information by utilizing the satellite remote sensing images is more and more, but the traditional identification method depends on a single spectrum or an original index dividing method or a single image, and is easy to cause wrong division or missing division. Disclosure of Invention Aiming at the defects in the prior art, the application provides a tea garden identification method based on multi-temporal remote sensing images according to the influence of the tea growing period and the agronomic operation period on the tea garden remote sensing spectrum. In order to achieve the above purpose, the present application adopts the following technical scheme: the application discloses a tea garden identification method based on multi-temporal remote sensing images, which comprises the following steps of: Determining a physical weather period and an agricultural operation period of tea, and collecting multi-temporal remote sensing image data of an area to be identified in the physical weather period and the agricultural operation period; calculating a normalized vegetation index value and a pre-constructed normalized target crop difference index value according to the remote sensing image data, and acquiring an altitude value of the area to be identified and a spectral characteristic value of each wave band in the remote sensing image data; Respectively calculating weights of normalized vegetation indexes, normalized target crop difference indexes, elevation features and spectrum features according to the obtained characteristic values; And classifying according to the weight of each characteristic value to obtain the identification result of the area to be identified. Preferably, the remote sensing image data is provided by a sentinel satellite number 2. Preferably, the acquiring the multi-temporal remote sensing image data of the area to be identified in the physical period and the agricultural operation period further comprises: And filtering images containing coastal aerosol, water vapor and a rolling wave band in the remote sensing image data, and carrying out radiation correction and atmosphere correction on the filtered remote sensing image data. Preferably, the method for constructing the normalized target crop difference index comprises the following steps: Collecting a training sample, wherein the training sample comprises a tea garden sample and each control group sample; comparing the tea garden sample with each comparison group sample, and respectively extracting remote sensing image wave band characteristics corresponding to the tea garden and each comparison group; constructing a normalized target crop difference index according to the wave band characteristics, wherein the normalized target crop difference index is as follows: NDCI=(Bmax-Bmin)/(Bmax+Bmin) Wherein B max represents band information in which the target crop band is at the highest point in all categories at the same time, and B min represents band information in which the target crop band is at the lowest point in all categories at the same time. Preferably, the calculating the weights of the normalized vegetation index, the normalized target crop difference index, the elevation feature and the spectral feature according to the obtained feature values respectively includes: Respectively constructing a plurality of second classifiers taking tea gardens and non-tea gardens as classification targets by taking the normalized vegetation index, the normalized target crop difference index, the elevation characteristic and the spectral characteristic as classification characteristics; and inputting the obtained characteristic values into each second classifier for training, and taking the sum of the classification accuracy of different classes under the same characteristic as the multiclass weight of the corresponding characteristic. Preferably, the inputting the weight of each feature value into a first classifier trained in advance to classify the feature value, to obtain the recognition result of the region to be recognized, includes: Combining points repeatedly appearing in different categories and points of unknown categories in the training result into a new data set, inputting the new data set into a first classifier which is trained in advance and takes multi-category weights of each feature as division basis