CN116168244A - Land utilization automatic classification method based on multi-source remote sensing data and cloud computing - Google Patents

Land utilization automatic classification method based on multi-source remote sensing data and cloud computing Download PDF

Info

Publication number
CN116168244A
CN116168244A CN202310132469.XA CN202310132469A CN116168244A CN 116168244 A CN116168244 A CN 116168244A CN 202310132469 A CN202310132469 A CN 202310132469A CN 116168244 A CN116168244 A CN 116168244A
Authority
CN
China
Prior art keywords
classification
data
land
remote sensing
land utilization
Prior art date
Legal status (The legal status 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 status listed.)
Pending
Application number
CN202310132469.XA
Other languages
Chinese (zh)
Inventor
金佳莉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Research Institute of Forestry of Chinese Academy of Forestry
Original Assignee
Research Institute of Forestry of Chinese Academy of Forestry
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 Research Institute of Forestry of Chinese Academy of Forestry filed Critical Research Institute of Forestry of Chinese Academy of Forestry
Priority to CN202310132469.XA priority Critical patent/CN116168244A/en
Publication of CN116168244A publication Critical patent/CN116168244A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a land utilization automatic classification method based on multi-source remote sensing data and cloud computing, which belongs to the technical field of remote sensing image classification and comprises the following steps: s1: sequentially reclassifying and superposing land utilization classified data; s2: carrying out layered sampling on land utilization classified data after superposition processing, and removing noise samples to obtain a sample data set; s3: generating an optimal remote sensing image by utilizing the Sentinel-2 image data set; s4: generating a multi-characteristic wave band combined image; s5: constructing and training a land utilization classification model; s6: and (5) carrying out land utilization automatic classification by utilizing a final land classification model, and verifying classification results. The automatic land utilization classification method can improve the availability of multi-source remote sensing data, quickly acquire reliable training samples, obtain real-time reliable land utilization classification results, and provide dynamic and accurate data guarantee for accurately defining and dynamically monitoring land utilization changes.

Description

Land utilization automatic classification method based on multi-source remote sensing data and cloud computing
Technical Field
The invention belongs to the technical field of remote sensing image classification, and particularly relates to an automatic land utilization classification method based on multi-source remote sensing data and cloud computing.
Background
The global rapid urban land utilization greatly changes urban land coverage and land utilization, faces the challenges of sustainable urban land utilization and increasingly aggravated ecological environment, and develops a large number of greening projects in the city, and the greening projects directly influence the continuous change of urban and suburban forest patterns, thereby influencing various ecological service functions provided by the urban and suburban forests for the city. Therefore, the relation between the urban forest component configuration and the function of providing various ecological services for urban environments and residents is explored, and the environmental factors influencing the relation are revealed, so that the method is a key for optimizing and improving the urban forest ecological service function. However, the conventional urban forest landscape pattern research is affected by data, and cannot simultaneously consider time resolution and space resolution, so that a large amount of financial resources and manpower are required for data acquisition, and the result is affected by scale effects, or is not representative in a small scale or is not floor-standing in a large scale.
In recent years, with the rapid development of remote sensing cloud computing platforms, massive remote sensing image sets, product data and the like are integrated, and the remote sensing cloud computing platforms can be directly called without downloading. Meanwhile, the remote sensing cloud computing platform has mass computing capability, can process mass data with high precision, large range and long time sequence, overturns the traditional data processing mode, provides several generations of convenience for combining mass remote sensing data, sample data and machine learning, solves the core problem of data dependence in urban forest landscape pattern research, and enables multi-scale pattern research to be possible.
At present, land utilization technology based on high-resolution remote sensing images is also used for obtaining land utilization classified products finally by downloading data, processing the data, collecting field samples and selecting classification methods and models. However, rapid increase of urban land utilization is urgently needed to update land utilization data in time. Meanwhile, a large amount of manpower and material resources are consumed in field data acquisition, the standards of the data samples of the sample platforms acquired in the non-field are different, and reasonable application is difficult to obtain. Therefore, automatic land utilization classification of multi-source remote sensing data and cloud computing by means of a remote sensing cloud computing platform is urgently needed.
In summary, the existing land utilization classification technology has poor timeliness, and cannot quickly obtain land utilization classification at any time period; meanwhile, the traditional data processing and acquiring modes greatly limit the use efficiency of data, and a stable and reliable training sample data set cannot be acquired rapidly. Therefore, there is a need for improvements in this technology.
Disclosure of Invention
The invention aims to solve the problems of low precision, poor timeliness and difficult acquisition of training sample data of the existing land utilization classification method based on remote sensing images, and provides a land utilization automatic classification method based on multi-source remote sensing data and cloud computing.
The technical scheme of the invention is as follows: the land utilization automatic classification method based on the multi-source remote sensing data and cloud computing comprises the following steps:
s1: collecting multi-source land utilization classification data of a research area, and sequentially carrying out reclassification and superposition treatment on the land utilization classification data to obtain superposition treated land utilization classification data;
s2: carrying out layered sampling on land utilization classification data after superposition processing, and removing noise samples by using a neighborhood analysis method to obtain a sample data set;
s3: acquiring a Sentinel-2 image data set, and generating an optimal remote sensing image by utilizing the Sentinel-2 image data set;
s4: calculating the characteristic wave band of the optimal remote sensing image according to the coverage category reclassified by land utilization classification data, and generating a multi-characteristic wave band combined image;
s5: constructing and training a land utilization classification model by utilizing the sample data set and the multi-characteristic wave band combined image, and generating a final land utilization classification model;
s6: and (5) carrying out land utilization automatic classification by utilizing a final land classification model, and verifying classification results.
Further, in step S1, land use classification data is reclassified into unused land, cultivated land, woodland, grassland, water body, and construction land;
in step S1, superposition processing is performed on the land use data subjected to reclassification, so as to obtain land use classification results of the same type, and the superposition processing is completed.
Further, in step S2, hierarchical sampling is performed according to the area ratio of land utilization classification data after the superposition processing, where the calculation formula is as follows:
Figure BDA0004084752280000021
wherein n is i A sample number representing class i land use classification data,
Figure BDA0004084752280000022
representing the accuracy of land utilization classification data s i The area of the i-th land use data is represented.
Further, in step S2, the specific method for removing the noise sample by using the neighborhood analysis method is as follows: and adjusting a neighborhood window according to the cell size of the land utilization class, and removing noise samples by utilizing 3*3 neighborhood analysis.
Further, step S3 comprises the sub-steps of:
s31: sequentially performing time screening, position screening and cloud content percentage sorting on the Sentinel-2 data set, and generating an optimal image set, wherein the calculation formula of image data in the optimal image set is as follows:
n＝10*m
wherein n represents the number of images of the optimal image data set, and m represents the number of images spanned by the research area;
s32: and acquiring a Sentinel-2:Cloud Probability data set, carrying out data combination on the Sentinel-2 image data set and the Sentinel-2:Cloud Probability data set, cutting by using a percentage probability threshold, and synthesizing by using a median to obtain the optimal remote sensing image.
Further, in step S4, the characteristic wave band of the optimal remote sensing image includes a normalized vegetation index NDVI, a bare soil index BSI, a vegetation water content index LSWI, and a corrected water body index MNDWI; combining the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI with the red wave band, the green wave band, the blue wave band, the near infrared wave band and the short wave infrared wave band of the optimal remote sensing image to generate a multi-characteristic wave band combined image;
the calculation formulas of the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI are respectively as follows:
NDVI＝(NIR-R)/(NIR+R)
BSI＝((SWIR2+R)-(NIR-BLUE))/((SWIR2+R)+(NIR-BLUE))
LSWI＝(SWIR1-NIR)/(SWIR1+NIR)
MNDWI＝(BLUE-SWIR1)/(BLUE+SWIR1)
where NIR denotes the near infrared band, R denotes the red band, SWIR1 denotes the first short wave infrared, SWIR2 denotes the second short wave infrared, and BLUE denotes the BLUE band.
Further, step S5 comprises the sub-steps of:
s51: taking 70% of sample data sets as training sample sets;
s52: and training the random forest classification model by utilizing the training sample set to the multi-characteristic wave band combined image, and adjusting the random forest classification model to generate a final land utilization classification model.
Further, in step S52, the specific method for generating the final land use classification model is as follows: and adjusting the number of the classification trees in the random forest classification model until the classification precision of the random forest classification model is not improved along with the increase of the number of the classification trees, and generating a final land utilization classification model by taking the number of the classification trees as the final tree value of the random forest classification model.
Further, in step S6, the specific method for verifying the classification result is as follows: taking 30% of sample data sets as test sample sets, calculating confusion matrixes of land utilization classification data of different categories according to classification results of land classification models, and calculating total classification accuracy OA, kappa coefficients and F1 scores of the test sample sets by using the confusion matrixes; if the total classification accuracy OA is greater than 90%, the Kappa coefficient is greater than 85% and the F1 score is greater than 85%, the classification result of the land classification model is verified to pass; the calculation formulas of the total classification precision OA, the Kappa coefficient and the F1 score are respectively as follows:
Figure BDA0004084752280000031
Kappa＝(OA-p e )/(1-p e )
F1＝2×UA×PA/(UA+PA)
Figure BDA0004084752280000032
wherein c represents the total number of categories, T i Representing the number of correctly classified samples for each category, q representing the total number of samples, p e Indicating occasional consistency errors, UA indicating user accuracy, PA indicating producer accuracy, a i Representing the number of real samples of each type b i The number of samples of each type obtained by prediction is represented, and n' represents the total number of samples.
The beneficial effects of the invention are as follows:
(1) The land utilization automatic classification method provides a method for acquiring real-time sample data sets based on multi-source remote sensing data, so that the efficiency of sample acquisition is greatly improved, and manpower and material resources are reduced; meanwhile, a remote sensing image generation network is built by combining with the Sentinel-2 image data set, so that the method still has good applicability to a period of severe cloud interference, and a remote sensing image with good quality is obtained;
(2) The automatic land utilization classification method utilizes random forests to generate a land utilization classification model, and carries out reverse adjustment on forest trees to obtain an optimal land utilization classification model;
(3) The automatic land utilization classification method can improve the availability of multi-source remote sensing data, quickly acquire reliable training samples, obtain real-time reliable land utilization classification results, and provide dynamic and accurate data guarantee for accurately defining and dynamically monitoring land utilization changes.
Drawings
FIG. 1 is a flow chart of an automatic land utilization classification method based on multi-source remote sensing data and cloud computing;
FIG. 2 is a schematic diagram of a multi-source land utilization classification product superposition analysis;
FIG. 3 is a schematic diagram of a multi-source remote sensing data generation real-time sample dataset.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings.
As shown in fig. 1, the invention provides an automatic land utilization classification method based on multi-source remote sensing data and cloud computing, which comprises the following steps:
s1: collecting multi-source land utilization classification data of a research area, and sequentially carrying out reclassification and superposition treatment on the land utilization classification data to obtain superposition treated land utilization classification data;
s2: carrying out layered sampling on land utilization classification data after superposition processing, and removing noise samples by using a neighborhood analysis method to obtain a sample data set;
s3: acquiring a Sentinel-2 image data set, and generating an optimal remote sensing image by utilizing the Sentinel-2 image data set;
s4: calculating the characteristic wave band of the optimal remote sensing image according to the coverage category reclassified by land utilization classification data, and generating a multi-characteristic wave band combined image;
s5: constructing and training a land utilization classification model by utilizing the sample data set and the multi-characteristic wave band combined image, and generating a final land utilization classification model;
s6: and (5) carrying out land utilization automatic classification by utilizing a final land classification model, and verifying classification results.
The method adopts multi-source remote sensing data hierarchical sampling to obtain real-time reliable sample data, and is flexible in application; in the remote sensing image processing, generating an optimal image set by utilizing a remote sensing image generating network; meanwhile, for the coverage type of land utilization classification, respectively calculating a normalized vegetation index (ndvi), a bare soil index (bsi), a vegetation water content index (lswi) and a corrected water body index (mdwi); and finally, selecting a random forest classifier, adjusting a forest tree according to the classification precision change to obtain an optimal classification model, and realizing automatic classification of land utilization of the multi-source remote sensing data on a remote sensing cloud platform.
In the embodiment of the present invention, in step S1, land use classification data is reclassified into unused land, cultivated land, woodland, grassland, water body, and construction land;
the land use classification data with high precision of multiple sources are reclassified according to classification requirements, the land use classification data should contain at least one category to be classified, and the higher the precision is, the better the effect is. Taking FROM-GLC10 data of Qinghai university and Dynamic World V1 data published by Google authorities as examples, the embodiment of the invention reclassifies the FROM-GLC10 data into unused lands, cultivated lands, woodlands, grasslands, water bodies and construction lands, as shown in table 1;
TABLE 1
FROM-GLC10 class Numerical value DynamicWorld class Numerical value Reclassifying categories Numerical value
Cultivated land 10 Water body 0 Unused land 0
Woodlands 20 Woodlands 1 Cultivated land 1
Grassland 30 Grassland 2 Woodlands 2
Shrubs (shrubs) 40 Submerged vegetation 3 Grassland 3
Wet land 50 Cultivated land 4 Water body 4
Water body 60 Shrubs (shrubs) 5 Building land 5
Moss source 70 Building land 6
Waterproof surface 80 Bare land 7
Bare land 90 Glacier/snow cover 8
Glacier/snow cover 100
In step S1, superposition processing is performed on the land use data subjected to reclassification, so as to obtain land use classification results of the same type, and the superposition processing is completed.
As shown in fig. 2, the reclassified land use data is subjected to superposition analysis, and the land use classification results of the same type are taken to obtain the land use classification products after superposition processing as shown in table 2.
TABLE 2
Figure BDA0004084752280000051
Figure BDA0004084752280000061
In the embodiment of the present invention, in step S2, hierarchical sampling is performed according to the area ratio of land utilization classification data after superposition processing, where the calculation formula is as follows:
Figure BDA0004084752280000062
wherein n is i A sample number representing class i land use classification data,
Figure BDA0004084752280000063
representing the accuracy of land utilization classification data s i The area of the i-th land use data is represented.
In the embodiment of the present invention, in step S2, a specific method for removing a noise sample by using a neighborhood analysis method is as follows: and adjusting a neighborhood window according to the cell size of the land utilization class, and removing noise samples by utilizing 3*3 neighborhood analysis.
The neighborhood analysis uses surrounding pixels as spatial data references to reduce single pixel data errors, and the neighborhood window can be adjusted according to the unit size of the land utilization class, and the window size is odd, such as 3*3, 5*5, 7*7, and the like.
In an embodiment of the present invention, step S3 comprises the sub-steps of:
s31: sequentially performing time screening, position screening and cloud content percentage sorting on the Sentinel-2 data set, and generating an optimal image set, wherein the calculation formula of image data in the optimal image set is as follows:
n＝10*m
wherein n represents the number of images of the optimal image data set, and m represents the number of images spanned by the research area;
eliminating the data sets which do not accord with the preset time period and the preset position, and completing time screening and position screening; the cloud content of the Sentinel-2 dataset is ranked according to the percentage.
S32: and acquiring a Sentinel-2:Cloud Probability data set, carrying out data combination on the Sentinel-2 image data set and the Sentinel-2:Cloud Probability data set, cutting by using a percentage probability threshold, and synthesizing by using a median to obtain the optimal remote sensing image. The Sentinel-2:Cloud Probability data collection was from the European air agency.
In the step S3, the optimal remote sensing image data set ordering is adopted, so that the edge effect of the image can be removed, and meanwhile, the cloud removal of the Sentinel-2:Cloud Probability data set can be changed according to a threshold value. For specific details reference is made to fig. 3.
In the embodiment of the present invention, in step S4, the characteristic wave band of the optimal remote sensing image includes a normalized vegetation index NDVI, a bare soil index BSI, a vegetation water content index LSWI, and a corrected water body index MNDWI; combining the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI with the red wave band, the green wave band, the blue wave band, the near infrared wave band and the short wave infrared wave band of the optimal remote sensing image to generate a multi-characteristic wave band combined image;
the calculation formulas of the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI are respectively as follows:
NDVI＝(NIR-R)/(NIR+R)
BSI＝((SWIR2+R)-(NIR-BLUE))/((SWIR2+R)+(NIR-BLUE))
LSWI＝(SWIR1-NIR)/(SWIR1+NIR)
MNDWI＝(BLUE-SWIR1)/(BLUE+SWIR1)
wherein NIR represents near infrared band, R represents red band, SWIR1 represents first short-wave infrared, SWIR2 represents second short-wave infrared, BLUE represents BLUE band; BLUE corresponds to the 2 nd band of Sentinel-2, R corresponds to the 4 th band of Sentinel-2, NIR corresponds to the 8 th band of Sentinel-2, SWIR1 corresponds to the 11 th band of Sentinel-2; SWIR2 corresponds to band 12 of Sentinel-2.
In an embodiment of the present invention, step S5 includes the sub-steps of:
s51: taking 70% of sample data sets as training sample sets;
s52: and training the random forest classification model by utilizing the training sample set to the multi-characteristic wave band combined image, and adjusting the random forest classification model to generate a final land utilization classification model.
In the embodiment of the present invention, in step S52, the specific method for generating the final land use classification model is as follows: and adjusting the number of the classification trees in the random forest classification model until the classification precision of the random forest classification model is not improved along with the increase of the number of the classification trees, and generating a final land utilization classification model by taking the number of the classification trees as the final tree value of the random forest classification model.
The random forest model has good robustness and high precision, and is widely applied to land utilization classification of remote sensing images. At the same time, the sample data set is set with a random parameter random and they are randomly split into 7:3, 70% are used to train random forest classification models, 30% are used to test classification model accuracy.
In the embodiment of the present invention, in step S6, the specific method for verifying the classification result is as follows: taking 30% of sample data sets as test sample sets, calculating confusion matrixes of land utilization classification data of different categories according to classification results of land classification models, and calculating total classification accuracy OA, kappa coefficients and F1 scores of the test sample sets by using the confusion matrixes; if the total classification accuracy OA is greater than 90%, the Kappa coefficient is greater than 85% and the F1 score is greater than 85%, the classification result of the land classification model is verified to pass; the calculation formulas of the total classification precision OA, the Kappa coefficient and the F1 score are respectively as follows:
Figure BDA0004084752280000071
Kappa＝(OA-p e )/(1-p e )
F1＝2×UA×PA/(UA+PA)
PA＝p ii /p i+
UA＝p jj /p j+
Figure BDA0004084752280000081
wherein c represents the total number of categories, T i Representing the number of correctly classified samples for each category, q representing the total number of samples, p e Indicating occasional consistency errors, UA indicating user accuracy, PA indicating producer accuracy, a i Representing the number of real samples of each type b i The number of samples of each type obtained by prediction is represented, and n' represents the total number of samples.
The accuracy is calculated by using a confusion matrix which is obtained by counting test samples, and is used for comparing a classification result with an actual measured value in the evaluation of the accuracy of an image supervision classification image, and then the accuracy of the classification result is displayed in a matrix form.
The embodiment of the invention takes the statistically obtained Beijing urban land utilization classification confusion matrix as an example, and is shown in table 3.
TABLE 3 Table 3
Figure BDA0004084752280000082
Calculating drawing precision by using the confusion matrix, wherein the evaluation indexes comprise: overall classification Accuracy (overlay Accuracy): means, for each random sample, the probability that the classified result is consistent with the test data type; kappa coefficient: is an index for consistency test and can also be used for measuring the classification effect. Since for classification problems, so-called consistency is whether the model prediction result and the actual classification result are consistent.
Based on table 3, the overall accuracy OA can be obtained as:
OA＝(272+272+295+274+294+240)/1771＝0.931
based on Table 3, kappa coefficients can be obtained as:
Kappa＝(0.931-0.1667)/(1-0.1667)＝0.916
calculating the precision of each type of land utilization by using the confusion matrix, wherein the evaluation indexes comprise: producer's Accuracy): optionally selecting a sample in the classification result, wherein the probability of the sample is consistent with the actual type of the ground; user precision (User's Accuracy): optionally a sample on the ground with the same probability as the classification result;
based on table 3, production accuracy PA can be obtained as shown in table 4:
TABLE 4 Table 4
Unused land 272/(272+3+0+0+0+27)＝0.901
Cultivated land 272/(1+272+7+6+0+6)＝0.925
Woodlands 295/(0+7+295+5+0+0)＝0.961
Grassland 274/(0+13+6+274+0+0)＝0.935
Water body 294/(0+0+0+0+294+1)＝0.997
Building land 240/(30+8+0+0+1+240)＝0.860
Based on table 3, the production accuracy UA can be obtained as shown in table 5:
TABLE 5
Unused land 272/(272+1+0+0+0+30)＝0.898
Cultivated land 272/(3+272+7+13+0+8)＝0.898
Woodlands 295/(0+7+295+6+0+0)＝0.958
Grassland 274/(0+6+5+274+0+0)＝0.961
Water body 294/(0+0+2+0+294+1)＝0.990
Building land 240/(27+6+0+0+1+240)＝0.876
Based on table 3, an F1 score can be obtained as shown in table 6:
TABLE 6
Unused land 2×0.901×0.898/(0.901+0.898)＝0.899
Cultivated land 2×0.925×0.898/(0.925+0.898)＝0.911
Woodlands 2×0.961×0.958/(0.961+0.958)＝0.959
Grassland 2×0.935×0.961/(0.935+0.961)＝0.948
Water body 2×0.997×0.990/(0.997+0.990)＝0.993
Building land 2×0.860×0.876/(0.860+0.876)＝0.863
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit thereof, and such modifications and combinations remain within the scope of the present disclosure.

Claims (9)

1. The land utilization automatic classification method based on the multi-source remote sensing data and cloud computing is characterized by comprising the following steps of:
s1: collecting multi-source land utilization classification data of a research area, and sequentially carrying out reclassification and superposition treatment on the land utilization classification data to obtain superposition treated land utilization classification data;
s2: carrying out layered sampling on land utilization classification data after superposition processing, and removing noise samples by using a neighborhood analysis method to obtain a sample data set;
s3: acquiring a Sentinel-2 image data set, and generating an optimal remote sensing image by utilizing the Sentinel-2 image data set;
s4: calculating the characteristic wave band of the optimal remote sensing image according to the coverage category reclassified by land utilization classification data, and generating a multi-characteristic wave band combined image;
s5: constructing and training a land utilization classification model by utilizing the sample data set and the multi-characteristic wave band combined image, and generating a final land utilization classification model;
s6: and (5) carrying out land utilization automatic classification by utilizing a final land classification model, and verifying classification results.
2. The automatic land use classification method based on multi-source remote sensing data and cloud computing according to claim 1, wherein in the step S1, land use classification data are reclassified as unused land, cultivated land, woodland, grassland, water body and construction land;
in the step S1, superposition processing is performed on the land use data subjected to reclassification, so as to obtain land use classification results of the same type, and the superposition processing is completed.
3. The automatic land use classification method based on multi-source remote sensing data and cloud computing according to claim 1, wherein in the step S2, hierarchical sampling is performed according to the area ratio of land use classification data after superposition processing, and the calculation formula is as follows:
Figure FDA0004084752270000011
wherein n is i A sample number representing class i land use classification data,
Figure FDA0004084752270000012
representing the accuracy of land utilization classification data s i The area of the i-th land use data is represented.
4. The land use automatic classification method based on multi-source remote sensing data and cloud computing according to claim 1, wherein in the step S2, the specific method for removing the noise sample by using the neighborhood analysis method is as follows: and adjusting a neighborhood window according to the cell size of the land utilization class, and removing noise samples by utilizing 3*3 neighborhood analysis.
5. The land use automatic classification method based on multi-source remote sensing data and cloud computing as claimed in claim 1, wherein said step S3 comprises the sub-steps of:
s31: sequentially performing time screening, position screening and cloud content percentage sorting on the Sentinel-2 data set, and generating an optimal image set, wherein the calculation formula of image data in the optimal image set is as follows:
n＝10*m
wherein n represents the number of images of the optimal image data set, and m represents the number of images spanned by the research area;
s32: and acquiring a Sentinel-2:Cloud Probability data set, carrying out data combination on the Sentinel-2 image data set and the Sentinel-2:Cloud Probability data set, cutting by using a percentage probability threshold, and synthesizing by using a median to obtain the optimal remote sensing image.
6. The land use automatic classification method based on multi-source remote sensing data and cloud computing according to claim 1, wherein in the step S4, the characteristic wave band of the optimal remote sensing image includes a normalized vegetation index NDVI, a bare soil index BSI, a vegetation moisture content index LSWI and a corrected water body index MNDWI; combining the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI with the red wave band, the green wave band, the blue wave band, the near infrared wave band and the short wave infrared wave band of the optimal remote sensing image to generate a multi-characteristic wave band combined image;
the calculation formulas of the normalized vegetation index NDVI, the bare soil index BSI, the vegetation water content index LSWI and the corrected water body index MNDWI are respectively as follows:
NDVI＝(NIR-R)/(NIR+R)
BSI＝((SWIR2+R)-(NIR-BLUE))/((SWIR2+R)+(NIR-BLUE))
LSWI＝(SWIR1-NIR)/(SWIR1+NIR)
MNDWI＝(BLUE-SWIR1)/(BLUE+SWIR1)
where NIR denotes the near infrared band, R denotes the red band, SWIR1 denotes the first short wave infrared, SWIR2 denotes the second short wave infrared, and BLUE denotes the BLUE band.
7. The automatic land use classification method based on multi-source remote sensing data and cloud computing as claimed in claim 1, wherein said step S5 comprises the sub-steps of:
s51: taking 70% of sample data sets as training sample sets;
s52: and training the random forest classification model by utilizing the training sample set to the multi-characteristic wave band combined image, and adjusting the random forest classification model to generate a final land utilization classification model.
8. The automatic land use classification method based on multi-source remote sensing data and cloud computing according to claim 7, wherein in the step S52, the specific method for generating the final land use classification model is as follows: and adjusting the number of the classification trees in the random forest classification model until the classification precision of the random forest classification model is not improved along with the increase of the number of the classification trees, and generating a final land utilization classification model by taking the number of the classification trees as the final tree value of the random forest classification model.
9. The automatic land utilization classification method based on multi-source remote sensing data and cloud computing according to claim 1, wherein in the step S6, the specific method for verifying the classification result is as follows: taking 30% of sample data sets as test sample sets, calculating confusion matrixes of land utilization classification data of different categories according to classification results of land classification models, and calculating total classification accuracy OA, kappa coefficients and F1 scores of the test sample sets by using the confusion matrixes; if the total classification accuracy OA is greater than 90%, the Kappa coefficient is greater than 85% and the F1 score is greater than 85%, the classification result of the land classification model is verified to pass; the calculation formulas of the total classification precision OA, the Kappa coefficient and the F1 score are respectively as follows:
Figure FDA0004084752270000031
Kappa＝(OA-p e )/(1-p e )
F1＝2×UA×PA/(UA+PA)
Figure FDA0004084752270000032
wherein c represents the total number of categories, T i Representing the number of correctly classified samples for each category, q representing the total number of samples, p e Indicating occasional consistency errors, UA indicating user accuracy, PA indicating producer accuracy, a i Representing the number of real samples of each type b i The number of samples of each type obtained by prediction is represented, and n' represents the total number of samples.
CN202310132469.XA 2023-02-09 2023-02-09 Land utilization automatic classification method based on multi-source remote sensing data and cloud computing Pending CN116168244A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310132469.XA CN116168244A (en) 2023-02-09 2023-02-09 Land utilization automatic classification method based on multi-source remote sensing data and cloud computing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310132469.XA CN116168244A (en) 2023-02-09 2023-02-09 Land utilization automatic classification method based on multi-source remote sensing data and cloud computing

Publications (1)

Publication Number Publication Date
CN116168244A true CN116168244A (en) 2023-05-26

Family

ID=86412903

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310132469.XA Pending CN116168244A (en) 2023-02-09 2023-02-09 Land utilization automatic classification method based on multi-source remote sensing data and cloud computing

Country Status (1)

Country Link
CN (1) CN116168244A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117541940A (en) * 2024-01-10 2024-02-09 日照市自然资源和规划局 Land utilization classification method and system based on remote sensing data

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117541940A (en) * 2024-01-10 2024-02-09 日照市自然资源和规划局 Land utilization classification method and system based on remote sensing data
CN117541940B (en) * 2024-01-10 2024-03-22 日照市自然资源和规划局 Land utilization classification method and system based on remote sensing data

Similar Documents

Publication Publication Date Title
CN112381861B (en) Forest land point cloud data registration and segmentation method based on foundation laser radar
CN108020211A (en) A kind of method of unmanned plane aeroplane photography estimation instruction plant biomass
CN109543630A (en) Remote sensing image forest land extracting method and system, storage medium, electronic equipment based on deep learning
CN108007438A (en) The estimating and measuring method of unmanned plane aeroplane photography remote sensing wetland plant biomass
CN109829425B (en) Farmland landscape small-scale ground feature classification method and system
CN109977991A (en) Forest resourceies acquisition method based on high definition satellite remote sensing
CN110176005B (en) Remote sensing image segmentation method based on normalized index and multi-scale model
CN105447274A (en) Method of performing coastal wetland drawing for medium-resolution remote sensing image by utilizing object-oriented classification technology
CN107247927A (en) A kind of remote sensing images coastline information extracting method and system based on K-T Transformation
CN116168244A (en) Land utilization automatic classification method based on multi-source remote sensing data and cloud computing
CN113723319A (en) Rice planting distribution extraction method based on Fengyun satellite data
CN117115669B (en) Object-level ground object sample self-adaptive generation method and system with double-condition quality constraint
Zheng et al. Single shot multibox detector for urban plantation single tree detection and location with high-resolution remote sensing imagery
WO2023213142A1 (en) Ecological quality evaluation and partitioning method and apparatus based on improved remote-sensed ecological indices
CN116502803A (en) Method and device for accounting physical quantity of land resource assets
CN114842356B (en) High-resolution earth surface type sample automatic generation method, system and equipment
CN115147726B (en) City form map generation method and device, electronic equipment and readable storage medium
CN116561509A (en) Urban vegetation overground biomass accurate inversion method and system considering vegetation types
CN115953685A (en) Multi-layer multi-scale division agricultural greenhouse type information extraction method and system
CN112347926B (en) High-resolution image city village detection method based on building morphology distribution
Xi et al. High-Confidence Sample Generation Technology and Application for Global Land-Cover Classification
CN107832805A (en) It is a kind of that technology of the volumetric position error on the evaluation influence of remote sensing soft nicety of grading is eliminated based on probability positions model
CN112509142B (en) Bean strain rapid three-dimensional reconstruction method based on phenotype-oriented precise identification
CN108985154A (en) Small size atural object sub-pixed mapping localization method and system based on image concentration class
CN110554406B (en) Method for inverting secondary forest structure parameters based on unmanned aerial vehicle stereo photogrammetry point cloud

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination