CN116563721B - Tobacco field extraction method based on layered classification thought - Google Patents
Tobacco field extraction method based on layered classification thought Download PDFInfo
- Publication number
- CN116563721B CN116563721B CN202310853315.XA CN202310853315A CN116563721B CN 116563721 B CN116563721 B CN 116563721B CN 202310853315 A CN202310853315 A CN 202310853315A CN 116563721 B CN116563721 B CN 116563721B
- Authority
- CN
- China
- Prior art keywords
- area
- image
- tobacco field
- region
- interpretation
- 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.)
- Active
Links
- 235000002637 Nicotiana tabacum Nutrition 0.000 title claims abstract description 87
- 238000000605 extraction Methods 0.000 title claims abstract description 23
- 244000061176 Nicotiana tabacum Species 0.000 title 1
- 241000208125 Nicotiana Species 0.000 claims abstract description 86
- 230000011218 segmentation Effects 0.000 claims abstract description 28
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000007637 random forest analysis Methods 0.000 claims abstract description 14
- 238000001228 spectrum Methods 0.000 claims abstract description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 24
- 230000012010 growth Effects 0.000 claims description 22
- 238000013178 mathematical model Methods 0.000 claims description 7
- 210000000746 body region Anatomy 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 9
- 230000003287 optical effect Effects 0.000 abstract description 6
- 230000000007 visual effect Effects 0.000 description 7
- 238000012549 training Methods 0.000 description 6
- 239000002689 soil Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000003698 anagen phase Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003306 harvesting Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000009331 sowing Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 241000607479 Yersinia pestis Species 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- -1 building lands Substances 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000003673 groundwater Substances 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
Classifications
-
- 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
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- 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
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a tobacco field extraction method based on a hierarchical classification idea, and belongs to the technical field of image processing. A tobacco field extraction method based on a layered classification idea comprises the following steps: acquiring a multispectral remote sensing image in a region to be identified; determining an interpretation area and establishing an interpretation mark; performing multi-scale segmentation on the obtained interpretation area to extract texture information in the interpretation area; operating multi-scale segmentation on the interpretation area through the obtained multi-spectrum remote sensing image and texture information so as to distinguish and segment tobacco field plots and other ground objects on the image, and obtaining a preliminary segmentation image; and establishing a random forest model, wherein the input characteristic of the random forest model is at least a preliminary segmentation image, and the output characteristic of the random forest model is a tobacco field distribution area in the area to be identified. The method can reduce the weight of the artificially reported planting plots, solve the problem of objectivity of data, and overcome the problem of lack of optical image data in the growing period so as to quickly obtain the range of the planting plots in the tobacco field.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a tobacco field extraction method based on a hierarchical classification idea.
Background
The method can quickly and accurately acquire the tobacco planting area information, is beneficial to the work such as tobacco growth monitoring, pest and disease damage monitoring, yield estimation and the like, and has important significance for decision and planning of tobacco planting management. Because the difference between the tobacco field plots and other planting plots is small, the existing tobacco field plot acquisition method mainly relies on manual investigation to report step by step, and is low in efficiency and poor in timeliness.
In order to solve the problems of low tobacco planting area information acquisition efficiency and poor timeliness, a remote sensing technology is generally adopted at present to rapidly and accurately acquire tobacco field plot information; the method comprises the basic flow of obtaining high-resolution images of tobacco field plots by aerial photography or satellite images, and then processing the images by using an image processing technology to finish the extraction and identification of the tobacco field plots.
In a rainy region, due to the influence of weather, the quality of an obtained satellite image is low, the problem of lack of an optical image in a key growth period often occurs, a radar remote sensing image is required to be supplemented for tobacco field identification, the complexity of radar remote sensing image identification work is high, in the existing method for identifying by the radar remote sensing image, long-time sequence data of crops to be identified are required, the time sequence data often cover a whole season from sowing to harvesting, and identification work cannot be completed in a short period.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a tobacco field extraction method based on a layered classification idea, which can realize the reduction of the weight of a manually reported planting field, solve the problem of objectivity of data, and overcome the problem of lack of optical image data in the growth period so as to quickly obtain the real tobacco field planting field range.
The tobacco field extraction method based on the layered classification idea comprises the following steps: acquiring a multispectral remote sensing image in a region to be identified; determining an interpretation area and establishing an interpretation mark; performing multi-scale segmentation on the obtained interpretation area to extract texture information in the interpretation area; operating multi-scale segmentation on the interpretation area through the obtained multi-spectrum remote sensing image and texture information so as to distinguish and segment tobacco field plots and other ground objects on the image, and obtaining a preliminary segmentation image; and establishing a random forest model, wherein the input characteristic of the random forest model is at least a preliminary segmentation image, and the output characteristic of the random forest model is a tobacco field distribution area in the area to be identified.
As a further improvement of the present invention, a method of performing a multi-scale segmentation of an interpretation region includes: converting the image map extracted with the texture information in the interpretation region into a gray level image; dividing the region roughly by recognizing the gray level on the gray level image to obtain a target region A, a region C to be divided and a region B; obtaining an NDVI curve of a sample local block in a target area A and each pixel point in an area C to be segmented through a multispectral remote sensing image; and obtaining a specific dividing line about the region A and the region B in the region C according to the obtained NDVI curve so as to obtain an image of the region A after the specific dividing line is divided, namely a preliminary divided image.
As a further improvement of the invention, the NDVI curve of one local block in the target area A is obtained and is the NDVI curve in a certain growth stage of tobacco crops; the obtained NDVI curve of each pixel point in the region C to be segmented is the NDVI curve in the same time period as the growth stage.
As a further improvement of the present invention, the step of obtaining a specific parting line includes:
analyzing and obtaining the variation trend of the NDVI curve of the sample land parcels in the area A;
obtaining at least one slope K1, kn of the NDVI curve of the sample plot within region a, n being an integer greater than or equal to 1;
Respectively comparing the trend of the NDVI curve of each pixel point in the area C with the trend of the NDVI curve of the area A to obtain all the pixel points with the same trend of the curves in the area C;
calculating the slope K11, the third order and the Kn1 of the NDVI curve of each pixel point respectively;
k1,..kn and K11,..kn 1 were compared, respectively; when it is satisfied that K11 falls within [ K1-a, K1+a ],. When Kn+a ], judging that the pixel point belongs to a tobacco field area;
and (3) obtaining pixel points in all the areas C, which are judged to be the tobacco field areas, and connecting a series of pixel points farthest from the area A, wherein the connecting lines are specific dividing lines.
As a further improvement of the present invention, the step of determining the interpretation area comprises: respectively identifying non-potential tobacco field distribution areas in the area to be identified, wherein the non-potential tobacco field distribution areas at least comprise a steep gradient area, a permanent water body area, a water impermeable surface area and a potential tobacco field distribution area; and setting a mask on the non-potential tobacco field distribution area to be shielded, wherein the non-shielded part is the interpretation area.
As a further improvement of the present invention, the step of identifying the steeper region of slope includes: the basic data selects a digital elevation model, and identifies an area with gradient larger than 15 degrees, namely, a mathematical model is set: SLOPE >15 deg. to obtain a steeper SLOPE region.
As a further improvement of the present invention, the step of identifying the permanent water body region includes: the basic data are respectively selecting a first-period winter multispectral image and a first-period summer multispectral image in the area to be identified, respectively calculating NDWI of the first-period winter multispectral image and the first-period summer multispectral image, and setting a mathematical model: NDWI >0; for a certain pixel, the method can meet the requirement that when the index value of the summer NDWI is greater than 0 or the index value of the winter NDWI is greater than 0, the pixel is a water mask area which needs to be shielded.
As a further improvement of the present invention, there is further provided a road removing step including: acquiring road data, wherein the road data is from openstreetmap open source data; setting a 20-meter buffer area for the expressway network, and setting a 5-meter buffer area for the common highway to establish a face vector to be removed; and (3) using an erasing tool in arcmap 10.3.3 to erase road network areas of the tobacco field distribution areas obtained by classifying the random forest model.
As a further improvement of the present invention, further comprising a misclassification step of removing misclassification, the misclassification step comprising: setting a minimum upper pattern area, namely, pattern spots smaller than 10 mu are considered as misclassifications, and the pattern spots need to be removed.
As a further improvement of the invention, the multispectral remote sensing image data comprise sentinel No. 210 m spatial resolution data and environment disaster reduction satellite 16m spatial resolution data.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, the manual reporting weight is reduced, and the remote sensing data is introduced as a main data source for identifying the tobacco field distribution area, so that the objectivity and accuracy of the data can be improved;
aiming at the situation that the optical images are missing in a key growth period and the acquired satellite images are low in quality and the radar remote sensing image data are required to be supplemented due to the fact that weather influences are caused in a cloudy rain area, the characteristics of covering mulching films on tobacco fields in a foggy area in winter are utilized, only cloud-free images in a short period of one growth period or even less than one growth period of crops are selected to serve as data bases for tobacco field identification, the images in the selected period are only required to be ensured to be clear and free of weather images, and the growth state of the crops can be reflected;
Meanwhile, the data to be collected only need to be selected in a short period of one growth stage or even less than one growth stage, so that a period of no cloud and rain can be specially selected when the data is selected, the situation that acquired satellite images are low in quality and even lack of optical images in a key growth period due to the fact that the cloud and rain are more than one period in a part of areas is avoided, the probability of optical image deletion is reduced, the time range requirement on the data to be acquired is reduced, the data acquirability is improved, the data precision is improved, and the identification precision is improved;
Aiming at the problems that in the existing identification method, a time sequence image set is directly established to carry out ground object classification so as to extract the misclassification and confusion in the target ground object, the invention establishes a plurality of ground object masks based on the thought of layered classification, thereby reducing the problems of misclassification and confusion and further improving the accuracy and precision of tobacco field identification; in addition, the range of the classification is narrowed, the calculated amount is reduced, and the classification precision is improved;
In addition, the data acquisition process is simplified, and only images of crops in a short period of one growth stage or even less than one growth stage are selected as data bases by utilizing the characteristic of covering mulching films on tobacco fields in Fujian areas, so that the total amount of data to be processed is reduced, and the difficulty and complexity of data acquisition are simplified; in addition, by utilizing the method of operating multi-scale segmentation, the area which finally needs to process data to obtain a specific segmentation line is only the area C, and compared with the method of identifying and processing the data in the whole area to be identified, the range of the data to be processed is greatly reduced, the total number of the data to be processed is reduced, the data processing pressure is reduced, the data processing efficiency is improved, and the data processing cost is reduced.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the present invention for obtaining a preliminary segmented image;
FIG. 3 is a visual interpretation of a forest in a1 month end cloud image according to the present invention;
FIG. 4 is a visual interpretation of a water body in a cloud-free image at the end of 1 month according to the present invention;
FIG. 5 is a visual interpretation of the architectural terrain in a 1 month end cloud image according to the present invention;
FIG. 6 is a visual interpretation of bare soil in a1 month end cloudless image according to the present invention;
FIG. 7 is a visual interpretation of the overwintering crop in the 1 month end cloud image according to the present invention;
FIG. 8 is a visual interpretation of tobacco field plots in a1 month end cloud image according to the present invention;
FIG. 9 is a diagram of an image after multi-scale segmentation according to the present invention;
FIG. 10 is a view of an image without multi-scale segmentation according to the present invention;
FIG. 11 is a schematic view of road area culling according to the present invention;
FIG. 12 is a schematic view of a determined minimum upper graph area of the present invention;
FIG. 13 is a plot of tobacco field extraction results of the present invention;
FIG. 14 is a schematic view of a gray scale rough division region according to the present invention;
FIG. 15 is a schematic view of a gray scale rough division region according to the present invention;
FIG. 16 is a diagram of NDVI of region A of the present invention;
FIG. 17 is a diagram of NDVI of zone A and zone C of the present invention;
FIG. 18 is a schematic diagram of an accurately segmented region A of the present invention;
fig. 19 is a schematic view of an accurately segmented region a of the present invention.
Detailed Description
First embodiment: referring to fig. 1, a tobacco field extraction method based on a hierarchical classification idea is used for identifying a tobacco field region in a region to be identified, extracting the tobacco field region, and dividing and representing the tobacco field region on an image map;
in this embodiment, the area to be identified is selected as a tobacco field planting area in the Fujian area;
The tobacco field extraction method based on the layered classification idea comprises the following steps:
S1: the area to be identified comprises a gradient steeper area, a permanent water body area, a water-impermeable surface area and a potential tobacco field distribution area; respectively identifying a steep gradient zone, a permanent water zone and a water-impermeable surface zone, and arranging masks on the steep gradient zone, the permanent water zone and the water-impermeable surface zone for shielding, and reserving a potential tobacco field distribution zone for tobacco field identification;
Specifically, the step for identifying a steeper region of slope includes: selecting a digital elevation model by the basic data; downloading 30 m spatial resolution elevation data in the region to be identified on a geographic data space cloud platform, and calculating to obtain gradient data in the region to be identified by utilizing a gradient calculation function in arcmap 10.3.3 software; because tobacco planting needs comparatively flat topography, after comprehensive planting general knowledge, need reject the region that the slope is greater than 15, set up mathematical model promptly: SLOPE >15 DEG, the area obtained according to the formula is the steep area;
The step for identifying the permanent water body region includes: the basic data are respectively selected a first-stage winter multispectral image and a first-stage summer multispectral image in the area to be identified, NDWI (normalized water index) of the first-stage winter multispectral image and the first-stage summer multispectral image is calculated, and the calculation formula of the index is as follows:
NDWI=(NIR-Green)/(NIR+Green)
Wherein NIR represents the reflection value of the near infrared band, green represents the reflection value of the Green band, and in general, NDWI value is between-1 and 1, and the higher the value is, the greater the possibility of water body is; to reject the body of water, a mathematical model is set here: NDWI >0; for a certain pixel, the pixel is a water mask area to be shielded when the index value of the NDWI in summer is more than 0 or the index value of the NDWI in winter is more than 0; the reason that the two time point images are used for obtaining the union is that the fluctuation of the water level line caused by different seasons needs to be considered so as to remove the ground water body more completely;
The step for identifying the water impermeable surface area comprises: the step is completed by adopting the existing surface type data, wherein the data is open source data (downloading address: http:// data.stator loud.pcl.ac.cn/zh/resource/1);
setting a mathematical model: value=80; the water impermeable surface is screened to obtain a water impermeable surface area.
S2: acquiring a multispectral remote sensing image in a region to be identified, and taking the multispectral remote sensing image as a data base of supervision and classification of a subsequent non-mask region; due to the arrangement of the mask, the multispectral remote sensing image obtained at the moment is a multispectral remote sensing image of a potential tobacco field distribution area;
The multispectral remote sensing image data generally uses sentinel No. 210 m spatial resolution data and environment disaster reduction satellite 16m spatial resolution data, and preferably uses sentinel No. 2 satellite data;
Because the tobacco is subjected to four growth stages of seedling returning stage, rooting stage, vigorous stage and mature stage in total between sowing and harvesting, the acquired multispectral remote sensing image is the multispectral remote sensing image of any one growth stage, and the cloud rainfall is less in the time period for acquiring the multispectral remote sensing image, so that the acquisition of the cloud-free image is facilitated;
in this embodiment, the time of the multispectral remote sensing image data is selected to be a cloud-free image in1 month to 2 months.
S3: determining an interpretation zone, and establishing an interpretation mark: the interpretation area is a non-mask area; in order to perform preliminary extraction of tobacco field plots in a non-mask area, a training set needs to be formulated for supervision and classification, and 6 types of ground features are set to be used as training classifiers for the training set, which are respectively as follows: forest, water, construction land, bare soil, overwintering crops, tobacco field plots; 3-8 below are visual interpretation markers of specific classes in a1 month end cloud free image;
as can be seen from fig. 3-8, for three field types, bare soil is yellowish, winter crops are greenish, and tobacco field plots are grey due to the coverage of black and white color-striking films, and have spectral characteristics of a certain watertight surface.
S4: multi-scale segmentation method: performing multi-scale segmentation on the obtained interpretation region; specifically, setting the size of a segmentation window for the obtained interpretation region, wherein the size is usually 2 meters of image spatial resolution, the window shape is set to be square, and the aggregation method is set to be median, so as to extract texture information in the interpretation region;
The texture information is obtained through multi-scale segmentation and is used as a classification image basis, so that the influence of space variability is reduced, the stability of a classification result is improved, and the classification precision is enhanced; and as shown in fig. 9 and fig. 10, the images before and after multi-scale segmentation are respectively compared, so that noise points in the extraction result after multi-scale segmentation are smaller, unfilled holes are less likely to appear in continuous image spots, and the extraction efficiency of the tobacco field is improved.
S5: and (3) performing multi-scale segmentation on the texture information extracted in the step (S4) by utilizing the spectrum information obtained in the step (S2) so as to distinguish and segment the tobacco field land block on the image from other land features, thereby obtaining a preliminary segmentation image.
S6: establishing a random forest model, namely firstly establishing a multiband image by utilizing the spectrum information obtained in the step S2 and the preliminary segmentation image obtained in the step S5, and taking the multiband image as an image base map for supervision and classification;
and selecting a plurality of six-class ground object samples as a training set according to the interpretation mark determined in the step S3, training a random forest classifier, applying the training set to the image base map, obtaining a preliminary classification result, and extracting a tobacco field region.
S7: removing the road area in the tobacco field area in the preliminary classification result obtained in the step S6 through the road data, as shown in FIG. 11; the road data come from openstreetmap open source data, and a face vector to be removed is established by setting a 20-meter buffer area for the expressway network and a 5-meter buffer area for the common highway; and then using an erasing tool in arcmap 10.3.3 to perform preliminary treatment on the tobacco field area in the random forest classification result, and erasing the road network area.
S8: determining the minimum upper graph area, and removing the misclassified area: noise and free pixel points possibly mixed in the extraction result are mostly caused by misclassification, so that a minimum upper graph area is set, namely, the graph spots smaller than 10 mu are considered as misclassification and need to be removed;
Specifically, as shown in fig. 12, the processing result is converted into a vector file, and each land area is obtained by using the calculation geometry area function of arcmap 10.3.3 under the projection coordinate system gcs2000_3_degree_gk_cm_120e (EPSG: 4549), and the map spots smaller than 6667 square meters are eliminated.
S9: and a manual correction step: for the situation that the boundary of the high-precision tobacco field is required, additional manual correction is required; the base map data is generally modified by selecting 1 month base data with a spatial resolution of 5 meters or more; obtaining a final tobacco field extraction result after the step is completed; fig. 13 shows tobacco field extraction results in a certain region.
In step S5, the step of performing multi-scale segmentation on the texture information extracted in step S4 by using the spectral information obtained in step S2 includes:
Z1: converting the image map obtained in the step S4 into a gray level image, wherein the value of each pixel point in the converted gray level image represents the brightness value of the point; the region where the set of pixel points with the same gray level is located is the region where the same ground object sample is located, so that the region can be roughly divided by recognizing the gray level on the gray level image, as shown in fig. 14 and 15 below;
The crops planted in the area A are the same, the crops in the area A are tobacco, and the area A is the area where tobacco fields are finally needed to be obtained; the crop planted in the area B is another crop, or the area B can be ground samples such as bare soil, construction land and water body;
The area C is positioned between the area A and the area B, the area C is a demarcation area between the area A and the area B, the condition that the area A crop and the area B crop exist simultaneously possibly exists, the gray level of the area C is also between the area A and the area B, and the specific parting line of the area A crop and the area B crop can be obtained only by obtaining the parting line in the area C;
z2: obtaining a cloud-free image of a sample local block in the area A in a certain growth stage to obtain an NDVI curve of crops in the area A in the growth stage, namely an NDVI curve of the crops in the tobacco field in the growth stage; as shown in fig. 16, the NDVI curve shows a tendency to decrease, slowly increase, and rapidly increase during this growth phase;
The NDVI value of the sample land block can be obtained by taking the average value of the sum of all pixel points;
Wherein ndvi= (NIR-Red)/(nir+red), green is Green band, red is Red band, NIR is near infrared band, SWIR is short wave near infrared band;
Z3: obtaining at least one slope K1,..kn, n of the NDVI curve, n being an integer greater than or equal to 1; since the NDVI curve shows a tendency to fall first, then rise slowly, and finally rise rapidly during this growth phase, the resulting slope is at least three: k1, K2, K3;
Z4: obtaining an NDVI curve of each pixel point in the region C in the same time period as the growth stage;
Z5: respectively comparing the trend of the NDVI curve of each pixel point in the area C with the trend of the NDVI curve of the area A to obtain all the pixel points with the same trend of the curve in the area C, wherein the obtained NDVI curve on each pixel point shows the trend of descending firstly, then slowly ascending and finally rapidly ascending;
The method is mainly used for screening bare soil, building lands, water bodies and partial forest crops or overwintering crops in potential tobacco field areas, and can be used for distinguishing most of ground object samples from tobacco field plots so as to screen out most of ground object samples needing to be distinguished;
z6: calculating the slope K11, the third order and the Kn1 of the NDVI curve of each pixel point respectively; wherein the time period of the NDVI curve portion represented by K11 is the same as the time period of the NDVI curve portion represented by K1, as shown in fig. 17;
Z7: k1,..kn and K11,..kn 1 were compared, respectively; when K11 is satisfied and falls on [ K1-a, K1+a ],. The main. and Kn1 fall on [ Kn-a, kn+a ], judging that the pixel belongs to a tobacco field area, wherein a is an allowable error value, and a is set by a user independently;
Z8: screening out pixel points in the area C, which are judged to be the tobacco field area, and connecting a series of pixel points farthest from the area A, wherein the connecting lines divide a part of the area C close to the area A into the area A, and a part of the area C close to the area B into the area B, the connecting lines are specific dividing lines, and thick black line parts in the following figures 18 and 19 are specific ranges of the finally obtained area A;
z9: according to the method, all the areas A in the potential tobacco field area are obtained, and the obtained boundaries of the areas A are processed accurate dividing lines, so that the coverage range of the areas A can be accurately obtained, namely, tobacco field parts in the potential tobacco field area can be accurately extracted, and preliminary divided images can be obtained.
Claims (6)
1. The tobacco field extraction method based on the layered classification idea is characterized by comprising the following steps of: acquiring a multispectral remote sensing image in a region to be identified; determining an interpretation area and establishing an interpretation mark; performing multi-scale segmentation on the obtained interpretation area to extract texture information in the interpretation area; operating multi-scale segmentation on the interpretation area through the obtained multi-spectrum remote sensing image and texture information so as to distinguish and segment tobacco field plots and other ground objects on the image, and obtaining a preliminary segmentation image; establishing a random forest model, wherein the input characteristic of the random forest model is at least a preliminary segmentation image, and the output characteristic of the random forest model is a tobacco field distribution area in the area to be identified;
The method for performing operation multi-scale segmentation on the interpretation area comprises the following steps: converting the image map extracted with the texture information in the interpretation region into a gray level image; dividing the region roughly by recognizing the gray level on the gray level image to obtain a target region A, a region C to be divided and a region B; obtaining an NDVI curve of a sample local block in a target area A and each pixel point in an area C to be segmented through a multispectral remote sensing image; obtaining a specific dividing line about the region A and the region B in the region C according to the obtained NDVI curve so as to obtain an image of the region A after being divided by the specific dividing line, namely a preliminary divided image;
Obtaining an NDVI curve of a local block in the target area A as an NDVI curve of a tobacco crop in a certain growth stage; the obtained NDVI curve of each pixel point in the region C to be segmented is the NDVI curve in the same time period as the growth stage;
the step of determining an interpretation zone includes: respectively identifying non-potential tobacco field distribution areas in the area to be identified, wherein the non-potential tobacco field distribution areas at least comprise a steep gradient area, a permanent water body area, a water impermeable surface area and a potential tobacco field distribution area; setting a mask on a non-potential tobacco field distribution area for shielding, wherein the non-shielded part is an interpretation area;
the step of identifying the permanent water body region includes: the basic data are respectively selecting a first-period winter multispectral image and a first-period summer multispectral image in the area to be identified, respectively calculating NDWI of the first-period winter multispectral image and the first-period summer multispectral image, and setting a mathematical model: NDWI >0; for a certain pixel, the method can meet the requirement that when the index value of the summer NDWI is greater than 0 or the index value of the winter NDWI is greater than 0, the pixel is a water mask area which needs to be shielded.
2. The tobacco field extraction method based on the hierarchical classification concept as claimed in claim 1, wherein: the step of obtaining the specific parting line comprises the following steps:
analyzing and obtaining the variation trend of the NDVI curve of the sample land parcels in the area A;
obtaining at least one slope K1, kn of the NDVI curve of the sample plot within region a, n being an integer greater than or equal to 1;
Respectively comparing the trend of the NDVI curve of each pixel point in the area C with the trend of the NDVI curve of the area A to obtain all the pixel points with the same trend of the curves in the area C;
calculating the slope K11, the third order and the Kn1 of the NDVI curve of each pixel point respectively;
k1,..kn and K11,..kn 1 were compared, respectively; when it is satisfied that K11 falls within [ K1-a, K1+a ],. When Kn+a ], judging that the pixel point belongs to a tobacco field area;
and (3) obtaining pixel points in all the areas C, which are judged to be the tobacco field areas, and connecting a series of pixel points farthest from the area A, wherein the connecting lines are specific dividing lines.
3. The tobacco field extraction method based on the hierarchical classification concept as claimed in claim 1, wherein: the step of identifying a steeper region of slope includes: the basic data selects a digital elevation model, and identifies an area with gradient larger than 15 degrees, namely, a mathematical model is set: SLOPE >15 deg. to obtain a steeper SLOPE region.
4. The tobacco field extraction method based on the hierarchical classification concept as claimed in claim 1, wherein: the method also comprises a road eliminating step, wherein the road eliminating step comprises the following steps: acquiring road data, wherein the road data is from openstreetmap open source data; setting a 20-meter buffer area for the expressway network, and setting a 5-meter buffer area for the common highway to establish a face vector to be removed; and using the erasing tool in arcmap 10.3.3 to erase road network areas of the tobacco field distribution areas obtained by classifying the random forest model.
5. The tobacco field extraction method based on the hierarchical classification concept as claimed in claim 1, wherein: the method also comprises a step of removing the misclassification area, wherein the step of removing the misclassification area comprises the following steps: setting a minimum upper pattern area, namely, pattern spots smaller than 10 mu are considered as misclassifications, and the pattern spots need to be removed.
6. The tobacco field extraction method based on the hierarchical classification concept as claimed in claim 1, wherein: the multispectral remote sensing image data comprise sentinel No. 210 m spatial resolution data and environment disaster reduction satellite 16m spatial resolution data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310853315.XA CN116563721B (en) | 2023-07-12 | 2023-07-12 | Tobacco field extraction method based on layered classification thought |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310853315.XA CN116563721B (en) | 2023-07-12 | 2023-07-12 | Tobacco field extraction method based on layered classification thought |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116563721A CN116563721A (en) | 2023-08-08 |
| CN116563721B true CN116563721B (en) | 2024-09-10 |
Family
ID=87498709
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310853315.XA Active CN116563721B (en) | 2023-07-12 | 2023-07-12 | Tobacco field extraction method based on layered classification thought |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116563721B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117636175B (en) * | 2024-01-24 | 2024-09-27 | 北京香田智能科技有限公司 | Data processing method and device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108363949A (en) * | 2017-12-27 | 2018-08-03 | 二十世纪空间技术应用股份有限公司 | A kind of cotton remote-sensing monitoring method based on phenology analysis |
| CN111709379A (en) * | 2020-06-18 | 2020-09-25 | 谢国雪 | Remote sensing image-based hilly area citrus planting land plot monitoring method and system |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8340420B2 (en) * | 2009-10-05 | 2012-12-25 | National Taiwan University | Method for recognizing objects in images |
| CN103093233A (en) * | 2012-12-03 | 2013-05-08 | 中国环境科学研究院 | Forest classification method based on object-oriented high-resolution remote sensing image |
| CN107392115B (en) * | 2017-06-30 | 2021-01-12 | 中原智慧城市设计研究院有限公司 | Traffic sign identification method based on hierarchical feature extraction |
| CN113009485B (en) * | 2021-03-10 | 2024-07-02 | 安徽皖南烟叶有限责任公司 | Remote sensing tobacco field identification method based on improved vegetation index |
| CN114005048A (en) * | 2021-11-07 | 2022-02-01 | 福建师范大学 | Multi-temporal data-based land cover change and thermal environment influence research method |
| CN114332596A (en) * | 2021-11-15 | 2022-04-12 | 杭州领见数字农业科技有限公司 | Overwintering crop identification method and device |
| CN113837673A (en) * | 2021-11-29 | 2021-12-24 | 中国环境科学研究院 | Ecological protection red line marking method for carbon sequestration function of land ecosystem |
| CN114549967A (en) * | 2022-01-10 | 2022-05-27 | 杭州领见数字农业科技有限公司 | Extraction method and device for garlic planting area |
| CN115861629A (en) * | 2022-12-07 | 2023-03-28 | 南京市测绘勘察研究院股份有限公司 | High-resolution farmland image extraction method |
| CN116129276A (en) * | 2023-03-16 | 2023-05-16 | 重庆市气象科学研究所(重庆市生态气象和卫星遥感中心、重庆市农业气象中心) | Remote sensing fine classification method for main grain crops in terrain complex region |
-
2023
- 2023-07-12 CN CN202310853315.XA patent/CN116563721B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108363949A (en) * | 2017-12-27 | 2018-08-03 | 二十世纪空间技术应用股份有限公司 | A kind of cotton remote-sensing monitoring method based on phenology analysis |
| CN111709379A (en) * | 2020-06-18 | 2020-09-25 | 谢国雪 | Remote sensing image-based hilly area citrus planting land plot monitoring method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116563721A (en) | 2023-08-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106384081B (en) | Slope farmland extraction method and system based on high-resolution remote sensing image | |
| US7639842B2 (en) | Remote sensing and probabilistic sampling based forest inventory method | |
| CN106951836B (en) | crop coverage extraction method based on prior threshold optimization convolutional neural network | |
| US8111924B2 (en) | Remote sensing and probabilistic sampling based method for determining the carbon dioxide volume of a forest | |
| US7212670B1 (en) | Method of feature identification and analysis | |
| CN109919083B (en) | Early automatic winter wheat mapping method based on Sentinel-2 image data | |
| CN109063754B (en) | Remote sensing image multi-feature joint classification method based on OpenStreetMap | |
| CN113033670A (en) | Method for extracting rice planting area based on Sentinel-2A/B data | |
| CN116563721B (en) | Tobacco field extraction method based on layered classification thought | |
| CN114612794A (en) | Remote sensing identification method for land covering and planting structure in finely-divided agricultural area | |
| CN111882573A (en) | Cultivated land plot extraction method and system based on high-resolution image data | |
| Arcidiacono et al. | Classification of crop-shelter coverage by RGB aerial images: A compendium of experiences and findings | |
| CN112329733B (en) | Winter wheat growth monitoring and analyzing method based on GEE cloud platform | |
| CN117437489A (en) | Urban green space extraction method based on decision tree model | |
| CN109446965B (en) | Automatic tobacco field identification method based on unmanned aerial vehicle data | |
| Grigillo et al. | Classification based building detection from GeoEye-1 images | |
| CN115512159A (en) | Object-oriented high-resolution remote sensing image earth surface coverage classification method and system | |
| CN115760885A (en) | High-canopy-density wetland forest parameter extraction method based on consumption-level unmanned aerial vehicle image | |
| Hnatushenko et al. | Identification of Objects on Satellite Images Using the Image Texture Properties. | |
| CN116168290B (en) | Arbor-shrub grass classification method based on high-resolution remote sensing image and three-dimensional data | |
| CN118570672A (en) | FVC data set generation method and device based on unmanned aerial vehicle long time sequence image | |
| CN118470441A (en) | Multi-time-phase high-resolution remote sensing image-based farmland classification method and system | |
| Wang et al. | Individual Tree Crown Extraction from High Resolution Image Based on a Threshold-Improved Watershed Algorithm | |
| CN118781494A (en) | Deep learning-based farmland backup resource information extraction method and system | |
| MUSTAFA et al. | Estimating of Diameter at Breast Height for Scattered Pinusbrutiaten. Trees using Remote Sensing Techniques, in Zawita Sub-District, Duhok, Kurdistan Region-Iraq |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |