CN112183428A - Wheat planting area segmentation and yield prediction method - Google Patents
Wheat planting area segmentation and yield prediction method Download PDFInfo
- Publication number
- CN112183428A CN112183428A CN202011081232.6A CN202011081232A CN112183428A CN 112183428 A CN112183428 A CN 112183428A CN 202011081232 A CN202011081232 A CN 202011081232A CN 112183428 A CN112183428 A CN 112183428A
- Authority
- CN
- China
- Prior art keywords
- planting area
- wheat planting
- image
- remote sensing
- segmentation
- 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
Links
- 241000209140 Triticum Species 0.000 title claims abstract description 108
- 235000021307 Triticum Nutrition 0.000 title claims abstract description 108
- 230000011218 segmentation Effects 0.000 title claims abstract description 61
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012545 processing Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000005855 radiation Effects 0.000 claims description 8
- 239000002689 soil Substances 0.000 claims description 6
- 238000001556 precipitation Methods 0.000 claims description 5
- 238000001308 synthesis method Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 4
- 230000008020 evaporation Effects 0.000 claims description 4
- 238000001704 evaporation Methods 0.000 claims description 4
- 230000014509 gene expression Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000003595 spectral effect Effects 0.000 claims description 4
- 229930002875 chlorophyll Natural products 0.000 claims description 3
- 235000019804 chlorophyll Nutrition 0.000 claims description 3
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000002310 reflectometry Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 abstract description 12
- 238000013507 mapping Methods 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 10
- 238000001228 spectrum Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000013136 deep learning model Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 4
- 230000002068 genetic effect Effects 0.000 description 4
- 238000012417 linear regression Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 239000000443 aerosol Substances 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Educational Administration (AREA)
- Life Sciences & Earth Sciences (AREA)
- Agronomy & Crop Science (AREA)
- Animal Husbandry (AREA)
- Marine Sciences & Fisheries (AREA)
- Mining & Mineral Resources (AREA)
- Primary Health Care (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a wheat planting area segmentation and yield prediction method, which utilizes the existing open multispectral remote sensing data and satellite monitoring meteorological environment data to realize the accurate segmentation and scientific yield estimation method of a winter wheat planting area, thereby avoiding the cost on time, labor and financial resources brought by large-scale on-site mapping and meteorological monitoring. Meanwhile, the model has better mobility, not only can be applied to the division of the planting area of winter wheat and the prediction of yield, but also can be effectively applied to other crops, and provides possibility for realizing agricultural modernization and business.
Description
Technical Field
The invention belongs to the technical field of image processing and remote sensing, and particularly relates to a wheat planting area segmentation and yield prediction method.
Background
The growth of wheat is influenced by various external environmental factors such as solar radiation, atmosphere, hydrology, soil and the like. By comprehensively applying professional knowledge in the field of remote sensing and an optimization algorithm in mathematics, quantitative representation of external environment factors related to wheat growth and biological growth characteristics expressed by the wheat through multiple spectra are obtained, and therefore the division of a planting area and the construction of an estimated yield model are achieved. As more growth environment data need to be collected and counted, a large amount of manpower and material resources are inevitably consumed for manual on-site monitoring by using an instrument, and reliable and stable monitoring data can be obtained by converting remote sensing data into related index indexes of crop growth through an inversion method. At present, meteorological monitoring data generated by remote sensing satellite sensors represented by MODIS are widely used, and data such as absorption of vegetation chlorophyll and aerosol in the atmosphere can be obtained from the MODIS. Therefore, how to accurately divide the planting area of the target crop from the multi-spectrum remote sensing image is a problem which needs to be solved urgently at present.
At present, papers and patents related to agricultural assessment mainly rely on spectral data in multispectral images to perform simple formula calculation in the aspect of planting area segmentation, and the method can better segment the planting areas of crops under the conditions of less data volume and more knowledge of related fields. However, when the remote sensing related knowledge is not deep enough and the processing of crops in a large area is required, a lot of time is consumed for manually adjusting the parameters in the model, and the automation degree is not high. Therefore, how to realize a model which can be flexibly applied to other areas and is more automatic is an urgent problem to be solved for remote sensing image area crop segmentation.
The Chinese patent with the application number of 201210133136.0 provides a winter wheat yield estimation method for assimilating the characteristic of a leaf area index time sequence curve, a WOFOST model is adopted to predict the yield of winter wheat in a unit area within the scope of the Shijiazhuan city, but the scheme does not involve the division of a winter wheat planting area, the yield of the unit area is predicted for the whole area grid, no difference exists in the area grid, only the Leaf Area Index (LAI) of a remote sensing satellite MODIS sensor is used, the WOFOST model is optimized by fitting the time sequence curve of the LAI, and the yield simulation is seriously low only by depending on the MODIS LAI to perform the assimilation of the model. The Chinese patent with the application number of 201910094036.3 provides a regional winter Wheat estimation method based on remote sensing phenological assimilation and particle swarm optimization algorithm, a scheme based on MODIS LAI index is also adopted, a crop growth model adopts MCWLA-Wheat, and an LAI time sequence curve is also adopted.
Therefore, the current thesis and patent still use a simple one-dimensional linear equation based on multispectral remote sensing data or use a genetic algorithm to find and optimize a prediction model for the method for estimating the yield of the winter wheat. However, the crop growth characteristic index obtained based on multispectral calculation and the meteorological environment data volume obtained by the remote sensing satellite sensor are large, and the large data volume cannot be fitted only by using a simple one-dimensional linear equation; moreover, the use of algorithms like genetic algorithms does not have strict mathematical convergence and theoretical proof of finding the optimal value, and it cannot be guaranteed that the model can obtain a good prediction result. Therefore, an urgent need exists for solving the problem of how to apply an algorithm with better accuracy and fitting data characteristics to crop estimation.
Disclosure of Invention
In view of the above, the invention provides a wheat planting area segmentation and yield prediction method, which utilizes the existing open multispectral remote sensing data and satellite monitoring meteorological environment data to realize accurate segmentation and scientific estimation method of the winter wheat planting area, thereby avoiding the cost on time, labor and financial resources brought by large-scale on-site mapping and meteorological monitoring.
A wheat planting region segmentation and yield prediction method comprises the following steps:
(1) acquiring multispectral remote sensing images historically acquired by a target area for years and meteorological environment information of a wheat planting area in the images and carrying out data cleaning work;
(2) carrying out image preprocessing work of radiometric calibration and atmospheric correction on the multispectral remote sensing image, and converting the image into an RGB format;
(3) dividing the multispectral remote sensing image processed in the step (2) into a plurality of image blocks, and marking pixels belonging to a wheat planting area in the image blocks;
(4) establishing a convolutional neural network, inputting the image blocks into the neural network one by one to train the image blocks so as to obtain a segmentation model for distinguishing wheat planting areas in the image;
(5) calculating crop growth characteristics of the wheat planting area according to the spectral information of the wheat planting area in the multispectral remote sensing image;
(6) establishing a polynomial relation between the yield of the wheat planting area and meteorological environment information and crop growth characteristic indexes, and fitting the polynomial by using historical data to obtain a prediction model for measuring and calculating the yield of the wheat planting area;
(7) and acquiring a multispectral remote sensing image of a target area to be predicted at present, preprocessing the image and dividing the multispectral remote sensing image into a plurality of image blocks, inputting each image block into the segmentation model to obtain a wheat planting area in the image, further acquiring current meteorological environment information of the wheat planting area, calculating current crop growth characteristics of the wheat planting area, and finally inputting the current meteorological environment information and the crop growth characteristics into the prediction model to calculate the yield of the wheat planting area.
Further, the meteorological environment information in the step (1) includes solar radiation energy, surface reflectivity, monthly average surface temperature, monthly surface temperature extremum, monthly average air temperature, monthly temperature extremum, monthly average precipitation, soil moisture, monthly average atmospheric pressure, monthly average atmospheric evaporation, and monthly average steam pressure.
Further, the data cleaning work on the meteorological environment information in the step (1) comprises the steps of removing abnormal values, complementing missing values and normalizing numerical values.
Further, the multispectral remote sensing image is converted into an RGB format by adopting a standard false color synthesis method in the step (2), so that the loss of the wave band which plays an important role in the statistics of the crop growth characteristics in the multispectral remote sensing image can be ensured as little as possible.
Further, the multispectral remote sensing image is divided into a plurality of image blocks by overlapping in the step (3) by using a sliding window method, and overlapping is needed between adjacent divided blocks, because the overlapping needs to be introduced in the subsequent segmentation reconstruction process, so that noise errors possibly generated at the boundary of the image by the segmentation model are eliminated.
Further, in the step (4), the minimum binary cross entropy loss function is used as an optimization target, and a stochastic gradient descent method is adopted to train the convolutional neural network.
Further, the crop growth characteristics of the wheat planting area in the step (5) comprise a normalized vegetation index NDVI, a green chlorophyll vegetation index GCVI and an enhanced vegetation index EVI, and the specific expressions are as follows:
wherein: NIR is the average value of the near-infrared wave band of each pixel point in the wheat planting area, RED is the average value of the infrared wave band of each pixel point in the wheat planting area, GREEN is the average value of the GREEN wave band of each pixel point in the wheat planting area, and BLUE is the average value of the BLUE wave band of each pixel point in the wheat planting area.
Further, the polynomial expression in the step (6) is as follows:
y=β0x0+β1x1+β2x2+…+βnxn
wherein: y is the yield of the wheat planting area, x0~xnCorresponding to each index value beta of the wheat planting area about meteorological environment information and crop growth characteristics0~βnIs x0~xnThe corresponding polynomial coefficients.
Compared with the similar planting region segmentation and yield estimation technology based on the multispectral remote sensing image at present, the method has the following advantages:
1. the deep learning model in the computer vision field is applied to the multispectral remote sensing image, and the original multispectral remote sensing image is converted by adopting a standard false color synthesis method, so that the model can process the multispectral remote sensing image; compared with the method for segmenting the winter wheat planting area based on the inversion data at present, the accuracy of segmentation can be improved by using the model, and misjudgment caused by similarity of crop spectra is avoided.
2. The remote sensing data and the meteorological environment data used in the invention can be obtained from a website of a Chinese academy remote sensing and digital globe institute or NASA, and the quality of the data is better, so that the manpower and time cost is saved compared with manual monitoring.
3. The inversion vegetation index and meteorological environment data of winter wheat growth are comprehensively considered, and multispectral remote sensing data and meteorological environment data are collected according to the winter wheat growth period of the historical year and about every half month; compared with a one-dimensional linear model only depending on NDVI indexes in a related paper, the method has better robustness and accuracy;
4. according to the method, a linear regression algorithm is used to obtain a polynomial model about the multi-spectral data, the meteorological data and the winter wheat yield by combining the vegetation index and the meteorological data which are obtained based on the multi-spectral calculation; compared with the method based on the genetic algorithm in the similar patents, the method can obtain more accurate yield estimation results.
Drawings
FIG. 1 is a schematic flow chart of a wheat planting region segmentation and yield prediction method of the present invention.
FIG. 2 is a schematic diagram of the segmentation result of the 2018-year winter wheat planting area of Zhengzhou city obtained by using the segmentation model of the invention.
Detailed Description
In order to more specifically describe the present invention, the following detailed description is provided for the technical solution of the present invention with reference to the accompanying drawings and the specific embodiments.
As shown in FIG. 1, the invention introduces a method applied to multi-spectral remote sensing image segmentation processing and winter wheat yield estimation, which comprises the following specific steps:
step S1: and acquiring data required by the winter wheat planting region segmentation and yield estimation model, wherein the data comprises multispectral remote sensing data of the target region history and meteorological environment data of the target region history.
Considering the interval of different growth stages of the winter wheat and the time of the remote sensing satellite for one circle around the earth, multispectral remote sensing images of the region and corresponding meteorological environment data are collected from the planting time of the winter wheat and once every half month, so that the multispectral remote sensing images can cover all the growth stages of the winter wheat, including a green returning period, a heading period and the like. The meteorological environment data required to be collected are time-synchronized with the collected multispectral remote sensing data, the time-synchronization comprises solar radiation, surface-month average temperature and extremum, month average precipitation, soil moisture, month average atmospheric pressure, month average atmospheric evaporation capacity, month average steam pressure and the like, and the data can be obtained from a remote sensing satellite data website disclosed by NASA.
Step S2: and for each acquired remote sensing image, segmenting the large remote sensing image into small blocks by adopting a sliding window method, wherein an overlapping area is formed between the adjacent small blocks, and the size of the sliding window needs to be matched with the input size of a deep learning model segmented in the later winter wheat planting area. The deep learning model for the winter wheat planting area segmentation can use a U-Net or a model based on U-Net improvement. The U-Net model is used for semantic segmentation of images in the field of computer vision and is widely applied to the aspect of processing medical images.
Step S3: the constructed segmentation model and the processed multispectral remote sensing image used for training are used in the step, and the specific optimization process is as follows:
firstly, constructing a semantic segmentation model by using convolution and a full connection layer, wherein a multispectral remote sensing image is cut into small regions with adjacent regions mutually overlapped after being preprocessed in the steps and then input into the segmentation model; the segmentation model needs to divide pixels on the remote sensing image into a winter wheat planting area and other areas, and a minimized binary cross entropy loss function is used as an optimization target, wherein the formula of the binary cross entropy loss function is as follows:
the object processed by the segmentation model is each pixel point on the remote sensing image, and y in the loss functioniThe correct class label that indicates a certain pixel point,a category label representing the pixel point prediction; inputting the preprocessed and labeled remote sensing image into a segmentation model, and optimizing the segmentation model by using a random gradient descent methodAnd (4) molding.
Stopping the training optimization of the segmentation model when one of the following conditions is satisfied:
(a) when the binary cross entropy loss function value does not decrease on the training data set;
(b) the training round of the model on the data set exceeded 20 rounds.
And finally obtaining a segmentation model of the winter wheat planting area through the data preparation and processing and the construction and training of the segmentation model.
The evaluation index of the model can be the accuracy of the pixel class after segmentation or an IoU evaluation index, and the two indexes are expressed as follows:
although the segmentation model can be evaluated well using pixel accuracy and is also essentially the same as the optimization goal of the model, using the IoU index can supplement the evaluation index using only pixel accuracy, since the planted area is a continuous area. It should be noted that, in consideration of different spectrum and visual characteristics exhibited by the winter wheat at different growth stages, corresponding segmentation models need to be trained for the winter wheat at different growth stages respectively; in the training stage of the model, the actual planting area of the current year is taken as a target; in the prediction stage, for the prediction year, the prediction about the winter wheat planting area is respectively obtained by using the segmentation models of the winter wheat in several growth stages, and finally the average value is calculated, so that the result is more reliable.
Step S4: and step (3) accumulating the vegetation indexes such as NDVI, GCVI, EVI and the like in each remote sensing image according to the segmentation result of the winter wheat planting area obtained in the step (S3), wherein the calculation formula is as follows:
the vegetation index mainly uses four bands of near infrared, red, green and blue, wherein the reflection spectrum of the vegetation is mainly concentrated on the near infrared band, so that the vegetation can show red in an RGB format image synthesized by adopting standard false colors, and the vegetation index can be used for better modeling the growth characteristics of winter wheat.
Step S5: after the processing of the remote sensing image of the target area, the training of the segmentation model and the processing of the vegetation characteristic index are completed in the previous step, the meteorological environment characteristic data of the target area is processed at this stage. The meteorological environment characteristic data required to be collected by the target area comprise indexes such as atmospheric pressure, steam pressure, monthly average air temperature, monthly air temperature extreme value, air humidity, solar radiation, earth surface reflection, soil humidity, monthly precipitation and the like. Before being applied to a prediction model, the data needs to be preprocessed, wherein the preprocessing comprises missing values, abnormal value elimination, numerical value normalization and the like. The meteorological data have a close correlation with the final yield of winter wheat, and the specific numerical correlation is measured as shown in table 1.
TABLE 1
Step S6: the calculation of the vegetation index in the winter wheat planting area and the preprocessing operation of the meteorological environment characteristics are already calculated and completed in the previous steps. In the step, a linear regression model is constructed, and the vegetation index, the meteorological environment characteristics, the winter wheat planting area and the current year yield which are obtained by processing and calculating the steps in the historical year are input. Linear regression is a widely used machine learning prediction model, and its basic form is as follows:
y=β0x0+β1x1+β2x2+…+βnxn
wherein: y represents a prediction result, the model is trained by using a least square method, the optimization target of the final model is a minimum mean square error, the error describes the error between the predicted yield and the actual yield of a target area, and the model is optimized by using a random gradient descent method, wherein the formula of an error function of the model is as follows:
wherein: xiIs x in the above formula0,x1,……xnRepresents that β is the coefficient corresponding to each term x. The estimated termination condition of the model training process is a preset maximum iteration number or the number of successive rounds of loss function values of the model on the training set does not decrease.
Step S7: collecting multispectral remote sensing images and corresponding meteorological data for the years to be predicted of the target area at the time frequency of about half a month interval, sequentially completing the operations of segmentation of wheat planting areas, statistics of vegetation indexes, preprocessing of meteorological environment data and normalization according to the steps, and finally sending the processed data into a prediction and yield estimation model to obtain the predicted yield of the winter wheat of the target area.
In the following, we will exemplify the specific application of the method of the present invention by using Zheng City of Henan province as a research area.
Step S1: the method comprises the steps of selecting Zheng city and county areas of each place and city of Henan province as research areas, wherein 6 districts, 5 county-level cities and one county are administered in the Zheng city, the research areas are located between 112 degrees 42 'to 114 degrees 14' at east longitude and 34 degrees 16 'to 34 degrees 58' at northern latitude, the landform is mainly plain, and the areas are located in the main wheat producing areas of the original regions. The local climate belongs to the temperate zone seasonal climate, and the annual average temperature is 15.6 ℃; the average rainfall is 542.15 mm all year round, the frost-free period is 209 days, and the sunshine time all year round is about 1869.7 hours. The meteorological environment data with enough precision and large time span can be found from the earth remote sensing satellite data disclosed by NASA, wherein the data are classified into atmosphere, atmospheric temperature, wind, earth climate indexes, earth surface, hydrology and the like according to categories. According to the correlation analysis between the meteorological environment indexes and the wheat yield, selecting the meteorological environment indexes with large winter wheat yield images, and acquiring meteorological environment data of a target area at about every half month, wherein the meteorological environment data comprise monthly average air temperature (DEG C), monthly air temperature extreme values (DEG C), solar radiation (MJ/square meter), atmospheric pressure (Kpa), steam pressure (Kpa), soil moisture content (%), monthly evaporation capacity (cm), monthly precipitation capacity (mm), earth surface reflectivity (%), monthly average earth surface temperature (DEG C) and monthly earth surface temperature extreme values (DEG C). The data collected once every half month can contain the data of each stage of winter wheat growth as much as possible, including the green returning period, heading period and the like of the winter wheat, the data can be obtained from the open website of NASA, and the time interval of the data is enough to meet the collection frequency of about half month, so that the requirement of the invention is met. The multispectral remote sensing image of the target area can be acquired from a website of a national space agency earth observation and data center, and the area to be acquired is selected from the website according to the time to be acquired, so that the multispectral remote sensing image covering the area can be acquired.
Step S2: and preprocessing the collected meteorological environment data of the target area. Firstly, the collected multispectral remote sensing image of the target area does not only contain the target area Zhengzhou city, and the Zhengzhou city area needs to be cut from the multispectral remote sensing image according to the geographical boundary of the administrative district of the Zhengzhou city. The specific operation steps are as follows: firstly, acquiring a longitude and latitude file of an administrative division boundary in GeoJSON format of an administrative division in Zhengzhou city, wherein the coordinates of the file are the same as the geographic coordinate standard used by a multispectral remote sensing image by using WSN-84; then, the geographic information of the administrative boundaries is used, GDAL library programming is adopted, or professional software such as ENVI or ArcGIS is used for cutting the remote sensing image, the pixel value outside the target area is set to be 0, and the multispectral remote sensing image can be obtained from sensors of remote sensing satellites of different models, so that operations such as radiometric calibration, atmospheric correction and the like are required to be applied to the collected multispectral remote sensing image to convert relative radiation into absolute radiation.
Step S3: the collected multispectral remote sensing data has tens of thousands of pixel points in length and width, and cannot be directly input into the segmentation model, and generally multispectral remote sensing images have 4, 8 and 16 wave bands, so that the multispectral remote sensing images also need to be converted into images in an RGB format. Firstly, converting a remote sensing image into an image in an RGB format, and adopting a standard false color synthesis method in the specific implementation, wherein the standard false color synthesis method uses red, green, blue and near-infrared wave bands of a multispectral remote sensing image; the green band contains more reflection spectrum information of vegetation, and the near-infrared band contains the most reflection spectrum information of the vegetation; the method adopting standard false color synthesis can better help to identify the ground features and retain more reflection spectrum information of the ground vegetation.
Step S4: after the step S3 of converting the multispectral image into the RGB image, the image needs to be cut into small image blocks by a sliding window method, and there is an overlap between the adjacent small blocks. Compared with the common method of dividing the grids for segmenting the remote sensing image, the method of sliding windows is adopted, and the adjacent areas are overlapped, so that the noise error generated by the segmentation model at the edge of each small image block can be eliminated better.
Step S5: after the multispectral data is converted and segmented in the step S4, the image of each small block is processed by adopting a segmentation model, and a winter wheat planting area is obtained. In the embodiment, a U-Net model is used, and the U-Net semantic segmentation model is widely applied to the processing of medical images, and can still obtain a good segmentation effect particularly under the condition of unbalanced categories in the images; in the U-Net model, an input image is subjected to down-sampling and up-sampling operations, and finally a matrix with the same size as the input image is output, wherein each value in the matrix represents a class label of a corresponding pixel. After the winter wheat planting area in each small image area is divided by using the division model, the small images are placed according to the original positions of the small images, and the overlapped areas among the small images are subjected to fusion processing, so that noise possibly generated by the division model at the edge of the image is eliminated, and the result is shown in fig. 2.
Step S6: and preprocessing the acquired and processed meteorological environment, including abnormal data processing, missing value processing, numerical value normalization processing and the like. In addition, according to the segmentation result of the winter wheat planting region in the step S5 and the vegetation index indexes such as NDVI, GCVI, EVI and the like obtained by accumulative calculation on the regions, the polynomial relationship between the variables and the yield of the winter wheat in the target region is constructed by combining the meteorological environment data obtained by processing and the planting area of the winter wheat, the problem is solved by using a least square method, and finally a polynomial model close to the optimum can be obtained.
The embodiment of the invention adopts a winter wheat planting region segmentation and linear regression yield estimation method based on deep learning, fully exerts the advantages of combining a deep learning model with a multispectral remote sensing image, and uses a machine learning method to search the internal relation between meteorological environment variables and winter wheat yield. In the segmentation of the winter wheat planting area, crop vegetation spectral information contained in the multispectral remote sensing image is fully utilized, and a segmentation task of the winter wheat planting area is better completed by the aid of a segmentation model. And a regression algorithm in machine learning is used based on the segmentation result and meteorological environment data, so that a more optimal solution can be found compared with the traditional experience-based and genetic algorithm. In specific implementation, based on historical data of Zhengzhou city, data of the Zhengzhou city statistical bureau are used, data in 2018 are used as model verification and evaluation, a trained model is used for segmenting the winter wheat planting area of the Zhengzhou city, the area of the winter wheat planting area of the Zhengzhou city in 2018 is 1676654592 square meters, and the error of the area of the winter wheat planting area of the Zhengzhou city statistical bureau in the year 1655000000 square meters is less than 3 percent; the yield of the winter wheat is estimated by the yield estimation model, the finally predicted yield is 79.76706083 ten thousand tons, and the error of the 78.8 ten thousand tons of the data of the Zheng State City statistical bureau in 2018 is less than 3%.
The method can avoid the cost of manpower monitoring of the traditional method, greatly improve the growth monitoring of winter wheat in each growth stage and the prediction of yield data, and provide a quick, convenient and scientific data basis for the scientific decision of tracking and regulating the grain situation of relevant departments of the country.
The method fully utilizes the existing open multispectral remote sensing data and satellite monitoring meteorological environment data, realizes accurate segmentation and scientific estimation of the winter wheat planting area, and avoids the cost on time, labor and financial resources brought by large-scale on-site mapping and meteorological monitoring. Meanwhile, the model has better mobility, not only can be applied to the division of the planting area of winter wheat and the prediction of yield, but also can be effectively applied to other crops, and provides possibility for realizing agricultural modernization and business.
The embodiments described above are presented to enable a person having ordinary skill in the art to make and use the invention. It will be readily apparent to those skilled in the art that various modifications to the above-described embodiments may be made, and the generic principles defined herein may be applied to other embodiments without the use of inventive faculty. Therefore, the present invention is not limited to the above embodiments, and those skilled in the art should make improvements and modifications to the present invention based on the disclosure of the present invention within the protection scope of the present invention.
Claims (8)
1. A wheat planting region segmentation and yield prediction method comprises the following steps:
(1) acquiring multispectral remote sensing images historically acquired by a target area for years and meteorological environment information of a wheat planting area in the images and carrying out data cleaning work;
(2) carrying out image preprocessing work of radiometric calibration and atmospheric correction on the multispectral remote sensing image, and converting the image into an RGB format;
(3) dividing the multispectral remote sensing image processed in the step (2) into a plurality of image blocks, and marking pixels belonging to a wheat planting area in the image blocks;
(4) establishing a convolutional neural network, inputting the image blocks into the neural network one by one to train the image blocks so as to obtain a segmentation model for distinguishing wheat planting areas in the image;
(5) calculating crop growth characteristics of the wheat planting area according to the spectral information of the wheat planting area in the multispectral remote sensing image;
(6) establishing a polynomial relation between the yield of the wheat planting area and meteorological environment information and crop growth characteristic indexes, and fitting the polynomial by using historical data to obtain a prediction model for measuring and calculating the yield of the wheat planting area;
(7) and acquiring a multispectral remote sensing image of a target area to be predicted at present, preprocessing the image and dividing the multispectral remote sensing image into a plurality of image blocks, inputting each image block into the segmentation model to obtain a wheat planting area in the image, further acquiring current meteorological environment information of the wheat planting area, calculating current crop growth characteristics of the wheat planting area, and finally inputting the current meteorological environment information and the crop growth characteristics into the prediction model to calculate the yield of the wheat planting area.
2. The wheat planting area segmentation and yield prediction method of claim 1, wherein: the meteorological environment information in the step (1) comprises solar radiation energy, earth surface reflectivity, monthly average earth surface temperature, monthly earth surface temperature extreme value, monthly average air temperature, monthly air temperature extreme value, monthly average precipitation, soil moisture, monthly average atmospheric pressure, monthly average atmospheric evaporation capacity and monthly average steam pressure.
3. The wheat planting area segmentation and yield prediction method of claim 1, wherein: and (2) performing data cleaning work on the meteorological environment information in the step (1) and including removal of abnormal values, completion of missing values and normalization processing of logarithmic values.
4. The wheat planting area segmentation and yield prediction method of claim 1, wherein: and (3) converting the multispectral remote sensing image into an RGB format by adopting a standard false color synthesis method in the step (2).
5. The wheat planting area segmentation and yield prediction method of claim 1, wherein: and (3) dividing the multispectral remote sensing image into a plurality of image blocks in an overlapping manner by adopting a sliding window method.
6. The wheat planting area segmentation and yield prediction method of claim 1, wherein: and (4) taking the minimized binary cross entropy loss function as an optimization target, and training the convolutional neural network by adopting a random gradient descent method.
7. The wheat planting area segmentation and yield prediction method of claim 1, wherein: the crop growth characteristics of the wheat planting area in the step (5) comprise a normalized vegetation index NDVI, a green chlorophyll vegetation index GCVI and an enhanced vegetation index EVI, and the specific expression is as follows:
wherein: NIR is the average value of the near-infrared wave band of each pixel point in the wheat planting area, RED is the average value of the infrared wave band of each pixel point in the wheat planting area, GREEN is the average value of the GREEN wave band of each pixel point in the wheat planting area, and BLUE is the average value of the BLUE wave band of each pixel point in the wheat planting area.
8. The wheat planting area segmentation and yield prediction method of claim 1, wherein: the polynomial expression in the step (6) is as follows:
y=β0x0+β1x1+β2x2+…+βnxn
wherein: y is the yield of the wheat planting area, x0~xnCorresponding to each index value beta of the wheat planting area about meteorological environment information and crop growth characteristics0~βnIs x0~xnThe corresponding polynomial coefficients.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011081232.6A CN112183428A (en) | 2020-10-09 | 2020-10-09 | Wheat planting area segmentation and yield prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011081232.6A CN112183428A (en) | 2020-10-09 | 2020-10-09 | Wheat planting area segmentation and yield prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112183428A true CN112183428A (en) | 2021-01-05 |
Family
ID=73948025
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011081232.6A Pending CN112183428A (en) | 2020-10-09 | 2020-10-09 | Wheat planting area segmentation and yield prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112183428A (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906627A (en) * | 2021-03-15 | 2021-06-04 | 西南大学 | Green pricklyash peel identification method based on semantic segmentation |
CN113075359A (en) * | 2021-03-18 | 2021-07-06 | 塔里木大学 | Walnut growth prediction system based on satellite and unmanned aerial vehicle remote sensing combination |
CN113537645A (en) * | 2021-08-23 | 2021-10-22 | 苏州憨云智能科技有限公司 | Soybean yield prediction method based on machine learning fusion satellite and weather data |
CN113592664A (en) * | 2021-08-05 | 2021-11-02 | 四川省农业科学院农业信息与农村经济研究所 | Crop production space prediction simulation method, equipment, model and storage medium |
CN114445797A (en) * | 2021-12-29 | 2022-05-06 | 深圳云天励飞技术股份有限公司 | Night driving vision auxiliary method and related equipment |
CN114463637A (en) * | 2022-02-07 | 2022-05-10 | 中国科学院空天信息创新研究院 | Winter wheat remote sensing identification analysis method and system based on deep learning |
CN115600771A (en) * | 2022-12-09 | 2023-01-13 | 中化现代农业有限公司(Cn) | Crop yield estimation method, device, equipment and storage medium |
CN115797788A (en) * | 2023-02-17 | 2023-03-14 | 武汉大学 | Multimodal railway design element remote sensing feature extraction method based on deep learning |
CN116485040A (en) * | 2023-06-13 | 2023-07-25 | 中国农业大学 | Seed vitality prediction method, system, electronic equipment and storage medium |
CN117058541A (en) * | 2023-08-03 | 2023-11-14 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101858971A (en) * | 2010-06-02 | 2010-10-13 | 浙江大学 | Rice yield remote sensing estimation method based on MODIS data |
CN109784300A (en) * | 2019-01-28 | 2019-05-21 | 中国平安财产保险股份有限公司四川分公司 | A kind of crops science survey production method and system |
CN109919395A (en) * | 2019-04-01 | 2019-06-21 | 安徽大学 | A kind of winter wheat yield monitoring method based on short cycle remote sensing area data |
CN110245327A (en) * | 2019-03-28 | 2019-09-17 | 国智恒北斗好年景农业科技有限公司 | A kind of yield of wheat remote sensing estimation method based on GF-1 data reconstruction |
CN110309985A (en) * | 2019-07-10 | 2019-10-08 | 北京师范大学 | A kind of crop yield prediction technique and system |
CN110309969A (en) * | 2019-06-28 | 2019-10-08 | 河南农业大学 | Based on the monitoring of the winter wheat Spring frost of Internet of Things and remote-sensing inversion and production prediction method |
CN110378521A (en) * | 2019-06-28 | 2019-10-25 | 河南农业大学 | The building and application of Henan northeast Yield Forecast of Winter Wheat model |
CN111062526A (en) * | 2019-12-09 | 2020-04-24 | 北京师范大学 | Winter wheat yield per unit prediction method and system |
CN111666815A (en) * | 2020-05-06 | 2020-09-15 | 武汉大学 | Automatic garlic planting information extraction method based on Sentinel-2 remote sensing image |
-
2020
- 2020-10-09 CN CN202011081232.6A patent/CN112183428A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101858971A (en) * | 2010-06-02 | 2010-10-13 | 浙江大学 | Rice yield remote sensing estimation method based on MODIS data |
CN109784300A (en) * | 2019-01-28 | 2019-05-21 | 中国平安财产保险股份有限公司四川分公司 | A kind of crops science survey production method and system |
CN110245327A (en) * | 2019-03-28 | 2019-09-17 | 国智恒北斗好年景农业科技有限公司 | A kind of yield of wheat remote sensing estimation method based on GF-1 data reconstruction |
CN109919395A (en) * | 2019-04-01 | 2019-06-21 | 安徽大学 | A kind of winter wheat yield monitoring method based on short cycle remote sensing area data |
CN110309969A (en) * | 2019-06-28 | 2019-10-08 | 河南农业大学 | Based on the monitoring of the winter wheat Spring frost of Internet of Things and remote-sensing inversion and production prediction method |
CN110378521A (en) * | 2019-06-28 | 2019-10-25 | 河南农业大学 | The building and application of Henan northeast Yield Forecast of Winter Wheat model |
CN110309985A (en) * | 2019-07-10 | 2019-10-08 | 北京师范大学 | A kind of crop yield prediction technique and system |
CN111062526A (en) * | 2019-12-09 | 2020-04-24 | 北京师范大学 | Winter wheat yield per unit prediction method and system |
CN111666815A (en) * | 2020-05-06 | 2020-09-15 | 武汉大学 | Automatic garlic planting information extraction method based on Sentinel-2 remote sensing image |
Non-Patent Citations (1)
Title |
---|
郭锐 等: "山东省冬小麦单产监测与预报方法研究", 《农业机械学报》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906627A (en) * | 2021-03-15 | 2021-06-04 | 西南大学 | Green pricklyash peel identification method based on semantic segmentation |
CN113075359A (en) * | 2021-03-18 | 2021-07-06 | 塔里木大学 | Walnut growth prediction system based on satellite and unmanned aerial vehicle remote sensing combination |
CN113592664B (en) * | 2021-08-05 | 2023-07-18 | 四川省农业科学院农业信息与农村经济研究所 | Crop production space prediction simulation method, device, model and storage medium |
CN113592664A (en) * | 2021-08-05 | 2021-11-02 | 四川省农业科学院农业信息与农村经济研究所 | Crop production space prediction simulation method, equipment, model and storage medium |
CN113537645A (en) * | 2021-08-23 | 2021-10-22 | 苏州憨云智能科技有限公司 | Soybean yield prediction method based on machine learning fusion satellite and weather data |
CN113537645B (en) * | 2021-08-23 | 2023-11-24 | 苏州憨云智能科技有限公司 | Soybean yield prediction method based on machine learning fusion satellite and weather data |
CN114445797A (en) * | 2021-12-29 | 2022-05-06 | 深圳云天励飞技术股份有限公司 | Night driving vision auxiliary method and related equipment |
CN114463637A (en) * | 2022-02-07 | 2022-05-10 | 中国科学院空天信息创新研究院 | Winter wheat remote sensing identification analysis method and system based on deep learning |
CN115600771A (en) * | 2022-12-09 | 2023-01-13 | 中化现代农业有限公司(Cn) | Crop yield estimation method, device, equipment and storage medium |
CN115797788A (en) * | 2023-02-17 | 2023-03-14 | 武汉大学 | Multimodal railway design element remote sensing feature extraction method based on deep learning |
CN115797788B (en) * | 2023-02-17 | 2023-04-14 | 武汉大学 | Multimodal railway design element remote sensing feature extraction method based on deep learning |
CN116485040A (en) * | 2023-06-13 | 2023-07-25 | 中国农业大学 | Seed vitality prediction method, system, electronic equipment and storage medium |
CN116485040B (en) * | 2023-06-13 | 2023-09-08 | 中国农业大学 | Seed vitality prediction method, system, electronic equipment and storage medium |
CN117058541A (en) * | 2023-08-03 | 2023-11-14 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
CN117058541B (en) * | 2023-08-03 | 2024-02-13 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112183428A (en) | Wheat planting area segmentation and yield prediction method | |
CN111598019B (en) | Crop type and planting mode identification method based on multi-source remote sensing data | |
Domenikiotis et al. | Early cotton yield assessment by the use of the NOAA/AVHRR derived Vegetation Condition Index (VCI) in Greece | |
Becker-Reshef et al. | A generalized regression-based model for forecasting winter wheat yields in Kansas and Ukraine using MODIS data | |
El Hajj et al. | Integrating SPOT-5 time series, crop growth modeling and expert knowledge for monitoring agricultural practices—The case of sugarcane harvest on Reunion Island | |
Jaafar et al. | Crop yield prediction from remotely sensed vegetation indices and primary productivity in arid and semi-arid lands | |
CN108458978B (en) | Sensitive waveband and waveband combination optimal tree species multispectral remote sensing identification method | |
CN114821362B (en) | Multi-source data-based rice planting area extraction method | |
CN111209871B (en) | Rape planting land remote sensing automatic identification method based on optical satellite image | |
CN109711102A (en) | A kind of crop casualty loss fast evaluation method | |
CN109063660B (en) | Crop identification method based on multispectral satellite image | |
CN115128013A (en) | Soil organic matter content space prediction evaluation method based on partition algorithm | |
CN114663489A (en) | Crop leaf area index remote sensing inversion method and system under constraint of space-time characteristics of land blocks | |
CN116824384A (en) | Soybean identification method based on standard curve | |
CN117197668A (en) | Crop lodging level prediction method and system based on deep learning | |
Xie | Combining CERES-Wheat model, Sentinel-2 data, and deep learning method for winter wheat yield estimation | |
Liu et al. | Open-air grape classification and its application in parcel-level risk assessment of late frost in the eastern Helan Mountains | |
Boschetti et al. | Estimation of rice production at regional scale with a Light Use Efficiency model and MODIS time series. | |
CN114694041A (en) | Hyperspectral identification method for cotton phytotoxicity and spider mite insect damage | |
Khodjaev et al. | Combining multiple UAV-Based indicators for wheat yield estimation, a case study from Germany | |
CN117152645A (en) | Wheat rust monitoring method based on unmanned aerial vehicle multispectral image depth characteristics | |
WO2023195863A1 (en) | Methods and systems for estimating crop yield from vegetation index data | |
CN115063610B (en) | Soybean planting area identification method based on Sentinel-1 and 2 images | |
Singla et al. | Sugarcane ratoon discrimination using LANDSAT NDVI temporal data | |
CN116124774A (en) | Method for predicting nitrogen content of canopy based on unmanned aerial vehicle spectrum multi-source data |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210105 |