CN115861831A - Time series remote sensing data crop identification method based on crop phenological knowledge - Google Patents
Time series remote sensing data crop identification method based on crop phenological knowledge Download PDFInfo
- Publication number
- CN115861831A CN115861831A CN202211423961.4A CN202211423961A CN115861831A CN 115861831 A CN115861831 A CN 115861831A CN 202211423961 A CN202211423961 A CN 202211423961A CN 115861831 A CN115861831 A CN 115861831A
- Authority
- CN
- China
- Prior art keywords
- crop
- remote sensing
- phenological
- time
- sensing data
- 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
Images
Abstract
The invention relates to a crop identification method based on time series remote sensing data of crop phenology knowledge, which comprises the following steps: s2, constructing a time sequence remote sensing data characteristic extraction model based on an LSTM network, and integrating a full convolution neural network in the LSTM network; s3, constructing a neural network integrating the phenological knowledge and time sequence remote sensing data, namely a PST-LSTM model, based on a multi-modal learning framework; s4, acquiring training sample data, and training and optimizing parameters of the PST-LSTM model; and S5, identifying the remote sensing image to be identified based on the trained PST-LSTM model. The invention overcomes the problems of large dependence of areas and data and weak mobility in crop identification in the traditional method, and improves the identification precision and the applicability of crops.
Description
Technical Field
The invention relates to the field of automatic classification of crops by remote sensing images, in particular to a crop identification method based on time series remote sensing data of crop phenological knowledge.
Background
Grain safety is used as an important basis of national safety, and scientific and effective development of crop planting is a key means for guaranteeing grain safety. The spatial distribution information of crop planting has important significance on aspects such as agricultural modern production, planting structure spatial configuration optimization and the like, and provides basic data support for individual farmers and related departments to develop agricultural digital management. The existing crop planting space distribution data acquisition method is large in manual dependence, time-consuming, labor-consuming and poor in timeliness, and the intelligent agricultural requirements are difficult to meet. The automatic classification of crop planting based on satellite remote sensing data is an effective scheme for solving the problem. At present, remote sensing data crop planting automatic classification methods mostly utilize a single phenological feature rule or a conventional time series classification method. Few studies have comprehensively utilized the phenological, temporal and spatial characteristics of crops, and particularly integrated multimodal information in a unified framework. The existing identification method has the problems of large dependence on regions and data and weak mobility, and the large-scale and cross-region application of remote sensing data crop automatic classification is limited.
Disclosure of Invention
In view of the above, the present invention aims to provide a crop identification method based on time series remote sensing data of crop phenological knowledge, which overcomes the problems of large area and data dependence and weak mobility in crop identification in the conventional method, and improves the accuracy and applicability of crop identification.
In order to achieve the purpose, the invention adopts the following technical scheme:
a crop identification method based on time series remote sensing data of crop phenology knowledge comprises the following steps:
s1, constructing a phenological index of a key growth period according to a crop growth curve, and extracting phenological characteristics of crops;
s2, constructing a time sequence remote sensing data feature extraction model based on the LSTM network, and integrating a full convolution neural network in the LSTM network;
s3, constructing a neural network integrating the phenological knowledge and time sequence remote sensing data, namely a PST-LSTM model, based on a multi-modal learning framework;
s4, acquiring training sample data, and training and optimizing parameters of the PST-LSTM model;
and S5, identifying the remote sensing image to be identified based on the trained PST-LSTM model.
Further, the step S1 specifically includes:
step S11: acquiring a time sequence growth curve of the target crop index, and fitting the growth curve by using a GAM method;
step S12: acquiring phenological information of target crops in key phenological periods such as a sowing season, a mature season and a harvesting season, and performing characteristic analysis on the crop index time sequence curve according to the phenological information;
step S13: and constructing a phenological characteristic index according to the characteristic of the crop index time sequence curve, and extracting a crop phenological variable.
Further, the step S2 specifically includes: constructing a remote sensing data time sequence feature extractor based on an LSTM network, and extracting the time feature of crop growth; on the basis of an LSTM network, a remote sensing data spatial feature extractor is constructed based on a full convolution neural network module, and spatial features of crop types are extracted.
Further, in the LSTM network, a remote sensing data time sequence feature extractor is constructed to extract time features of crop growth, specifically:
(1) Constructing a bidirectional feature fusion module, wherein the module consists of 2 encoders and 1 series layer, and each encoder comprises a space-time convolution layer, a normalization layer, an activation function and an attention mechanism layer;
(2) Extracting time characteristics and phenological characteristics of an input data set based on a bidirectional characteristic fusion module, and performing bidirectional characteristic fusion processing;
(3) And selecting the LSTM as a basic network, and extracting the time characteristics of crop growth by combining an attention mechanism to obtain deep time domain information of the fusion characteristics.
Further, on the basis of the LSTM network, a remote sensing data spatial feature extractor is constructed based on a full convolution neural network module, and spatial features of crop types are extracted, specifically:
(1) Carrying out dimension replacement on the remote sensing time sequence data to convert the remote sensing time sequence data from (N, Q, M) to (N, M, Q); wherein N is the total number of samples, Q is the maximum time step, and M is the number of variables processed in each time step;
(2) And constructing a full convolution neural network module, extracting the spatial characteristics of the crop type, wherein the full convolution neural network module consists of 3 encoders and 1 global average pooling layer, each encoder comprises a space-time convolution layer, a normalization layer and an activation function, and the first two encoders are ended by an attention mechanism.
Further, the step S3 specifically includes:
step S31: constructing a multi-mode learning framework, and integrating phenological knowledge and time sequence remote sensing data;
step S32: and on the basis of a multi-mode learning framework, the phenological, temporal and spatial characteristics output by the LSTM module and the full convolution neural network module are fused, and the classified output is finished by using a full connection layer.
Further, step S4 specifically includes:
step S41: based on the high-resolution optical image, training sample data is collected by using a visual interpretation method;
step S42: constructing training sample data, assigning a target crop label to be 1, assigning a non-target crop label to be 0, and setting the dimension of the time sequence data set to be (N, Q) t ×M) The dimensionality of the data set of the physical variables is (N, Q) p ) N is the total number of samples, Q t For the time series data set maximum time step, M is the number of variables processed in each time step, Q p The total number of the phenological variables;
step S43: training the model based on a training sample set, wherein an Adam optimizer is adopted in training, the initial learning rate is set to be Lr, the training batch is set to be B, and the iteration number is set to be R;
step S44: analyzing the influence of different phenological variable combinations on the recognition result, determining the optimal combination, obtaining a corresponding model weight file, and obtaining a trained PST-LSTM model;
compared with the prior art, the invention has the following beneficial effects:
1. the invention constructs a unified learning framework based on multi-mode learning, integrates the phenological knowledge and time sequence remote sensing data, automatically fuses and extracts phenological-time-space characteristics of crops, and greatly improves the recognition capability of a model to the crops. The method can solve the problems of a large amount of false extraction, extraction omission and serious fragmentation in crop identification in the traditional method, and improves the mobility of the model.
2. The invention further improves the crop identification precision by automatically combining the phenological knowledge and the time series remote sensing data, and provides technical support for the production management of the agricultural department.
Drawings
FIG. 1 is a schematic flow chart of a method according to an embodiment of the present invention.
FIG. 2 is a diagram showing the structure of a phenological spatiotemporal-short memory model (PST-LSTM) according to an embodiment of the present invention.
Fig. 3 is a diagram of the structure of LSTM unit in the model of the embodiment of the present invention.
Fig. 4 is a planting period diagram of a target crop extracted according to an embodiment of the present invention.
Fig. 5 is a schematic diagram of time series curves and index construction of the target crop SAR image VH and VV indexes extracted in the embodiment of the present invention.
FIG. 6 is a detailed diagram of a part of the recognition result according to the embodiment of the present invention.
FIG. 7 is a detailed diagram of migration results according to an embodiment of the present invention.
Detailed Description
The invention is further explained by the following embodiments in conjunction with the drawings.
In the embodiment, a Sentinel-1A time series SAR remote sensing image of a research area is obtained, and preprocessing operation is performed on the image, wherein the preprocessing step comprises operations of track file correction, radiation correction, terrain correction, speckle filtering and the like.
Referring to fig. 1, the invention provides a crop identification method based on time series remote sensing data of crop phenological knowledge, which comprises the following steps:
step S1: constructing a phenological index of a key growth period according to a tobacco crop growth curve, and extracting crop phenological characteristics;
step S2: constructing a remote sensing data time sequence feature extractor based on a long and short memory neural network model (LSTM), and extracting the time feature of crop growth;
and step S3: on the basis, a remote sensing data spatial feature extractor is constructed based on a full convolution neural network module (FCN), and spatial features of crop types are extracted;
and step S4: based on a multi-mode learning framework, integrating the phenological knowledge and time sequence remote sensing data, and creating a neural network which integrates phenological, time and space, namely a phenological space-time-long memory model (PST-LSTM);
step S5: and acquiring training sample data, and training and optimizing parameters of the PST-LSTM model.
Step S6: carrying out automatic identification on crops in different regions based on the trained PST-LSTM model, and verifying the effectiveness and the mobility of the model;
in this embodiment, the step S1 specifically includes the following steps:
step S11: acquiring time sequence growth curves of indexes such as VH and VV of the tobacco crops based on SAR time sequence data, and fitting the growth curves by using a GAM method;
step S12: acquiring phenological information of the tobacco in key phenological periods such as a film covering period, a transplanting period, a topping period, a growing period and a mature harvesting period, and performing characteristic analysis on time sequence curves such as VH and VV according to the phenological information:
the tobacco VH value is reduced due to the effect of the greenhouse film on the tobacco during the film mulching period. After the transplanting period, the tobacco grows, the leaves are enlarged, and the VH value is increased. In the topping period, the tobacco growth is limited by the top advantages, the VH curve has local minimum values, after topping, the tobacco continues to grow, and the VH value continuously rises. By the mature harvest period, tobacco begins to be harvested, and the VH value is suddenly reduced. The VV backscattering coefficient change of the tobacco in the phenological period has the same trend as the VH coefficient change;
step S13: according to the characteristics of the tobacco VH and VV time sequence curves, constructing phenological characteristic indexes V1, V2 and V3, and extracting crop phenological variables:
firstly, characteristic indexes V1 and V3 are constructed according to the phenological characteristics of tobacco in a mature harvesting period, wherein the V1 index is the sum of the slope of the periodic VH image in the mature harvesting period of the tobacco.
The V3 index is the difference between VV values at the end of tobacco harvest and at the beginning of tobacco harvest.
V3＝VV End of harvest -VV At the beginning of harvesting
Secondly, constructing a characteristic index V2 according to the phenological characteristics that the tobacco VH value reaches a local minimum value before the transplanting period and at the topping period, wherein V2 is the difference between the two minimum values.
V2＝VH Minimum value of capping period -VH Minimum before transplanting period
In this embodiment, in step S11, the GAM is defined as follows:
GAM is a multivariate regression model in a non-parametric form, and has great flexibility, and the expression is as follows:
g(u)＝β 0 +s 1 (X 1 )+s 2 (X 2 )+…+s n (X n )
in the formula, s i (. Cndot.) is a nonparametric smooth function with a nuclear functionNumber, smooth spline function, etc.
In this embodiment, the step S2 specifically includes the following steps:
step S21: constructing a bidirectional feature fusion module (BIFFM), wherein the BIFFM is composed of 2 encoders and 1 series layer, and each encoder comprises a space-time convolution layer, a normalization layer, an activation function and an attention mechanism layer;
step S22: extracting time characteristics and phenological characteristics of the input data set based on a bidirectional characteristic fusion module, and performing bidirectional characteristic fusion processing;
step S23: selecting LSTM as a basic network, acquiring deep time domain information of the fusion characteristics by combining an attention mechanism, and extracting the time characteristics of crop growth;
in this embodiment, in step S23, the LSTM is defined as follows:
the LSTM unit has a gating mechanism inside to determine the stay and update of information. The input data of the LSTM unit of the t time step is composed of the current input signal x t And previous time output signal h t-1 And (4) forming. In the LSTM, input data are output through a sigmod function, a tanh function and the like, and the specific calculation formula is as follows:
f t ＝σ(W f ·[h t-1 ，x t ]+b f )
i t ＝σ(W i ·[h t-1 ，x t ]+b i )
o t ＝σ(W o ·[h t-1 ，x t ]+b o )
in the formula (f) t 、i t And o t A forgetting gate, an input gate and an output gate; w is a group of f 、W i 、W o And W c Is a weighting factor; b f 、b i 、b o And b c Is a bias vector.
In this embodiment, the step S3 specifically includes the following steps:
step S31: carrying out dimensionality replacement on the remote sensing time sequence data to enable the remote sensing time sequence data to be converted from (N, Q, M) to (N, M, Q), wherein N is the total number of samples, Q is the maximum time step, M is the variable number processed in each time step, and when M is smaller than Q, dimensionality replacement operation can improve the performance of a model;
step S32: and on the basis of the step S31, constructing a full convolution neural network module and extracting the spatial characteristics of the crop type. The full convolution neural network module is composed of 3 encoders and 1 global average pooling layer, each encoder comprises a space-time convolution layer, a normalization layer and an activation function, and the first two encoders are ended by an attention mechanism;
in this embodiment, step S4 specifically includes the following steps:
step S41: on the basis of the steps S2 and S3, a multi-modal learning framework is constructed, and the phenological knowledge and the time sequence remote sensing data are integrated;
step S42: on the basis of a multi-mode learning framework, the phenological, temporal and spatial characteristics output by the LSTM module and the full convolution neural network module are fused, and classified output is completed by using a full connection layer;
in this embodiment, step S5 specifically includes the following steps:
step S51: based on the Sentinel-2 image, training sample data is collected by using a visual interpretation method;
step S52: based on step S51, a training sample set is made, and the tobacco label is assigned as 1 and the non-tobacco label is assigned as 0. The time series dataset dimension is (N, Q) t X M) and the dimensionality of the data set of the objective variables is (N, Q) p ) N =71227 is the total number of samples, Q t =15 is timing sequenceMaximum time step of the data set, M =2 is the number of variables processed in each time step, Q p The total number of the phenological variables;
step S53: training the model based on a training sample set, wherein an Adam optimizer is adopted in training, and the initial learning rate is set to be 10 -3 Training batch is set to be 128, and iteration number is set to be 250;
step S54: analyzing the influence of different phenological variable combinations on the recognition result, determining the optimal combination, obtaining a corresponding model weight file, and obtaining a trained PST-LSTM model;
in particular, in the present embodiment, the 15-scene Sentinel-1A radar time-series remote sensing image from 2 to 7 months in 2020 is used as a research area in samming city ninghuan prefecture in fujian province, and VH and VV backscatter coefficient values are obtained after preprocessing. This example used 29234 tobacco samples and 41993 non-tobacco samples for model training.
As shown in fig. 2, for the structure diagram of the phenological space-time-long memory model (PST-LSTM) constructed in this embodiment, the model extracts the time characteristics of crop growth based on the long-short memory neural network module (LSTM), integrates the full convolution neural network module (FCN) on the LSTM, extracts the spatial characteristics of crop types, and finally fuses the phenological, time, and spatial characteristics in the multi-modal learning framework to obtain the spatial distribution of crops. The PST-LSTM input is time sequence remote sensing image data and a phenological characteristic data set, fusion of phenological characteristics and time characteristics is completed through a bidirectional characteristic fusion module (BiFFM), the BiFFM is composed of 2 encoders and 1 serial layer, and each encoder comprises a space-time convolution layer, a normalization layer, an activation function and an attention mechanism layer. And obtaining the time characteristics of the crop growth through the LSTM module by the fused characteristics. The FCN module is composed of 3 encoders and 1 global average pooling layer, each encoder comprises a space-time convolution layer, a normalization layer and an activation function, the first two encoders are ended by a CBAM attention mechanism, and time sequence remote sensing image data are processed by the FCN module to obtain spatial characteristics of crop types. And finally, the climate, time and space characteristics of the fused crops of the series layers are utilized to complete the identification of the target crops.
As shown in FIG. 5, constructed phenological characteristic indexes V1, V2 and V3 are shown in the present example according to the tobacco VH and VV time-series curve characteristics. The V1 index is the sum of the VH gradual slopes during the tobacco mature harvest. The V2 index is the difference between the minimum VH values of tobacco at 2 months (before the transplanting period) and at 4 months (topping period). The V3 index is the difference between VV values at the end of tobacco harvest and at the beginning of tobacco harvest.
Fig. 6 is a partial detail view of tobacco identified in ninghuan county, san ming city, fujian province according to this embodiment. In this embodiment, 3 different sets of physical and environmental variable combinations are set for model training, specifically, V1+ V2, V1+ V3, and V1+ V2+ V3. As can be seen from the figure, the tobacco identification results added with different climatic variables have larger difference, but most tobacco planting areas can be still identified, and the identification precision is higher. Through precision verification, the model weight of the V1+ V3 combination is optimal, and the overall precision reaches 94.95%, so that the model is selected as a migration prediction model.
Fig. 7 is a detailed diagram of migration results in pucheng prefecture, longyan prefecture, and Anfu prefecture, jian prefecture, of Fujian province, according to an embodiment of the present invention. As can be seen from the figure, most of the tobacco-growing areas in three areas were identified using the proposed method, with higher mobility.
The above description is only a preferred embodiment of the present invention, and all equivalent changes and modifications made in accordance with the claims of the present invention should be covered by the present invention.
Claims (7)
1. A crop identification method based on time series remote sensing data of crop phenology knowledge is characterized by comprising the following steps:
s1, constructing a phenological index of a key growth period according to a crop growth curve, and extracting phenological characteristics of crops;
s2, constructing a time sequence remote sensing data feature extraction model based on the LSTM network, and integrating a full convolution neural network in the LSTM network;
s3, constructing a neural network integrating the phenological knowledge and time sequence remote sensing data, namely a PST-LSTM model, based on a multi-modal learning framework;
s4, acquiring training sample data, and training and optimizing parameters of the PST-LSTM model;
and S5, identifying the remote sensing image to be identified based on the trained PST-LSTM model.
2. The crop recognition method based on the crop phenological knowledge time series remote sensing data as claimed in claim 1, wherein the step S1 is specifically:
step S11: acquiring a time sequence growth curve of the target crop index, and fitting the growth curve by adopting a GAM method;
step S12: acquiring phenological information of target crops in key phenological periods such as a sowing season, a mature season and a harvesting season, and performing characteristic analysis on the crop index time sequence curve according to the phenological information;
step S13: and constructing a phenological characteristic index according to the characteristic of the crop index time sequence curve, and extracting a crop phenological variable.
3. The crop recognition method based on the crop phenological knowledge time series remote sensing data as claimed in claim 1, wherein the step S2 is specifically: constructing a remote sensing data time sequence feature extractor based on an LSTM network, and extracting the time feature of crop growth; on the basis of an LSTM network, a remote sensing data spatial feature extractor is constructed based on a full convolution neural network module, and spatial features of crop types are extracted.
4. The crop recognition method based on the crop phenological knowledge time series remote sensing data as claimed in claim 3, wherein the LSTM network is used to construct a remote sensing data time series feature extractor to extract the time features of crop growth, specifically:
(1) Constructing a bidirectional feature fusion module, wherein the module consists of 2 encoders and 1 series layer, and each encoder comprises a space-time convolution layer, a normalization layer, an activation function and an attention mechanism layer;
(2) Extracting time characteristics and phenological characteristics of an input data set based on a bidirectional characteristic fusion module, and performing bidirectional characteristic fusion processing;
(3) And selecting the LSTM as a basic network, and extracting the time characteristics of crop growth by combining an attention mechanism to obtain deep time domain information of the fusion characteristics.
5. The crop recognition method based on crop phenology knowledge time series remote sensing data as claimed in claim 3, wherein the remote sensing data spatial feature extractor is constructed based on a full convolution neural network module on the basis of an LSTM network to extract spatial features of crop types, and specifically comprises:
(1) Carrying out dimension replacement on the remote sensing time sequence data to convert the remote sensing time sequence data from (N, Q, M) to (N, M, Q); wherein N is the total number of samples, Q is the maximum time step, and M is the number of variables processed in each time step;
(2) And constructing a full convolution neural network module, extracting the spatial characteristics of the crop type, wherein the full convolution neural network module consists of 3 encoders and 1 global average pooling layer, each encoder comprises a space-time convolution layer, a normalization layer and an activation function, and the first two encoders are ended by an attention mechanism.
6. The crop recognition method based on the crop phenological knowledge time series remote sensing data as claimed in claim 1, wherein the step S3 is specifically:
step S31: constructing a multi-mode learning framework, and integrating phenological knowledge and time sequence remote sensing data;
step S32: and on the basis of a multi-mode learning framework, the phenological, temporal and spatial characteristics output by the LSTM module and the full convolution neural network module are fused, and the classified output is finished by using a full connection layer.
7. The crop recognition method based on the time series remote sensing data of crop phenology knowledge according to claim 1, wherein the step S4 is specifically:
step S41: based on the high-resolution optical image, training sample data are collected by using a visual interpretation method;
step S42: constructing training sample data, assigning a target crop label to be 1, assigning a non-target crop label to be 0, and setting the dimension of the time sequence data set to be (N, Q) t X M) and the dimensionality of the data set of the objective variables is (N, Q) p ) N is the total number of samples, Q t For the time series data set maximum time step, M is the number of variables processed in each time step, Q p The total number of the climatic variables is;
step S43: training the model based on a training sample set, wherein an Adam optimizer is adopted in the training, the initial learning rate is set to be Lr, the training batch is set to be B, and the iteration number is set to be R;
step S44: analyzing the influence of different phenological variable combinations on the recognition result, determining the optimal combination, obtaining the corresponding model weight file, and obtaining the trained PST-LSTM model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211423961.4A CN115861831A (en) | 2022-11-15 | 2022-11-15 | Time series remote sensing data crop identification method based on crop phenological knowledge |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211423961.4A CN115861831A (en) | 2022-11-15 | 2022-11-15 | Time series remote sensing data crop identification method based on crop phenological knowledge |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115861831A true CN115861831A (en) | 2023-03-28 |
Family
ID=85663446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211423961.4A Pending CN115861831A (en) | 2022-11-15 | 2022-11-15 | Time series remote sensing data crop identification method based on crop phenological knowledge |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115861831A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116863341A (en) * | 2023-08-24 | 2023-10-10 | 中国农业科学院农业资源与农业区划研究所 | Crop classification and identification method and system based on time sequence satellite remote sensing image |
| CN117216444A (en) * | 2023-09-06 | 2023-12-12 | 北京林业大学 | Vegetation weather parameter extraction method and device based on deep learning |
-
2022
- 2022-11-15 CN CN202211423961.4A patent/CN115861831A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116863341A (en) * | 2023-08-24 | 2023-10-10 | 中国农业科学院农业资源与农业区划研究所 | Crop classification and identification method and system based on time sequence satellite remote sensing image |
| CN116863341B (en) * | 2023-08-24 | 2024-01-26 | 中国农业科学院农业资源与农业区划研究所 | Crop classification and identification method and system based on time sequence satellite remote sensing image |
| CN117216444A (en) * | 2023-09-06 | 2023-12-12 | 北京林业大学 | Vegetation weather parameter extraction method and device based on deep learning |
| CN117216444B (en) * | 2023-09-06 | 2024-04-19 | 北京林业大学 | Vegetation weather parameter extraction method and device based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110287944B (en) | Crop pest monitoring method based on multispectral remote sensing image of deep learning | |
| CN115861831A (en) | Time series remote sensing data crop identification method based on crop phenological knowledge | |
| Sharma et al. | Wheat crop yield prediction using deep LSTM model | |
| CN102668899B (en) | Crop planting mode recognition method | |
| CN107423850B (en) | Regional corn maturity prediction method based on time series LAI curve integral area | |
| CN111898922B (en) | Multi-scale crop yield assessment method and system | |
| CN115829162B (en) | Crop yield prediction method, device, electronic equipment and medium | |
| Acevedo-Opazo et al. | Assessment of an empirical spatial prediction model of vine water status for irrigation management in a grapevine field | |
| CN115222100A (en) | Crop yield prediction method based on three-dimensional cyclic convolution neural network and multi-temporal remote sensing image | |
| BAI et al. | Growth simulation and yield prediction for perennial jujube fruit tree by integrating age into the WOFOST model | |
| CN114140692A (en) | Fresh corn maturity prediction method based on unmanned aerial vehicle remote sensing and deep learning | |
| CN115018105A (en) | Winter wheat meteorological yield prediction method and system | |
| Veeragandham et al. | A review on the role of machine learning in agriculture | |
| CN111798132A (en) | Dynamic farmland monitoring method and system based on multi-source time sequence remote sensing depth coordination | |
| CN116543316A (en) | Method for identifying turf in paddy field by utilizing multi-time-phase high-resolution satellite image | |
| CN117114374B (en) | Intelligent agricultural irrigation management system based on weather prediction | |
| CN111275567B (en) | SENP cotton yield estimation method based on unmanned aerial vehicle image and estimation model construction method | |
| CN111579565B (en) | Agricultural drought monitoring method, system and storage medium | |
| CN116151454A (en) | Method and system for predicting yield of short-forest linalool essential oil by multispectral unmanned aerial vehicle | |
| CN116451823A (en) | Apple yield prediction method based on meteorological master control factors | |
| CN115661674A (en) | Crop irrigation information extraction method based on multisource satellite soil humidity data | |
| CN114821300A (en) | Method for judging plant protection operation time for preventing wheat scab based on multispectral remote sensing of unmanned aerial vehicle | |
| CN116842351B (en) | Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment | |
| CN116542403B (en) | Crop yield prediction method, device, electronic equipment and readable storage medium | |
| Mohan et al. | Temporal convolutional network based rice crop yield prediction using multispectral satellite 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 |