CN112580595A - Double-cropping rice field remote sensing identification method based on Shapelet - Google Patents
Double-cropping rice field remote sensing identification method based on Shapelet Download PDFInfo
- Publication number
- CN112580595A CN112580595A CN202011598570.7A CN202011598570A CN112580595A CN 112580595 A CN112580595 A CN 112580595A CN 202011598570 A CN202011598570 A CN 202011598570A CN 112580595 A CN112580595 A CN 112580595A
- Authority
- CN
- China
- Prior art keywords
- shapelet
- rice field
- double
- time sequence
- cropping rice
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
- G06F16/258—Data format conversion from or to a database
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A double-cropping rice field remote sensing identification method based on Shapelet comprises the following steps: 1, preparing and preprocessing data to form a rice field vegetation index time sequence data set of a research area; 2, determining key points of a time sequence according to the double cropping rice growth period, and acquiring a plurality of subsequences of the vegetation index time sequence by taking the key points as endpoints to form a Shapelet candidate set; 3, carrying out importance evaluation on the candidate Shapelets, and selecting K optimal Shapelets; 4 Shapelet conversion. Calculating the distances between all time sequences and K Shapelets, and converting each time sequence into a vector consisting of K distances; 5, constructing a convolutional neural network classifier, and training the convolutional neural network by using a training sample; and 6, importing the converted time sequence into a classifier for learning, identifying the type of the rice field, and generating a double-cropping rice field distribution map. The invention adopts a Shapelet-based method to carry out remote sensing identification of the double cropping rice field, can effectively improve the identification precision of the double cropping rice field and provides an effective means for rapidly monitoring the planting area of the double cropping rice.
Description
Technical Field
The invention relates to the field of double cropping rice planting, in particular to a double cropping rice field remote sensing identification method based on Shapelet.
Background
The rice is one of the most important grain crops in China, 60% of people in China use rice as staple food, and the shortage of the yield and the area directly influences the national grain safety and social stability. Therefore, the development of double cropping rice production has an important guarantee effect on national food safety. Since the 70 and 80 th ages of the last century, the two-season rice paddy field planting system has gone through an iterative process from single-season rice to two-season rice and from two-season rice to single-season rice. In recent years, the trend of 'double-to-single' is more obvious under the influence of rural labor shortage and low economic benefit of double-cropping rice, and the method has obvious influence on the food yield in China. Therefore, the method develops the research of the rice planting system, accurately identifies the rice planting change, and has important significance for formulating regional grain policies and ensuring national grain safety.
The rice full growth period comprises 7 main growth periods, such as a seedling period, a tillering period, an elongation period, a booting period, a heading period, a filling period, a mature period and the like, wherein each growth period has different spectral characteristics. By extracting the long-term potential characteristics of different growth periods of rice through multi-temporal remote sensing data, the difference of the spectral characteristics of the rice and other crops can be identified, and the accurate identification of the rice planting system is realized. The identification of the rice planting system in the regional scale by adopting remote sensing time sequence data is a difficult point of the current research. Firstly, the time-space heterogeneity of the rice phenological characteristics on the regional scale is caused by the difference of rice varieties, local climate and terrain; secondly, cloud shadow problems, weather influences and noise introduced by the sensor cause distortion of remote sensing time sequence data, so that deviation of the phenological characteristics of the rice is caused.
Shapelet refers to a subsequence that is distinguishable in time series. The method based on the time series Shapelet has the following advantages in two aspects: (1) shapelet can fully illustrate the difference between the growth characteristics of crops such as rice and the like and the growth characteristics of other crops, has stronger identification performance and can be effectively used for crop identification; (2) shapelet compares the similarity between subsequences of the time series, both reducing the computational complexity and preserving sufficient growth characteristics to identify crop types. Therefore, the Shapelet-based time series classification method can effectively solve the problems existing in the current multi-temporal remote sensing data crop identification.
Disclosure of Invention
Aiming at the defects of the existing work, the invention provides a double-cropping rice field remote sensing identification method based on Shapelet.
The invention relates to a double-cropping rice field remote sensing identification method based on Shapelet, which comprises the following steps:
step 1: preparing and preprocessing data to form a rice field vegetation index time sequence data set of a research area;
step 2: determining a key point of a time sequence according to the double cropping rice growth period, and acquiring a plurality of subsequences of the vegetation index time sequence by taking the key point as an endpoint to form a Shapelet candidate set;
and step 3: carrying out importance evaluation on the candidate Shapelets, and selecting K optimal Shapelets;
and 4, step 4: shapelet conversion. Calculating the distances between all time sequences and the K Shapelets, taking each time sequence as an object, taking the distance between the time sequence and the K Shapelets as K attributes of each object, and converting each time sequence into a vector consisting of K attribute values;
and 5: constructing a convolutional neural network classifier, and training a convolutional neural network by using a training sample;
step 6: and (4) leading the converted time sequence into a classifier, identifying the rice field type, and generating a double-cropping rice distribution map.
In the step 1, the concrete steps are as follows:
and collecting the remote sensing data with higher quality, and obtaining a vegetation index time sequence data set of the research area through preprocessing such as geometric correction, radiation correction, combination and cutting, S-G filtering processing and the like.
In the step 2, the concrete steps are as follows:
and determining key time points according to transplanting, tillering, jointing, booting, heading aligning and maturing time of the double cropping rice. And extracting the sub-sequence of the vegetation index time sequence by a sliding window method, indirectly controlling the length of the sub-sequence by changing and adjusting the size of the window, and deleting the sub-sequence which does not contain the key point to form a Shapelet candidate set.
In the step 3, the concrete steps are as follows:
1) determining the correlation of 2 Shapelets by using a Pearson correlation coefficient, and generating a correlation matrix of a Shapelet candidate set;
2) using a formulaIG= Entropy – E 1 – E 2Calculating the information gain of each candidate Shapelet and sequencing;
3) screening K optimal Shapelets according to the information gain result and the correlation matrix of the Shapelets;
wherein the content of the first and second substances,IGthe gain of the information representing the Shapelet,Entropyin order to be the entropy of the signal,,Cwhich represents the total of the categories of the content,n i the number of samples for each of the categories,Nin order to be the total number of samples,N 1、N 2andE 1、E 2respectively, the number of 2 mutually disjoint subset samples and the entropy into which the data set is partitioned by sharelet. The value of K is 1/2-2/3 of the number of key points.
In the step 5, the concrete steps are as follows:
in Python IDLE, calling scimit-learn packet to construct convolutional neural network classifier, importing training data, and training classifier.
In the step 6, the concrete steps are as follows:
and importing the converted time sequence into a classifier, identifying the rice field, and calling gdal in Python IDLE to generate a GTiff file.
The scheme of the invention has at least the following beneficial effects:
the invention provides a double-cropping rice field remote sensing identification method based on Shapelet, which comprises the steps of converting rice field vegetation index grid data into vegetation index time sequence data, screening Shapelet by adopting a Pearson correlation coefficient matrix and an information gain method, constructing a neural network classifier, realizing effective classification of the rice field vegetation index time sequence data, and identifying the double-cropping rice field. The method of Shapelet can effectively overcome the difference of vegetation index curves caused by the difference of rice varieties and management, and increase the accuracy of double cropping rice identification; the Shapelet screening scheme adopting the Pearson correlation coefficient and the information gain not only reduces the calculated amount, but also keeps the growth characteristics of identifying the double-cropping rice field; the classification method based on the neural network is also beneficial to improving the identification precision of the double cropping rice field.
Drawings
FIG. 1 is a flow chart of the double cropping rice field remote sensing identification method based on Shapelet of the present invention.
Fig. 2 is a schematic diagram of the extraction of key points in the time series proposed by the present invention.
FIG. 3 shows 8 optimal Shapelets extracted from the double cropping rice field identification in 2018 New Ganjian.
Fig. 4 is a distribution diagram of double cropping rice field in 2018, new dry county.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to describe the double-cropping rice field remote sensing identification method based on Shapelet in detail, definitions of time series, subsequence, time series-subsequence distance, split point, entropy, information gain, Shapelet and the like are introduced.
Definition 1: time series and subsequences: for a length ofmTime series ofT,TEach subsequence of (1)SIs fromTContinuous interception starting at an arbitrary position. Wherein the length of the setting subsequence islThe starting point ispThen the subsequence is represented asS={t p , t p+1 , …, t p l+-1},1≤P≤m-l+1。
Definition 2: time series to subsequence distance: the distance between a shorter time series sub-sequence and a longer time series needs to be calculated from time series to sub-sequence. The short time series are slid over the long series to obtain the shortest distance between the two, called time series to sub-series distance, and defined as。
Definition 3: splitting point: one split point is a tuple (s, t),sA sub-sequence is represented that is,ta threshold value representing the distance. A splitting point: (s, t) Data setDIs divided into two disjoint subsets, whereinD left = {x: x ∈D, subdist(s, x) ≤ t},D right = { x: x ∈D, subdist(s, x) > t}。
Definition 4: entropy: hypothesis data setDIn which comprisesNA time series of bars, the number of classes beingCClass IIIC i In a data setDTherein is provided withn i An example, andn 1 + n 2 + … + n c = Nthen data setDThe entropy of (d) is defined as:。
definition 5: information gain: a splitting point: (s, t) An information gain ofI(s, t) = E(D) – E(D left ) – E(D right )。
Definition 6: data setDOne Shapelet in is composed ofDAn example ofTA subsequence ofSAnd a threshold value of the split pointtThe formed tuple (s, t) It attempts to assemble the data setDDivided into two different groups. A sharelet is a subsequence that can best distinguish between different classes among all split points.
So that the manner in which the above recited features of the present invention can be understood and readily understood, a more particular description of the invention, briefly summarized above, may be had by reference to embodiments, some of which are illustrated in the appended drawings, wherein:
as shown in FIG. 1, the remote sensing identification method for double cropping rice field based on Shapelet of the invention comprises the following steps:
in the first step, EVI (enhanced vegetation index) data (MOD 13Q 1) of rice field MODIS in the study area were collected. The 23-phase EVI data of 2018 years in the data set was selected to construct the EVI time series set. Data was downloaded from the NASA website (https:// ladsweb. nascom. NASA. gov /). And adopting MRT (MODIS reproduction tool) software to perform mosaic, shearing, projection and resampling on MOD13Q1 data, and adopting S-G filtering to reconstruct EVI data to generate an EVI time sequence data set of the research area. In Python IDLE, gdal is called to read the EVI data and store as a two-dimensional array. The spatial resolution of the MOD13Q1 data is 250m, mixed pixels of double cropping rice and single cropping rice are generated in the rice field type learning process, and therefore, the rice field is divided into three types of double cropping rice, single cropping rice and single and double mixed cropping rice in the embodiment;
secondly, determining the key point of the time sequence (see fig. 2) according to the growth period (see table 1) of the double cropping rice and the time point of MOD13Q1 data, acquiring the subsequence of the time sequence in a time window sliding mode, and generating a Shapelet candidate set;
TABLE 1 double cropping Rice growth period
Variety of (IV) C | Transplanting period | Tillering stage | Jointing stage | Heading period | Full heading period | Maturity stage |
Early rice | 4.20 | 5.8 | 5.28 | 6.6 | 6.12 | 7.12 |
Late rice | 7.20 | 8.6 | 8.28 | 9.7 | 9.19 | 11.2 |
Thirdly, firstly, a Pearson correlation matrix of a Shapelet candidate set is calculated, and subsequences with the same length and Shapelets with shorter lengths are selected from longer Shapelets to calculate Pearson coefficients; according to the formulaIG= Entropy – E 1 – E 2Calculating the information gain of each candidate Shapelet and sequencing; selecting 8 Shapelets with the highest information gain (see FIG. 3);
fourthly, adopting a formula according to the 8 optimal Shapelets generated in the third stepCalculating the distance between each time series and the 8 Shapelets, converting the time series into a vector with 8 attribute values,T i =[dis i 1,,dis i 2,,…,dis i 8,]. The representing method not only retains the advantages of Shapelet time sequence data classification, but also reduces the data volume, simplifies the operation, reduces the operation cost, and simultaneously retains the precedence relationship of time sequence data, thereby ensuring the classification precision;
and fifthly, calling scimit-spare packets in Python IDLE to construct a convolutional neural network classifier, importing training data and training the classifier. The relevant parameters of the convolutional neural network are set as follows: 3 convolution layers with the size of 3 multiplied by 3 are arranged, and each layer is provided with 64, 128 and 256 filters; setting 3 pooling layers with the size of 3 multiplied by 3; setting a full connection layer with the size of 1 multiplied by 3, and adopting a Softmax function for classification; the learning rate was set to 0.01; performing parameter updating by adopting an SGD (Stochastic Gradient Description) function to reduce the value of a loss function, wherein in an SFG (sequential Forward production) optimization algorithm, a batch value is set to be 64, a weight attenuation factor is set to be 0.0001, and an iteration value is set to be 22500;
and sixthly, introducing the converted time sequence into the convolutional neural network classifier, identifying the rice field type, calling gdal in Python IDLE, converting the classification result into a GTiff file, wherein the classification result is shown in FIG. 4, the classification precision is 0.91%, and the kappa coefficient is 0.94.
In addition to the above embodiments, the present invention may have other embodiments. All technical solutions formed by adopting equivalent substitutions or equivalent transformations fall within the protection scope of the claims of the present invention.
Claims (7)
1. A double-cropping rice field remote sensing identification method based on Shapelet is characterized in that: the method comprises the following steps:
step 1: preparing and preprocessing data to form a rice field vegetation index time sequence data set of a research area;
step 2: determining a key point of a time sequence according to the double cropping rice growth period, and acquiring a plurality of subsequences of the vegetation index time sequence by taking the key point as an endpoint to form a Shapelet candidate set;
and step 3: carrying out importance evaluation on the candidate Shapelets, and selecting K optimal Shapelets;
and 4, step 4: shapelet conversion; calculating the distances between all time sequences and the K Shapelets, taking each time sequence as an object, taking the distance between the time sequence and the K Shapelets as K attributes of each object, and converting each time sequence into a vector consisting of K attribute values;
and 5: constructing a convolutional neural network classifier, and training a convolutional neural network by using a training sample;
step 6: and (4) leading the converted time sequence into a classifier, identifying the type of the rice field, and generating a double-cropping rice field distribution map.
2. The double cropping rice field remote sensing identification method based on Shapelet as claimed in claim 1, characterized in that: in the step 1, the remote sensing data of the research area is subjected to geometric correction, radiation correction, mosaic, cutting, projection and resampling, vegetation indexes are calculated, and an S-G filtering is adopted to generate a vegetation index time sequence data set of the research area in Python IDLE.
3. The double cropping rice field remote sensing identification method based on Shapelet as claimed in claim 1, characterized in that: in step 2, determining key points of the time sequence according to the double cropping rice growing period, taking the key points as endpoints, and obtaining a plurality of subsequences of the vegetation index time sequence through a sliding time window to form a Shapelet candidate set.
4. The double cropping rice field remote sensing identification method based on Shapelet as claimed in claim 1, characterized in that: the step 3 comprises the following steps:
firstly, calculating a Pearson correlation matrix of a Shapelet candidate set;
second, according to the formulaIG= Entropy –E1– E 2Calculating the information gain of each candidate Shapelet and sequencing;
thirdly, the K Shapelets with the highest information gain are selected by combining the Pearson correlation coefficient matrix.
5. A substrate according to claim 1The remote sensing identification method for the double-cropping rice field in Shapelet is characterized by comprising the following steps: in step 4, Shapelet conversion is carried out; according to the formulaAnd calculating the distances between all the time series and the K Shapelets, taking each time series as an object, taking the distances between the time series and the K Shapelets as K attributes of each object, and converting each time series into a vector consisting of K attribute values.
6. The double cropping rice field remote sensing identification method based on Shapelet as claimed in claim 1, characterized in that: in the Python IDLE, calling scimit-spare packet to construct a convolutional neural network classifier, importing training data and training the classifier.
7. The double cropping rice field remote sensing identification method based on Shapelet as claimed in claim 1, characterized in that: and 6, in a Python IDLE, importing the vegetation index time sequence of the rice field to be identified after Shapelet conversion into a convolutional neural network classifier, identifying the double cropping rice field, and calling a gdal packet to generate GTiff type data and a double cropping rice field distribution map.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011598570.7A CN112580595A (en) | 2020-12-30 | 2020-12-30 | Double-cropping rice field remote sensing identification method based on Shapelet |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011598570.7A CN112580595A (en) | 2020-12-30 | 2020-12-30 | Double-cropping rice field remote sensing identification method based on Shapelet |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112580595A true CN112580595A (en) | 2021-03-30 |
Family
ID=75144164
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011598570.7A Pending CN112580595A (en) | 2020-12-30 | 2020-12-30 | Double-cropping rice field remote sensing identification method based on Shapelet |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112580595A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116563720A (en) * | 2023-07-12 | 2023-08-08 | 华中师范大学 | Single-double-season rice sample automatic generation method for cooperative optical-microwave physical characteristics |
CN117407733A (en) * | 2023-12-12 | 2024-01-16 | 南昌科晨电力试验研究有限公司 | Flow anomaly detection method and system based on countermeasure generation shapelet |
-
2020
- 2020-12-30 CN CN202011598570.7A patent/CN112580595A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116563720A (en) * | 2023-07-12 | 2023-08-08 | 华中师范大学 | Single-double-season rice sample automatic generation method for cooperative optical-microwave physical characteristics |
CN116563720B (en) * | 2023-07-12 | 2023-10-03 | 华中师范大学 | Single-double-season rice sample automatic generation method for cooperative optical-microwave physical characteristics |
CN117407733A (en) * | 2023-12-12 | 2024-01-16 | 南昌科晨电力试验研究有限公司 | Flow anomaly detection method and system based on countermeasure generation shapelet |
CN117407733B (en) * | 2023-12-12 | 2024-04-02 | 南昌科晨电力试验研究有限公司 | Flow anomaly detection method and system based on countermeasure generation shapelet |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111598019B (en) | Crop type and planting mode identification method based on multi-source remote sensing data | |
CN110175931B (en) | Method for rapidly extracting crop planting area and phenological information in large range | |
CN104572886B (en) | The financial time series similarity query method represented based on K line charts | |
CN112580595A (en) | Double-cropping rice field remote sensing identification method based on Shapelet | |
CN115829162B (en) | Crop yield prediction method, device, electronic equipment and medium | |
CN109063660B (en) | Crop identification method based on multispectral satellite image | |
CN109800921B (en) | Regional winter wheat yield estimation method based on remote sensing phenological assimilation and particle swarm optimization | |
CN113642464B (en) | Time sequence remote sensing image crop classification method combining TWDTW algorithm and fuzzy set | |
CN103971199B (en) | A kind of remote sensing ranking method of a wide range of Grain Growth Situation | |
CN116227758B (en) | Agricultural product maturity prediction method and system based on remote sensing technology and deep learning | |
CN116824384A (en) | Soybean identification method based on standard curve | |
CN114494909B (en) | Method and system for generating spatial distribution diagram of soybean growing season | |
CN115861831A (en) | Time series remote sensing data crop identification method based on crop phenological knowledge | |
CN115861629A (en) | High-resolution farmland image extraction method | |
CN113052433B (en) | Crop unit yield estimation method based on key time phase and farmland landscape characteristic parameters | |
CN114332492B (en) | Vegetation feature extraction and selection method based on long-time sequence | |
CN112215184B (en) | Method for detecting yield of camellia oleifera fruit trees based on three-dimensional laser scanner | |
CN115937705A (en) | South single-season and double-season rice identification method based on multi-source data and phenological characteristics | |
CN115205701A (en) | Method for extracting sample based on rice phenological characteristics | |
CN114663791A (en) | Branch recognition method for pruning robot in unstructured environment | |
CN111695533B (en) | Remote sensing mapping method for automatically monitoring planting areas of winter wheat and summer corn year by year | |
CN115495988B (en) | Soil remote sensing inversion method based on optimal time window selection | |
CN116843368B (en) | Marketing data processing method based on ARMA model | |
CN114782835B (en) | Crop lodging area proportion detection method and device | |
CN116485784B (en) | Winter idle field remote sensing extraction method based on time sequence NDVI |
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 |