CN110503287A - Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form - Google Patents
Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form Download PDFInfo
- Publication number
- CN110503287A CN110503287A CN201810498045.4A CN201810498045A CN110503287A CN 110503287 A CN110503287 A CN 110503287A CN 201810498045 A CN201810498045 A CN 201810498045A CN 110503287 A CN110503287 A CN 110503287A
- Authority
- CN
- China
- Prior art keywords
- time
- meteorological
- data
- distance
- sequence
- 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
Classifications
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Mining
Abstract
The present invention provides a kind of method that multivariable rice Meteorological series carry out similarity analysis in the crucial phenological period, belongs to agricultural weather the field of data mining.For linear drift problem caused by the temperature minimax exceptional value in Meteorological series, sequence Wave crest and wave trough erroneous matching problem and phenological period cause different in size caused by mutation rainfall etc., introduce dynamic time warping algorithm, the design for increasing form and mixing gradient, to improve the accuracy and simplicity of measurement, while meteorological data variation tendency and exceptional value are highlighted on crop influence.Specific steps are as follows: according to rice phenological period data cutting Meteorological series；Meteorological series first derivative second dervative calculates；It is given a mark at a distance from the original series and gradient sequence at different levels of dynamic form penalty based on form constant factor；Weighted blend original series are at a distance from gradients at different levels；It weights each meteorological variables sequence distance and forms distance between polynary Meteorological series；It is ranked up according to distance metric marking.
Description
One, technical field
The invention belongs to agricultural weather data analysis fields, are agricultural weather domain knowledge and multivariate time series similitude
The crossing domain that the matching analysis algorithm be combined with each other.It is related to a kind of when can be used for non-linear, continuous, multivariable agricultural weather
Ordinal number can be used for meteorology of the rice within the time of infertility under different phenological according to similarity analysis is carried out within the crucial phenological period
Data similarity analysis.
Two, background technique
Meteorological index, such as the life of maximum temperature, minimum temperature, rainfall, sunshine time, radiant illumination to agricultural crops
It is long influence it is big, and include these indexs history meteorological data data can usually react somewhere for a period of time in weather
With the weather circumstance condition of crops, similarity measurement analysis is carried out according to the meteorological data data of historical years, it not only can be with
It realizes the supplement to meteorological observation missing data, while also can provide foundation for the automatic division of climatic province.Therefore, towards
The similarity analysis of agricultural weather data also becomes means commonly used by current agricultural weather data analysis.
The analysis of agricultural weather data similarity mainly by comparing maximum temperature in different regions or different time sections,
The difference of the distribution characteristics of the meteorological datas sample such as minimum temperature, rainfall, sunshine time and variation tendency reaches close
The purpose of degree sequence.Early stage meteorological data similarity analysis method is usual are as follows: utilizes mean temperature, standard deviation and anomaly etc.
Meteorology statistical indicator describes meteorological data variation characteristic, and the degree of approximation for comparing statistical indicator realizes similarity measurement, this kind of
Although method can effectively explain the feature of meteorological condition by statistical indicator on meteorology, when often considering one section
In meteorological data distribution characteristics and variation tendency itself that have ignored data, weaken the extremum of meteorological element to
Rice bring influences.Meteorological data temporal aspect is obvious simultaneously, has scholar to pass through the similarity distance degree of time series at present
Figureofmerit is to calculate the similarity degree of the meteorological condition of between year border or different regions or mutually from degree.
The core of the similarity analysis of agricultural weather time series data is the selection of distance metric method, often to maximum temperature,
The variables such as minimum temperature, illumination, rainfall carry out Euclidean distance calculating, then in the flat of each enterprising row distance of variable dimension
Sum.Local climate similitude divides mainly meteorological to objective area and surrounding area historical years using Euclidean distance
Data carry out similarity measurement, count the similar frequency, carry out region zones using similarity degree.Although this method is effectively simple,
But the meteorological variation characteristic of Various Seasonal is not accounted for.The highest temperature between year border is analyzed using Euclidean distance combination KNN
The variables meteorological data similarity degree such as degree, minimum temperature, rainfall and sunshine time utilizes the similar time meteorology number of history
According to unknown meteorological data in replacement target breeding time.But aiming at the problem that similarity measurement of agricultural weather data, this method
Often have ignored meteorological data exceptional value bring sequence data fluctuation abnormal problem, such as thermal extremes, extreme low temperature and drop
Accumulation mutation of rainfall etc., so that occurring the case where " wave crest-trough " is misfitted in similitude matching process.
Since there are series modalities caused by thermal extremes, extreme low temperature, rainfall mutation for agricultural weather data
The curve of cyclical fluctuations is abnormal, such as Wave crest and wave trough dislocation, thus common Euclidean distance not can solve the data with similar phenomenon away from
From measurement.The metric algorithm of dynamic time warping is introduced thus, and dynamic time warping algorithm is a kind of by unevenly distorting
Or bending solves similar sample sequence data originally and cannot do Similarity matching due to the linear drift that generates in time scale
The method of problem, but dynamic time warping is typically only capable to solve the measurement of single meteorological index, analyzes as discussed above, In
In real world, meteorological data belongs to multiple parameter data, there is the obvious mutation value in part in different variable dimensions.So,
The dynamic time warping method for measuring similarity for studying multidimensional, which becomes, to be solved in the research of agricultural weather integrated condition similarity analysis
One necessary work.The key of its similarity measurements flow function design is just to consider polynary meteorological temporal sequence simultaneously
It is listed in the correct matching for the form that the potential change information abundant of each variable dimension excavates and each dimension is fluctuated in the presence of complexity.
Based on meteorological data, there are the applications of different variation characteristics on each phenological period difference variable latitude of rice herein
Scene propose in conjunction with the rice phenological period segmentation with based on form and the multivariate time series for mixing gradient dynamic time warping
Method for measuring similarity.So that polynary meteorological data similarity analysis is closer to actual scene within the phenological period.
Three, summary of the invention
It is more the invention proposes a kind of rice based on form and the dynamic time warping for mixing gradient crucial phenological period
First meteorology time series data similarity measurements analysis method.Since there are bright for weather environment within the different phenological periods for rice
Aobvious variation and distributional difference, in order to guarantee the real world close to farming growth of following of similarity measurement, institute
It has carried out being divided according to the data of the whole story time in phenological period before carrying out meteorological data measurement in this way.And consider
To agricultural weather data in time scale, the sequence similarity of the accumulated value generation of the extreme value and rainfall of temperature
Distance inaccuracy caused by " wave crest-trough " abnormal matching problem of appearance is mixed, is proposed in dynamic time warping algorithm base
On plinth, in conjunction with Meteorological series form factor and be capable of depth excavate meteorological data internal information mixing gradient, establish
The method of new polynary agricultural weather data similarity measurement ensure that the meteorological data measurement in different phenological is more quasi-
Really.
Polynary gas of the present invention based on form with the rice for the dynamic time warping for mixing gradient crucial phenological period
As time series data similarity measurements analysis method, comprising the following steps:
Step 1: basic data prepares
The basic data that meteorological data similarity analysis need to prepare includes meteorological data (minimum temperature, maximum temperature, sunshine
When number or rainfall), different cultivars the phenologys period such as emergence, jointing, heading, maturation history corresponding date of seeding.
Day by day Temperature Series set in n growth period duration of rice:
T=< T1, T2…Tn> andWherein TiIndicate meteorological day by day in the growth period duration of rice in n time
Data,The high temperature of table, the lowest temperature, rainfall, sunshine time respectively.In order to emphasize agricultural weather data in time scale
On the front and back date between data interdependency, be expressed as follows using multivariate time series:
It is time span, 1≤i that wherein K, which is variables number, the N such as maximum temperature, minimum temperature, rainfall, sunshine time,
≤ K, 1≤j≤N, N indicate time span, the most variable numbers of the variable possessed that K is indicated, χijIndicate that i-th of variable exists
The actual observation value at jth time point.
The somewhere n kind growth period duration of rice date data:
Yn=< ySowing time, ySeeding stage, yTillering stage, yJointing stage, yBoot stage, yHeading stage, yFlorescence, yPustulation period, yMaturity period>, wherein y indicates entry into phenology
Phase start time, n indicate the time of institute's band, and gathering interior subscript indicates to define the phenological period of entrance.
Step 2: according to the division Meteorological series data in crop phenological period
Meteorological data feature itself and variation tendency feature within the different phenological periods is all different (such as Fig. 2), for
Entire breeding time carry out all meteorological datas carry out similarity measurements analysis can because this complex scene and there are biggish mistakes
Difference loses precision.So when for the meteorological data similarity analysis of specified kind crop, in order to enable measurement is more
Accurately, the different phenological whole story time progress data segmentation in real world according to the kind crop is needed.
The date data record of kind growth period duration of rice is specified in the somewhere n: into the time in certain class phenological period,
It is considered that when these two adjacent into the time in phenological period are that the kind entered needed for the latter phenological period for a period of time
Between, i.e. Yn=< ySowing time, ySeeding stage, yTillering stage, yJointing stage, yBoot stage, yHeading stage, yFlorescence, yPustulation period, yMaturity period> in, it is believed that the affiliated phenological period
It is as follows respectively: W<the seeding stage, n>=< ySowing time, ySeeding stage>, W<tillering stage, n>=< ySowing time, ySeeding stage>, WPhase<jointing stage, n>=< ySowing time, ySeeding stage>,
W<boot stage, n>=< ySowing time, ySeeding stage>, W<heading stage, n>=< ySowing time, ySeeding stage>, W<florescence, n>=< ySowing time, ySeeding stage>, W<the pustulation period, n>=< ySowing time,
ySeeding stage>, W<the maturity period, n>=< ySowing time, ySeeding stage>。
And according to the period in phenological period of above-mentioned expression, Temperature Series set day by day can be directed in n growth period duration of rice
Tday=< T1, T2...TnThe meteorological data segmentation of > progress different phenological, the different growing of available somewhere kind
Meteorological data set:
Sn=< SSowing time, SSeeding stage, STillering stage, SJointing stage, SBoot stage, SHeading stage, SFlorescence, SPustulation period, SMaturity period>, S indicates that the n-th time is corresponding
Meteorological data in the subscript affiliated phase in phenological period.
Step 3: the similarity measurement between target sample
The step 3 includes:
Agricultural weather data inherently have extremely strong complexity, are difficult inside statement from the data information on surface merely
Variation characteristic.In order to excavate the speed amplitude of variation tendency and variation of the agricultural weather data under different variables, introduce
The representation method of the data of Primordial Qi image data and its gradient.
The meteorological data of gradient more than step 3.1 describes
For agricultural weather data, carry out first as the raw value in step 1 multivariate time series indicate.For
Understanding of the enhancing to its variation tendency, introducing gradient in research indicates its variation tendency.First derivative is calculated first, it is public
Formula is as follows:
What i therein, j were respectively indicated be same sample same meteorological element variable under one group of time scale sequence q
Different corresponding points positions.
It repeats above-mentioned formula (1) and calculates the second dervative for obtaining the sequence, i.e., the speed and variation that it changes on sequence q
Amplitude expression, until all the points are calculated and terminate in the sequence.Thus constitute vectorTo indicate sequence
All information of q.
The design of step 3.2 form factor
Since the exceptional value in agrometeorological elements would generally bring very big influence to the growth of rice, so introducing shape
The basic goal of the state factor is to guarantee that sequence can kiss when measuring marking on the maximum similarity degree of waveform
It closes, highlights due to such as cumulant of rainfall and the high temperature low temperature exception bring sequence of temperature in agricultural weather variable
Fluctuation, influence brought by can be more when carrying out similitude weather data analysis the considerations of exceptional value.
Weight is assigned according to phase difference for the design of form factor, and mainly, is guaranteed when Wave crest and wave trough matches
Accurately, it and excludes because deliberately carrying out " wave crest-wave crest ", " trough-trough " is matched and had ignored between two correspondingly-shapeds
Existing phase difference, the Wave crest and wave trough matching that phase difference generates too far have exceeded a certain range and just lose meaning.
The step 3.2 is as follows:
The design of step 3.2.1 form factor constant
When the use of step very crucial in dynamic time warping algorithm being exactly the path matrix for creating m*n, two corresponding points
aiAnd bjDistance on matrix indicates.It is devised herein according to the phase difference between two o'clock and determines weighted value, in other words, when
aiAnd bjWhen point is close, weight assigns smaller value.In entire sequence distance calculating process, optimum distance is defined as 0.Since needing
Weight assignment is carried out according to the phase difference of two corresponding points, it is necessary to control coefrficient g be increased to the phase difference, also illustrated that
Control the empirical of the variation curvature of weighting function.
The design of step 3.2.2 dynamic form penalty
As described above, the phase difference utilized determines whether two o'clock approaches, it is expressed as | ai-bj|, in order to describe due to phase difference
Caused weight is in notable difference, introduces logic weighting function (MLWF), the wave crest according to caused by exceptional value in meteorological data
Trough irregular the case where occurring, it is as follows to construct penalty:
In order to further decrease the time complexity in calculating process, calculating is converted by the phase difference of corresponding points in research
Difference of the sequence of points of meteorological time series to be matched to known meteorological time series midpoint.Distance is indicated using c in text
The position of formation center [m/2].Comprehensively consider that form dynamic form penalty as follows:
The hybrid weight design of the meteorological data of gradient more than step 3.3 description
It is indicated by all information that previous step obtains q and then is calculated using formula (2) respectively in meteorological temporal sequence
Distance d under column original value, first derivative sequence and second dervative sequence0、d1And d2D is then carried out by weighting0、d1With
d2It is cumulative, obtain final the distance between two samples, specific formula is as follows:
D (i, j)=w0*d0(i, j)+w1*d1(i, j)+w2*d2(i, j) (4)
In summary all steps, it can be deduced that overall measure representation formula is as follows:
In formula, it is thus necessary to determine that parameter have g, w0, w1, w2.Pass through experiment (such as Fig. 3), it is determined that this four parameters are optimal
Combination: w0=1, w1=w2=2, g=0.2.
According to above-mentioned formula (5), calculating each meteorological variables of target year sequence are agreed to same in kind historical years with locality
In phenological period meteorological data data sample sequence of points to the distance between, export distance.
Step 3.4 constructs distance matrix, calculates shortest path strength using Dynamic Programming, is averaging distance.
It is cumulative using formula (5) calculated same time point distance based on weight 1, utilize formula (6) the filling distance square
Battle array is as follows:
It is as follows to obtain distance matrix:
Use dynamic programming algorithm, calculating matrix Dm*nMiddle shortest path, the method is as follows:
Step 3.4 successively calculates target time Meteorological series and all historical years Meteorological series within the same phenological period
Distance terminates until all historical years calculate, and exports all distance marking.
Step 4: being ranked up according to similarity measurement distance
Agricultural weather data under the specified crop varieties phenological period for calculating the target time according to the measure in step 3
The sample and other history years are obtained with the agricultural weather data under the same middle kind crop same phenological period of other historical years
The distance marking table of the corresponding sample of part.Then output is ranked up according to the score of marking table height from high to low, if distance is beaten
Split-phase is the same as work output side by side.First of output sequence indicate in specified sample with target time agricultural weather data
In the least similar time, the last one indicates the time most like with target time data agricultural weather data under particular cases.
It is as follows to improve major embodiment compared with existing agricultural weather data similarity measure by the present invention:
(1) redescribing for agricultural weather data is carried out using multivariate time series, ensure that in subsequent similitude point
During analysis, data are described closer to real physical world；
(2) the phenological period whole story according to specified kind has been carried out in the data handling procedure for carry out similarity analysis
Time segmentation, not only reduces the complexity of operation, at the same also contemplate crop under the different phenological to meteorological element according to
Rely degree；
(3) form factor (form constant factor and dynamic form punishment system are increased in carrying out similitude matching process
Number), it ensure that " wave crest-wave crest " of Meteorological series data, " trough-trough " are effectively matched；
(4) increase the mixing gradient that can be excavated in meteorological data in variation tendency information, it is abundant be utilized it is existing
Meteorological series data information so that similarity measurement is more accurate.
Four, Figure of description
Fig. 1 measures rice phenological period polynary meteorological data similarity method with the dynamic time warping for mixing gradient based on form
Flow chart
Fig. 2 2001 and Yixing force in 2005 educate round-grained rice rice anthesis Meteorological series variation tendency (minimum temperature (a), highest
Temperature (b), sunshine time (c) and rainfall (d))
The Yixing City Fig. 3 force educates round-grained rice rice classification accuracy result figure under the combination of each coefficient under different cluster numbers labels
Five, specific embodiment
The present invention is further described below by way of case study on implementation
Embodiment
By taking the force of Yixing City plantation educates round-grained rice rice varieties as an example, the local history meteorological data over the years in Yixing is educated for force
Round-grained rice rice varieties heading stage meteorological data carries out similar time analysis.In conjunction with attached drawing 1, present embodiment is illustrated:
Step 1: driving data prepares
The meteorological data historical data that selection is provided by national weather data sharing center, including max. daily temperature (DEG C),
Daily minimum temperature (DEG C), rainfall (mm), sunshine time (h) and corresponding area specify date in the phenological period over the years note of kind
Data are recorded, data description used is tested and is shown in Table 1, table 2.
Table 1 tests meteorological data and table is described in detail
Table 2 tests force and educates round-grained rice heading stage record date tables of data
Step 2: being divided according to the meteorological data in crop phenological period
Experiment divides the heading stage record data that the Yixing City force used educates the correspondence time of round-grained rice rice, as shown in table 2
Data corresponding meteorological data is split, obtain segmentation result, table 3 is the segmentation result example of heading stage in 1994.
Table 3 tests meteorological data and table is described in detail
Step 3: the similarity measurement between target sample
The step 3 includes:
Agricultural weather data inherently have extremely strong complexity, are difficult inside statement from the data information on surface merely
Variation characteristic.In order to excavate the speed amplitude of variation tendency and variation of the agricultural weather data under different variables, introduce
The representation method of the data of Primordial Qi image data and its gradient.
The meteorological data of gradient more than step 3.1 describes
For agricultural weather data, carry out first as the raw value in step 1 multivariate time series indicate.For
Understanding of the enhancing to its variation tendency, introducing gradient in research indicates its variation tendency.First derivative is calculated first, it is public
Formula is as follows:
What i therein, j were respectively indicated be same sample same meteorological element variable under one group of time scale sequence q
Different corresponding points positions.
It repeats above-mentioned formula (1) and calculates the second dervative for obtaining the sequence, i.e., the speed and variation that it changes on sequence q
Amplitude expression, until all the points are calculated and terminate in the sequence.Thus constitute vectorTo indicate sequence
All information of q.
The design of step 3.2 form factor
Since the exceptional value in agrometeorological elements would generally bring very big influence to the growth of rice, so introducing shape
The basic goal of the state factor is to guarantee that sequence can kiss when measuring marking on the maximum similarity degree of waveform
It closes, highlights due to such as cumulant of rainfall and the high temperature low temperature exception bring sequence of temperature in agricultural weather variable
Fluctuation, influence brought by can be more when carrying out similitude weather data analysis the considerations of exceptional value.
Weight is assigned according to phase difference for the design of form factor, and mainly, is guaranteed when Wave crest and wave trough matches
Accurately, it and excludes because deliberately carrying out " wave crest-wave crest ", " trough-trough " is matched and had ignored between two correspondingly-shapeds
Existing phase difference, the Wave crest and wave trough matching that phase difference generates too far have exceeded a certain range and just lose meaning.
The step 3.2 is as follows:
The design of step 3.2.1 form factor constant
When the use of step very crucial in dynamic time warping algorithm being exactly the path matrix for creating m*n, two corresponding points
aiAnd bjDistance on matrix indicates.It is devised herein according to the phase difference between two o'clock and determines weighted value, in other words, when
aiAnd bjWhen point is close, weight assigns smaller value.In entire sequence distance calculating process, optimum distance is defined as 0.Since needing
Weight assignment is carried out according to the phase difference of two corresponding points, it is necessary to control coefrficient g be increased to the phase difference, also illustrated that
Control the empirical of the variation curvature of weighting function.
The design of step 3.2.2 dynamic form penalty
As described above, the phase difference utilized determines whether two o'clock approaches, it is expressed as | ai-bj|, in order to describe due to phase difference
Caused weight is in notable difference, introduces logic weighting function (MLWF), the wave crest according to caused by exceptional value in meteorological data
Trough irregular the case where occurring, it is as follows to construct penalty:
In order to further decrease the time complexity in calculating process, calculating is converted by the phase difference of corresponding points in research
Difference of the sequence of points of meteorological time series to be matched to known meteorological time series midpoint.Distance is indicated using c in text
The position of formation center [m/2].Comprehensively consider that form dynamic form penalty as follows:
The hybrid weight design of the meteorological data of gradient more than step 3.3 description
It is indicated by all information that previous step obtains q and then is calculated using formula (2) respectively in meteorological temporal sequence
Distance d under column original value, first derivative sequence and second dervative sequence0、d1And d2D is then carried out by weighting0、d1With
d2It is cumulative, obtain final the distance between two samples, specific formula is as follows:
D (i, j)=w0*d0(i, j)+w1*d1(i, j)+w2*d2(i, j) (4)
In summary all steps, it can be deduced that overall measure representation formula is as follows:
In formula, it is thus necessary to determine that parameter have g, w0, w1, w2.Pass through experiment (such as Fig. 3), it is determined that this four parameters are optimal
Combination: w0=1, w1=w2=2, g=0.2.According to above-mentioned formula (5), each meteorological variables of target year sequence and locality are calculated
Agree in kind historical years in the same phenological period meteorological data data sample sequence of points to the distance between, export distance.
Step 3.4 constructs distance matrix, calculates shortest path strength using Dynamic Programming, is averaging distance.
The calculated same time point distance of formula 5 is utilized based on weight 1 is cumulative, using formula (6) the filling distance matrix,
It is as follows:
It is as follows to obtain distance matrix:
Use dynamic programming algorithm, calculating matrix Dm*nMiddle shortest path, the method is as follows:
Step 3.4 successively calculates target time Meteorological series and all historical years Meteorological series within the same phenological period
Distance terminates until all historical years calculate, and exports all distance marking.
According to above-mentioned formula (7), calculates in target year and local agreement kind historical years and agree to meteorological number in the phenological period
According to the distance between data sample, distance sequence is exported, what is calculated herein is that Yixing force educates round-grained rice rice 2011 and historical years
The distance of (1994-2011 is removed 2006,2007,2008) between the meteorological data in this period at heading stage is beaten
Point, export result such as table 4:
Table 4 2011 years and historical years the meteorological data similarity distance within heading stage are given a mark table
Step 4: being ranked up according to similarity measurement distance
Agricultural weather data under the specified crop varieties phenological period for calculating the target time according to the measure in step 3
The sample and other history years are obtained with the agricultural weather data under the same middle kind crop same phenological period of other historical years
The distance marking table of the corresponding sample of part.Then output is ranked up according to the score of marking table height from high to low.Output sequence
First indicate time least similar with target time agricultural weather data, the last one table in specified sample
Show the time most like with target time data agricultural weather data under particular cases.By time in table 4 according to marking from
It is high to Low to be ranked up, it obtains educating round-grained rice with Yixing force and gives a mark sequencing table in the similarity distance of 2011 heading stages and historical years,
Such as table 5:
Table 5 2011 years and historical years the meteorological data similarity distance within heading stage are given a mark sequencing table
| Time |
| 2005 |
| 2004 |
| 2002 |
| 2010 |
| 1994 |
| 2003 |
| 2000 |
| 1996 |
| 1997 |
| 1999 |
| 2001 |
| 1998 |
| 2009 |
| 1995 |
In table 5, it can be seen that it is least similar at 2011 heading stages and 2005 in historical years that Yixing force educates round-grained rice,
It is most like with nineteen ninety-five.
Claims (5)
1. measuring rice phenological period polynary meteorological data similarity method with the dynamic time warping for mixing gradient based on form
It mainly comprises the steps that
1) designated area meteorological data and this area specify the phenological period date data of kind to prepare
2) according to the division Meteorological series data in crop phenological period
3) based on form with mix the similitude between gradient DTW metric objective sample
4) it is ranked up according to similarity measurement distance marking.
2. according to claim 1 based on form and the crucial phenological period multivariable agricultural weather number of the crops for mixing gradient
According to similarity measurements analysis method, it is characterised in that carry out agricultural weather data according to according to multivariate time series in step 1)
It redescribes.In order to increase the interdependency in time scale, it is expressed as follows using multivariate time series:
Wherein K is the variables number such as maximum temperature, the lowest temperature base of a fruit, rainfall, sunshine time, and N is time span, 1≤i≤
K, 1≤j≤N, N indicate time span, the most variable numbers of the variable possessed that K is indicated, χijIndicate i-th of variable in jth
The actual observation value at time point.
3. according to claim 1 based on form and the crucial phenological period multivariable agricultural weather number of the crops for mixing gradient
According to similarity measurements analysis method, it is characterised in that divided in step 2) according to the meteorological data in crop phenological period.Crop is not
It is all different to the degree of dependence of meteorological element in the same phenological period, all meteorological datas, which are carried out, for entire breeding time carries out phases
Like property metric analysis, there is no actual meanings, and increase the complexity of calculating.So for specified kind crop
When meteorological data similarity analysis, in order to enable measurement is more accurate, need according to the kind crop in real world not
Data segmentation is carried out with the whole story time in phenological period.
What is recorded in the date data that kind growth period duration of rice is specified in the somewhere n is the time thought into certain phenological period,
It is considered that these two adjacent into the time in phenological period are the time needed for the kind enters the latter phenological period for a period of time,
That is Yn=< ySowing time, ySeeding stage, yTillering stage, yJointing stage, yBoot stage, yHeading stage, yFlorescence, yPustulation period, yMaturity period> in, it is believed that affiliated phenological period difference
It is as follows: W<the seeding stage, n>=< ySowing time, ySeeding stage>, W<tillering stage, n>=< ySowing time, ySeeding stage>, WPhase<jointing stage, n>=< ySowing time, ySeeding stage>, W<boot stage, n>=
<ySowing time, ySeeding stage>, W<heading stage, n>=< ySowing time, ySeeding stage>, W<florescence, n>=< ySowing time, ySeeding stage>, W<the pustulation period, n>=< ySowing time, ySeeding stage>,
W<the maturity period, n>=< ySowing time, ySeeding stage>。
And according to the period in phenological period of above-mentioned expression, Temperature Series set T day by day can be directed in n growth period duration of riceday=
<T1, T2...TnThe meteorological data segmentation of > progress different phenological, the gas of the different growing of available somewhere kind
Image data set:
Sn=< SSowing time, SSeeding stage, STillering stage, SJointing stage, SBoot stage, SHeading stage, SFlorescence, SPustulation period, SMaturity period>, S indicates the n-th time corresponding subscript
Meteorological data in the affiliated phase in phenological period.
4. according to claim 1 based on form and the crucial phenological period multivariable agricultural weather number of the crops for mixing gradient
According to similarity measurements analysis method, it is characterised in that the similarity measurement in step 3) between target sample.It calculates under different year
The key of similarity degree under the time serieses of meteorological datas such as the highest temperature is low, minimum temperature, rainfall, sunshine time is energy
It is enough to carry out sequences match in the sample meteorological data of different year, it is only guaranteed to have arrived " wave crest-wave crest ", " trough-trough "
Correct matching, the distance between obtained sample is just accurate.So needing to increase the shape to data sequence in calculating process
The matched control of state and hope obtain in more Meteorological series data variation tendency information.
The meteorological data of gradient more than step 3.1 describes
For agricultural weather data, carry out first as the raw value in step 1 multivariate time series indicate.In order to increase
By force to the understanding of its variation tendency, introducing gradient in research indicates its variation tendency.First derivative is calculated first, and formula is such as
Under:
What i therein, j were respectively indicated is that sequence q is not under one group of time scale for the same meteorological element variable of same sample
With corresponding points position.
It repeats above-mentioned formula (1) and calculates the second dervative for obtaining the sequence, i.e., the width of it changes on sequence q speed and variation
The expression of degree, until all the points are calculated and terminate in the sequence.Thus constitute vectorTo indicate the institute of sequence q
There is information.
The design of step 3.2 form factor
Since the exceptional value in agrometeorological elements would generally bring very big influence to the growth of crops, so introducing form
The basic goal of the factor is in order to guarantee that sequence can coincide when measuring marking on the maximum similarity degree of waveform, by force
The high temperature low temperature exception bring sequence fluctuation due to the cumulant and temperature of such as rainfall in agricultural weather variable, In are adjusted
The brought influence of exceptional value of can be more when progress similitude weather data analysis the considerations of.
Weight is assigned according to phase difference for the design of form factor, and mainly, it is accurate when Wave crest and wave trough matches to guarantee,
And it excludes because deliberately carrying out " wave crest-wave crest ", " trough-trough " is matched and had ignored existing between two correspondingly-shapeds
Phase difference, the Wave crest and wave trough matching that phase difference generates too far have exceeded a certain range and just lose meaning.
The step 3.2 is as follows:
The design of step 3.2.1 form factor constant
When the use of step very crucial in dynamic time warping algorithm being exactly the path matrix for creating m*n, two corresponding points aiAnd bj
Distance on matrix indicates.It is devised herein according to the phase difference between two o'clock and determines weighted value, in other words, work as aiAnd bj
When point is close, weight assigns smaller value.In entire sequence distance calculating process, optimum distance is defined as 0.Since needing basis
The phase difference of two corresponding points carries out weight assignment, it is necessary to increase control coefrficient g to the phase difference, also illustrate that control weight
The empirical of the variation curvature of function.
The design of step 3.2.2 dynamic form penalty
As described above, the phase difference utilized determines whether two o'clock approaches, it is expressed as | ai-bj|, in order to describe to cause due to phase difference
Weight be in notable difference, introduce logic weighting function (MLWF), the Wave crest and wave trough according to caused by exceptional value in meteorological data
Irregular the case where occurring, it is as follows to construct penalty:
In order to further decrease the time complexity in calculating process, converts the phase difference of corresponding points in research and calculate to quilt
Difference of the sequence of points of matched meteorological time series to known meteorological time series midpoint.Distance sequence is indicated using c in text
The position at center [m/2].Comprehensively consider that form dynamic form penalty as follows:
The hybrid weight design of the meteorological data of gradient more than step 3.3 description
It is indicated by all information that previous step obtains q and then is calculated respectively using formula (2) in meteorological time series original
Distance d under initial value, first derivative sequence and second dervative sequence0、d1And d2D is then carried out by weighting0、d1And d2's
It is cumulative, final the distance between two samples are obtained, specific formula is as follows:
D (i, j)=w0*d0(i, j)+w1*d1(i, j)+w2*d2(i, j) (4)
In summary all steps, it can be deduced that overall measure representation formula is as follows:
In formula, it is thus necessary to determine that parameter have g, w0, w1, w2.Pass through experiment, it is determined that the optimal combination of this four parameters: w0=
1, w1=w2=2, g=0.2.
According to above-mentioned formula (5), same phenology in each meteorological variables of target year sequence and local agreement kind historical years is calculated
In phase meteorological data data sample sequence of points to the distance between, export distance.
Step 3.4 constructs distance matrix, calculates shortest path strength using Dynamic Programming, is averaging distance.
It is cumulative using formula (5) calculated same time point distance based on weight 1, using formula (6) the filling distance matrix, such as
Under:
It is as follows to obtain distance matrix:
Use dynamic programming algorithm, calculating matrix Dm*nMiddle shortest path, the method is as follows:
Step 3.4 successively calculate target time Meteorological series and all historical years Meteorological series within the same phenological period away from
From until the calculating of all historical years terminates, and exporting all distance marking.
5. according to claim 1 based on form and the crucial phenological period multivariable agricultural weather number of the crops for mixing gradient
According to similarity measurements analysis method, it is characterised in that step 4) is ranked up according to similarity measurement distance.According in step 3
Under the specified crop varieties phenological period that measure calculates the target time agricultural weather data and other historical years it is same in
Agricultural weather data under the kind crop same phenological period obtain the sample and give a mark at a distance from the corresponding sample of other historical years
Table.Then output is ranked up according to the score of marking table height from high to low, if apart from identical work output side by side of giving a mark.It is defeated
First of sequence indicates the time least similar with target time agricultural weather data in specified sample out, last
The time most like with target time data agricultural weather data under a expression particular cases.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810498045.4A CN110503287A (en) | 2018-05-18 | 2018-05-18 | Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810498045.4A CN110503287A (en) | 2018-05-18 | 2018-05-18 | Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110503287A true CN110503287A (en) | 2019-11-26 |
Family
ID=68584193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810498045.4A Pending CN110503287A (en) | 2018-05-18 | 2018-05-18 | Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110503287A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461247A (en) * | 2020-04-09 | 2020-07-28 | 浙江国贸云商控股有限公司 | Feature data processing method and related device |
| CN113761023A (en) * | 2021-08-24 | 2021-12-07 | 国网甘肃省电力公司 | Photovoltaic power generation short-term power prediction method based on improved generalized neural network |
| CN117041303A (en) * | 2023-10-09 | 2023-11-10 | 睿兴科技(天津)有限公司 | Data transmission method and system for intelligent diagnosis of water pump |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003006274A (en) * | 2001-06-25 | 2003-01-10 | National Agricultural Research Organization | Judgement system for suitable crop and its kind to be introduced, and cultivation ground and time based on climate similarity at each growth stage |
| CN103942444A (en) * | 2014-04-29 | 2014-07-23 | 衣晓 | Parity 1-norm unequal length sequence similarity metric algorithm based on DTW |
| CN105913062A (en) * | 2016-04-26 | 2016-08-31 | 深圳前海云汉金融科技有限公司 | Big data trend curve local characteristics based matching method and terminal |
| CN105913326A (en) * | 2016-04-06 | 2016-08-31 | 南京农业大学 | Constraining knowledge and elite individual strategy genetic algorithm fusion-based crop growth period model variety parameter optimization method |
-
2018
- 2018-05-18 CN CN201810498045.4A patent/CN110503287A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003006274A (en) * | 2001-06-25 | 2003-01-10 | National Agricultural Research Organization | Judgement system for suitable crop and its kind to be introduced, and cultivation ground and time based on climate similarity at each growth stage |
| CN103942444A (en) * | 2014-04-29 | 2014-07-23 | 衣晓 | Parity 1-norm unequal length sequence similarity metric algorithm based on DTW |
| CN105913326A (en) * | 2016-04-06 | 2016-08-31 | 南京农业大学 | Constraining knowledge and elite individual strategy genetic algorithm fusion-based crop growth period model variety parameter optimization method |
| CN105913062A (en) * | 2016-04-26 | 2016-08-31 | 深圳前海云汉金融科技有限公司 | Big data trend curve local characteristics based matching method and terminal |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111461247A (en) * | 2020-04-09 | 2020-07-28 | 浙江国贸云商控股有限公司 | Feature data processing method and related device |
| CN113761023A (en) * | 2021-08-24 | 2021-12-07 | 国网甘肃省电力公司 | Photovoltaic power generation short-term power prediction method based on improved generalized neural network |
| CN117041303A (en) * | 2023-10-09 | 2023-11-10 | 睿兴科技(天津)有限公司 | Data transmission method and system for intelligent diagnosis of water pump |
| CN117041303B (en) * | 2023-10-09 | 2023-12-15 | 睿兴科技(天津)有限公司 | Data transmission method and system for intelligent diagnosis of water pump |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110503287A (en) | Rice phenological period polynary meteorological data similarity method is measured with the dynamic time warping for mixing gradient based on form | |
| CN105445229A (en) | Nitrogen balance spectroscopy-based wheat spring nitrogenous fertilizer application method, and construction method of nitrogen topdressing amount model thereof | |
| CN107329511B (en) | Based on the vegetable aquaculture luminous environment high efficiency regulatory method and system between suitable root warm area | |
| CN104199027B (en) | Method based on the polarimetric radar large area near real-time Monitoring Rice phenological period of compacting | |
| CN107356569A (en) | The construction method of method and its model based on chlorophyll fluorescence prediction wheat grain yield | |
| CN108169138A (en) | A kind of rice lodging monitoring method using thermal infrared images | |
| CN109615148B (en) | Method and system for determining meteorological yield of corn | |
| CN112598277B (en) | Method for evaluating yield difference of trans-regional reduced winter wheat and improving nitrogen fertilizer efficiency | |
| CN103971199B (en) | A kind of remote sensing ranking method of a wide range of Grain Growth Situation | |
| CN110874617A (en) | Method for establishing winter wheat leaf nitrogen content estimation model | |
| CN107941713A (en) | A kind of rice yield estimation method based on coupling crop modeling assimilation spectral reflectivity | |
| CN112735511A (en) | Cold damage cucumber PSII potential activity prediction method based on QGA-SVR | |
| CN108770614A (en) | A kind of Heat Tolerance of Wheat Cultivars evaluation method | |
| CN114118679B (en) | Crop yield per unit and growth evaluation method based on time sequence remote sensing data | |
| CN105654242A (en) | Fish swarm algorithm-based cucumber seedling stage carbon dioxide optimization regulation and control model, establishment method and application thereof | |
| CN113553767A (en) | Method and system for building greenhouse crop photosynthetic rate prediction model | |
| CN112992271A (en) | Method for rapidly predicting crop latitude adaptability indoors | |
| CN101464979A (en) | System and method for evaluating crop yield representation | |
| CN109472320B (en) | Method for automatically correcting variety parameters of crop growth period model under uncertain conditions | |
| JP7208503B2 (en) | Machine learning program, machine learning method and machine learning apparatus | |
| CN109615150B (en) | Method and system for determining rice meteorological output | |
| CN110853711A (en) | Whole genome selection model for predicting fructose content of tobacco and application thereof | |
| CN107609695B (en) | Crop yield remote sensing estimation method based on adjustable vegetation index | |
| Wu et al. | Crop yield estimation and irrigation scheduling optimization using a root-weighted soil water availability based water production function | |
| Khotimah | Performance of the K-nearest neighbors method on identification of maize plant nutrients |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 210095 803, building 7, Shanshui house, Zhongshan Garden City, Xuanwu District, Nanjing City, Jiangsu Province Applicant after: NANJING AGRICULTURAL University Address before: Weigang Xuanwu District of Nanjing Jiangsu province 210095 No. 1 Applicant before: NANJING AGRICULTURAL University |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |