CN108229065A - A kind of construction method and Forecasting Methodology of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature - Google Patents
A kind of construction method and Forecasting Methodology of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature Download PDFInfo
- Publication number
- CN108229065A CN108229065A CN201810119071.1A CN201810119071A CN108229065A CN 108229065 A CN108229065 A CN 108229065A CN 201810119071 A CN201810119071 A CN 201810119071A CN 108229065 A CN108229065 A CN 108229065A
- Authority
- CN
- China
- Prior art keywords
- leaf
- sample
- xylophyta
- photosynthetic rate
- net photosynthetic
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/231—Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
Abstract
The present invention provides the construction methods and Forecasting Methodology of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature, belong to field of bioinformatics.The present invention is clustered using the linkage that will be averaged and PAM clustering procedures combine, different sub- kinds is divided into the xylophyta of diverse geographic location, the difference of different sub- kinds is eliminated more reasonably to be predicted Net Photosynthetic Rate, the present invention is utilized respectively phenotypic data of the gradient boosting algorithm based on blade sample in every sub- kind and builds Net Photosynthetic Rate prediction model simultaneously, during model is established, present invention firstly provides gradient boosting algorithm iteration stopping criterion, xylophyta leaf morphology data for different sub- kinds provide the residual error critical value of algorithm iteration, establish the populus simonii Net Photosynthetic Rate prediction model of the improvement gradient boosting algorithm based on iteration stopping criterion, in the case of known leaf morphology data, there is ideal prediction result to the Net Photosynthetic Rate of the sub- kind of different populus simoniis.
Description
Technical field
The invention belongs to biological information fields, and in particular to a kind of net photosynthesis speed based on xylophyta leaf morphology feature
The construction method and Forecasting Methodology of rate prediction model.
Background technology
Net Photosynthetic Rate is to weigh the important indicator of photosynthesis of plant intensity, and weight is played the part of in plant growth and development process
It acts on (ShipleyB et al.2005).19th century, Sachs propositions first measure photosynthetic rate with half leaf method, this is earliest
Applied to the method for value in measuring photosynthesis, it does not need to complicated instrument and equipment, simple and easy to do.But it can be broken using half leaf method
Bad measured material, it is impossible to which METHOD FOR CONTINUOUS DETERMINATION, and minute is long, environmental condition is difficult to control, and leads to that determination efficiency is low, error is big
(Wei Yongsheng .2008).Sachs in 1883 initiates the photosynthetic rate that plant is measured with dry weight method, the initial stage seventies, the Chinese Academy of Sciences
Shanghai plant physiology research institute prevents photosynthate using the method (being known as improved dry weight method) of scald petiole base bast
It transports, makes measurement result closer to reality, so as to promote the application of dry weight method and research.But due to the difference of environmental condition,
Measured result is difficult to be compared to each other (RoderiekH.1978 under the conditions of different times；Wang Tian is into .1988；Pei Baohua et
al.1990)。
At present, CO2Infrared gas analysis method in absorption process is the widely used method for measuring plant Net Photosynthetic Rate
(yellow refined perfume .2003), development have passed through three phases.First stage is laboratory stage, starts the beginning of the fifties to apply, match
Relatively easy, only leaf chamber (or assimilation box) and the corresponding configurations such as gas circuit and air pump are put, takes artificial reading and calculating, such as U.S.
The infrared CO of BACKMAN companies2Analyzer, the FQ series CO of China Foshan analytical instrument factory production2Analyzer etc., later
Small-sized DC dual-purpose CO is developed again2Analyzer, such as the GXH 1 of Beijing Analytical Instrument Factory's production.Second stage is quotient
The product stage, in order to increase finding speed, energy multi-site determination, CO2Multipath conversion is configured on analyzer and corresponding record fills
It puts, to reach the 10 combination of channels formula of 6 channels and mono- 865 types of U.S. BAcKMAN of machine multiple spot channel measure, such as Japan's production
CO2Analyzer.Later in order to improve measurement accuracy and determination condition can be controlled, it is configured with the environmental control system of leaf chamber, such as moral
The MINICUVITTECO of state's production2-H2O measurement systems.Phase III is the portable stage.The generation in this stage is due to list
Piece machine, integrated circuit and Sensor Technology Development as a result, being launched in the mid-80.Microcontroller is by photosynthetic continuous mode
Involved in the physical quantitys such as CO2, temperature, humidity, PAR (photosynthetically active radiation) and flow carry out various operations, substantially increase
Determination efficiency.The data storage of configuration can store big amount measurement data, pass through additional leaf chamber, temperature, wet, light, CO2Concentration control
System processed carrys out Regulate Environment factor.The LCA-III types, the mono- IV types of LCA that have Adc Inc. more commonly used at present, U.S. LICOR
The LICOR-6200 and LICOR 1 (Jiang Gaoming, 1996) of company, mono- 301PS, CI-510 and CI 1 of CI of CID companies
And the BAU photosynthesis measurement systems of China Agricultural University's production.When LI-6400XT is used to measure Net Photosynthetic Rate, need upper
Photosynthetic photon flux density (PPFD) is set in 1600 μm of olm by 9 points to 11 points of the period of noon-2·s-1, concentration is set as
400μmol-1, when Net Photosynthetic Rate becomes constant (± 0.1), record the value of Net Photosynthetic Rate.As it can be seen that current existing net light
It is generally higher to environmental requirement to close rate determination method.
Invention content
In view of this, the purpose of the present invention is to provide a kind of sides for rapidly and accurately predicting xylophyta Net Photosynthetic Rate
Method, the method are not limited by instrument and environment.
In order to achieve the above-mentioned object of the invention, the present invention provides following technical scheme：
The present invention provides the construction method of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature, packets
Include following steps：
1) the xylophyta blade of the same species of China's covering entire geographical distribution of xylophyta is chosen, obtains woody plant
Object blade sample；
2) the leaf morphology feature of the xylophyta blade sample is measured, obtains the blade table of each xylophyta individual
Type data；
The leaf morphology feature includes leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf perimeter and the leaf shape factor；
3) based on the leaf morphology data, using average linkage clustering method and PAM clustering methods to xylophyta
Body carries out Clustering, obtains xylophyta individual phylogenetic group；
4) Net Photosynthetic Rate of xylophyta blade sample in the step 1) is measured, obtains the Net Photosynthetic Rate number of sample
According to；
5) using gradient boosting algorithm, according to iteration stopping criterion to the blade of each xylophyta individual of sample phylogenetic group
The Net Photosynthetic Rate data of phenotypic data and the sample build model, obtain Net Photosynthetic Rate prediction model；
The Net Photosynthetic Rate prediction model is shown in formula (1)；
Wherein, M represents iterations；xiRepresent six-vector (x1i, x2i, x3i, x4i, x5i, x6i), it corresponds to i-th respectively
The area of populus simonii sample, length, width, perimeter, length-width ratio and the leaf factor；RmjWhen representing the m times iteration, j-th of leaflet
Poplar sample set；cmjWhen representing the m times iteration, in j-th of populus simonii sample set, pass through minimumObtained constant；
There is no time restriction between the step 2) and step 4).
Preferably, construction method in the step 5), includes the following steps：
A. initialization constants c so thatMinimum,To lose letter
Number, N are sample number, yiNet Photosynthetic Rate value for i-th of populus simonii sample；I represents vector, is a six-vector (x1i,
x2i, x3i, x4i, x5i, x6i), x1i, x2i, x3i, x4i, x5i, x6iThe leaf area of i-th of xylophyta sample is corresponded to respectively, and leaf is long
Degree, leaf width degree, leaf perimeter, ratio of length to breadth and the leaf shape factor；
B. it to iterations m=1,2...M, if i=1,2...N, calculatesIt obtains
The negative gradient of loss function, using obtained negative gradient as the estimation of residual error, fm(x) for the m times iteration function when, according to rmi
Learn a regression tree；M is the m times iteration, and i is i-th of sample；Function when f (x) is the m-1 times iteration, first time iteration
When be f0(x), i.e.,And f (xi) it is by the leaf area of i-th of xylophyta sample, leaf length, leaf width degree, leaf
Perimeter, ratio of length to breadth and the leaf shape factor substitute into function f (x)；
C. so thatMinimum acquires cmj, update regression tree
Wherein, λ ∈ (0.001,0.01)；M be the m time iteration, j be j-th of region, the leaf of substantially several xylophytas
The set of piece sample；RmjDuring for the m times iteration, all samples have been divided into J set, and j represents j-th of set；
D. the residual error critical value of iteration stopping is calculatedWherein n be sample number, s2For sample side
Difference, α are significance；When residual error is less than σIt is criticalWhen, otherwise iteration stopping, carries out next iteration until residual error is less than σIt is critical,
Obtaining regression tree model is
Preferably, the side of Clustering is carried out in the step 3) to xylophyta individual using averagely linkage clustering method
Method includes the following steps：
1. the phenotypic characteristic data of xylophyta sample are chosen as the variable needed；
2. the variable data is normalized, normalization variable data is obtained；
3. calculate Distance matrix D using the normalization variable dataij；
4. utilize hclust () function pair Distance matrix D of R languageijImplementation level clusters, and two closest classes are drawn
Into one kind, wherein between class distance is defined as：Point and the average distance at another class midpoint in one class.
Preferably, the PAM clustering methods include the following steps：
A, sample is set as X { x (1), x (2) ... .. }；
B. k cluster centre is randomly selected in the sample first；
C. the distance to each cluster centre and then to the sample point in addition to cluster centre is calculated, sample is referred to distance
The nearest sample point of center of a sample, obtains initial cluster；
D. other sample points in each initial cluster in addition to the point at class center are calculated with the distance to other all the points
The minimum value is just realized primary cluster optimization by the minimum value of sum；
E. step d is repeated, until the position of cluster centre twice no longer changes, obtains cluster result.
Preferably, the quantity of xylophyta sample is 200~500 in the step 1).
Preferably, shown in step 2) the middle period form factor calculation formula such as formula (2)；
Preferably, when measure, will for 9 points to 11 points of the morning time of the measure of Net Photosynthetic Rate in the step 4)
Photosynthetic photon flux density is set in 1600 μm of olm-2·s-1, concentration is set as 400 μm of ol-1。
The present invention provides it is a kind of based on the prediction model that the construction method obtains to xylophyta Net Photosynthetic Rate
Forecasting Methodology includes the following steps：
A. the leaf morphology feature of xylophyta to be predicted is measured, obtains leaf morphology characteristic, the leaf morphology
Feature includes leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf perimeter and the leaf shape factor；
B. the leaf morphology data are inputted in the prediction model obtained in said program construction method, obtains prediction wood
The predicted value of this plant Net Photosynthetic Rate.
The present invention provides the construction method of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature, by
Different in the geographical location that the xylophyta of same species is distributed, the growing environment in different geographical locations makes the xylophyta
With different leaves phenotypic characteristic and Net Photosynthetic Rate, the present invention will be using will averagely linkage cluster and PAM clustering procedures combine, to not
Xylophyta with geographical location is divided into different sub- kinds, therefore eliminates the difference of different sub- kinds so as to net photosynthesis speed
Rate more reasonably predicted, at the same the present invention using the method for average linkage cluster and PAM clusters combine to blade sample into
Row cluster is so that more accurate to the division of sub- kind；The present invention is utilized respectively gradient boosting algorithm in every sub- kind simultaneously
Phenotypic data structure Net Photosynthetic Rate prediction model based on blade sample, during model is established, the present invention carries for the first time
Gradient boosting algorithm iteration stopping criterion is gone out, the xylophyta leaf morphology data for different sub- kinds provide algorithm iteration
Residual error critical value, establish the improvement gradient boosting algorithm based on iteration stopping criterion populus simonii Net Photosynthetic Rate prediction mould
Type in the case of known leaf morphology data, has the Net Photosynthetic Rate of the sub- kind of different populus simoniis ideal pre-
Survey result.
Description of the drawings
Fig. 1 is the cluster result figure that averagely links in embodiment 1；
Fig. 2 is PAM cluster result figures in embodiment 1.
Specific embodiment
The present invention provides the construction method of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature, packets
Include following steps：
1) the xylophyta blade of the same species of China's covering entire geographical distribution of xylophyta is chosen, obtains woody plant
Object blade sample；
2) the leaf morphology feature of the xylophyta blade sample is measured, obtains the blade table of each xylophyta individual
Type data；
The leaf morphology feature includes leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf perimeter and the leaf shape factor；
3) based on the leaf morphology data, using average linkage clustering method and PAM clustering methods to xylophyta
Body carries out Clustering, obtains xylophyta individual phylogenetic group；
4) Net Photosynthetic Rate of xylophyta blade sample in the step 1) is measured, obtains the Net Photosynthetic Rate number of sample
According to；
5) using gradient boosting algorithm, according to iteration stopping criterion to the blade of each xylophyta individual of sample phylogenetic group
The Net Photosynthetic Rate data of phenotypic data and the sample build model, obtain Net Photosynthetic Rate prediction model；
The Net Photosynthetic Rate prediction model is shown in formula (1)；
Wherein, M represents iterations；xiRepresent six-vector (x1i, x2i, x3i, x4i, x5i, x6i), it corresponds to i-th respectively
The area of populus simonii sample, length, width, perimeter, length-width ratio and the leaf factor；RmjWhen representing the m times iteration, j-th of leaflet
Poplar sample set；cmjWhen representing the m times iteration, in j-th of populus simonii sample set, pass through minimumObtained constant.
There is no time restriction between the step 2) and step 4).
The present invention chooses the xylophyta blade of the same species of China's covering entire geographical distribution of xylophyta, obtains wood
This plant leaf blade sample.
The present invention does not have the type of the xylophyta particular/special requirement, and method provided by the invention is applicable to this field
Any xylophyta kind known to technical staff.The xylophyta kind is preferably willow platymiscium, more preferably small
Ye Yang.In embodiments of the present invention, it is tested for the xylophyta selection populus simonii.
In the present invention, when the xylophyta be populus simonii when, the geographical distribution be preferably Shaanxi, Qinghai, Hebei,
Henan, Ningxia, Beijing and Inner Mongol.The quantity of the xylophyta blade is preferably 1000~1500.From the xylophyta
Individual representative is preferably chosen in blade and obtains xylophyta blade sample.The quantity of the xylophyta blade sample is preferably
200~500.The standard that the selection individual represents is healthy and complete as blade.
After obtaining xylophyta blade sample, the present invention measures the leaf morphology feature of the xylophyta blade sample,
Obtain the leaf morphology data of each xylophyta individual；The leaf morphology feature include leaf area, leaf length, leaf width degree,
Ratio of length to breadth, leaf perimeter and the leaf shape factor.
In the present invention, the measurement is preferably used with instrument and utilizes portable laser leaf area device (CI-202).
In the present invention, shown in the leaf shape factor calculation formula such as formula (2)；
The present invention is based on the leaf morphology data, using average linkage clustering method and PAM clustering methods to woody plant
Object individual carries out Clustering, obtains xylophyta individual phylogenetic group.
In the present invention, the method for carrying out Clustering to xylophyta individual using averagely linkage clustering method,
Include the following steps：
1. the phenotypic characteristic data of xylophyta sample are chosen as the variable needed；
2. the variable data is normalized, the variable data that obtains that treated；
3. treated described in, variable data calculates Distance matrix Dij；
4. the hclust () using R language adjusts the distance matrix DijImplementation level clusters, and two closest classes are divided into one
Class, wherein between class distance are defined as：Point and the average distance at another class midpoint in one class.
In the present invention, the PAM clustering methods include the following steps：
A, sample is set as X { x (1), x (2) ... .. }；
B. k cluster centre is randomly selected in the sample first；
C. the distance to each cluster centre and then to the sample point in addition to cluster centre is calculated, sample is referred to distance
The nearest sample point of center of a sample, obtains initial cluster；
D. to the distance sum of other sample points calculating in each cluster in addition to the point at class center to other all the points most
Small value, using the minimum value as new cluster, center just realizes primary cluster optimization；
E. step d is repeated, until the position of cluster centre twice no longer changes, obtains cluster result.
In the present invention, the cluster result result that two kinds of clustering methods obtain unanimously is divided according to result, still
The third clustering method FCM methods are used when the result that two kinds of clustering methods obtain is inconsistent, artificial contrast simultaneously chooses and FCM methods
Closer Clustering Model.
The present invention measures the Net Photosynthetic Rate of the xylophyta blade sample, obtains the Net Photosynthetic Rate data of sample.
In the present invention, the time of the measure of the Net Photosynthetic Rate is 9 points to 11 points of the morning.It will be photosynthetic during the measure
Photon flux density is set in 1600 μm of olm-2·s-1, concentration is set as 400 μm of ol-1.The measure of the Net Photosynthetic Rate is used
Instrument is preferably LI-6400XT.When Net Photosynthetic Rate becomes constant (± 0.1), the value of Net Photosynthetic Rate is recorded.In order to obtain
Accurate data measure the Net Photosynthetic Rate of each blade three times, and the measured value for the Net Photosynthetic Rate that is averaged, more accurate to provide
Data.
After obtaining leaf morphology characteristic and net photosynthetic rate data, the present invention is using gradient boosting algorithm to every
The leaf morphology data of a xylophyta individual of sample phylogenetic group and the Net Photosynthetic Rate data structure model of the sample, obtain
Net Photosynthetic Rate prediction model.
In the present invention, using the method for gradient boosting algorithm structure prediction model, following steps are preferably included：
A. initialization constants c so thatMinimum,To lose letter
Number, N are sample number, yiNet Photosynthetic Rate value for i-th of xylophyta sample；I represents vector, is a six-vector
(x1i, x2i, x3i, x4i, x5i, x6i), x1i, x2i, x3i, x4i, x5i, x6iThe leaf area of i-th of xylophyta sample, leaf are corresponded to respectively
Length, leaf width degree, leaf perimeter, ratio of length to breadth and the leaf shape factor；
B. it to iterations m=1,2...M, if i=1,2...N, calculatesIt obtains
The negative gradient of loss function, using obtained negative gradient as the estimation of residual error, fm(x) for the m times iteration function when, according to rmi
Learn a regression tree；M is the m times iteration, and i is i-th of sample；Function when f (x) is the m-1 times iteration, first time iteration
When be f0(x), i.e.,And f (xi) it is by the leaf area of i-th of xylophyta sample, leaf length, leaf width degree, leaf
Perimeter, ratio of length to breadth and the leaf shape factor substitute into function f (x)；
C. so thatMinimum acquires cmj, update regression tree
Wherein, λ ∈ (0.001,0.01)；M be the m time iteration, j be j-th of region, the leaf of substantially several xylophytas
The set of piece sample；RmjDuring for the m times iteration, all samples have been divided into J set, and j represents j-th of set；
D. the residual error critical value of iteration stopping is calculatedWherein n be sample number, s2For sample side
Difference, s are sample standard deviation, and α is significance；When residual error is less than σIt is criticalWhen, otherwise iteration stopping, it is straight to carry out next iteration
It is less than σ to residual errorIt is critical, obtaining regression tree model is
In the present invention, iteration stopping criterion is used in the gradient boosting algorithm for the first time.The iteration stopping criterion is such as
Under：
Under the hypothesis of gradient lift scheme, the residual error critical value σ of iteration stoppingIt is critical, the σIt is criticalSee formula (3),
Wherein, n is sample number, s2For sample variance, α is significance；
Card：To each predicted value, haveI=1,2..., n consider Hypothesis Testing Problem,
Choose test statistics
WhenWhen, χ2~χ2(n-1),
Then the region of rejection for the inspection that significance is α is
I.e.So
The present invention provides a kind of prediction sides of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature
Method includes the following steps：
A. the leaf morphology feature of xylophyta to be predicted is measured, obtains leaf morphology characteristic；
B. the leaf morphology data are inputted in the prediction model obtained in said program construction method, obtains prediction wood
The predicted value of this plant Net Photosynthetic Rate.
In the present invention, the leaf morphology feature preferably includes leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf week
The long and leaf shape factor.The assay method of the leaf morphology feature is with above-mentioned record, and this will not be repeated here.
With reference to embodiment to a kind of Net Photosynthetic Rate based on xylophyta leaf morphology feature provided by the invention
The construction method and Forecasting Methodology of prediction model are described in detail, but they cannot be interpreted as protecting model to the present invention
The restriction enclosed.
Embodiment 1
S101：In national entire populus simonii geographical distribution (Shaanxi, Qinghai, Hebei, Henan, Ningxia, Beijing, Inner Mongol),
1233 populus simonii samples are acquired, we choose 235 individuals therein and represent.Shandong Province of China province close county (36 ° of 23'N,
115 ° of 47'E) clone plant garden in, vegetative propagation is carried out to it with this 235 individual roots.In our current research, these nothings
The individual of pass is used for research and utilization leaf morphology data prediction Net Photosynthetic Rate.
S102：Blade area, length, width, perimeter, length-width ratio are measured using portable laser leaf area device (CI-202)
With this six leaf morphology data of the leaf factor.
S103：With reference to average linkage cluster and PAM clusters, based on the leaf morphology data that CI-202 measurements obtain, to 235
A division for representing individual and carrying out sub- kind.
The average linkage cluster：
1. the variable that choosing needs is geographical location difference；
2. a pair variable data is normalized；
3. calculate Distance matrix Dij；
4. utilize hclust () function pair DijImplementation level clusters, and two closest classes is divided into one kind, wherein between class
Distance definition is：Point and the average distance at another class midpoint in one class.
PAM is clustered：
1. sample is set as X { x (1), x (2) ... .. }；
2. randomly select k cluster centre in the sample first；
3. and then the distance to each cluster centre is calculated to the sample point outside except cluster centre, by sample be referred to away from
The sample point nearest from center of a sample, this just realizes initial cluster；
4. again to the distance sum of other sample points calculating in each class in addition to the point at class center to other all the points most
Small value, using the minimum point as new cluster, center just realizes primary cluster optimization；
5. repeating step 4, until the position of cluster centre twice no longer changes, this just completes final cluster.
S104：Maximum net photosynthetic rate in order to obtain, in 9 points to 11 points of the morning, by photosynthetic photon flux density (PPFD)
It is set in 1600 μm of olm-2·s-1, concentration is set as 400 μm of ol-1, Net Photosynthetic Rate is measured using LI-6400XT.When net
When photosynthetic rate becomes constant (± 0.1), the value of Net Photosynthetic Rate is recorded.Accurate data in order to obtain, we measure often
The Net Photosynthetic Rate of a blade three times, and the measured value for the Net Photosynthetic Rate that is averaged, to provide more accurately data.
S105：Individual specimen is represented for 235, in every sub- kind after being divided by S103, is utilized respectively ladder
Spend phenotypic data structure Net Photosynthetic Rate prediction model of the boosting algorithm based on blade sample.
S1051, initialization constants c so thatMinimum,
WhereinFor loss function, N is sample number,
yiNet Photosynthetic Rate value for i-th of populus simonii sample；I represents vector, is a six-vector (x1i, x2i,
x3i, x4i, x5i, x6i), x1i, x2i, x3i, x4i, x5i, x6iThe leaf area of i-th of xylophyta sample, leaf length, leaf are corresponded to respectively
Width, leaf perimeter, ratio of length to breadth and the leaf shape factor；
S1052. to iterations m=1,2...M, if i=1,2...N, r is calculatedmiFunction,
For the negative gradient of loss function, using the negative gradient of loss function as the estimation of residual error, fm(x) for the m times iteration function when,
According to rmiLearn a regression tree；
Wherein, m is the m times iteration, and i is i-th of sample；Function when f (x) is the m-1 times iteration, during first time iteration
For f0(x), i.e.,
f(xi) it is by the area of i-th of populus simonii sample, length, width, perimeter, length-width ratio and the leaf factor substitute into letter
Number f (x)；
S1053. pass through minimumAcquire cmj；
Wherein, m be the m time iteration, j be j-th of region, the set of substantially several populus simonii samples；
RmjDuring for the m times iteration, all samples have been divided into J set, and j represents j-th of set；
S1054. the residual error critical value of iteration stopping is calculatedWherein n be sample number, s2For sample
This variance, s are sample standard deviation, and α is significance.When residual error is less than σIt is criticalWhen, otherwise iteration stopping, is changed next time
Generation.
S1055. final regression tree model is
S106：Leaf morphology data (area, length, width, perimeter, the length of populus simonii to be predicted are measured using CI-202
Width than with the leaf factor).
S107：Based on populus simonii leaf morphology data to be predicted, the sub- kind divided using S103, with reference to average
Dynamic cluster and PAM clusters, sub- kind division is carried out to populus simonii to be predicted.
S108：By S107, the sub- kind of populus simonii to be predicted has been obtained, the ladder of S105 structures is used in this sub- kind
Spend lift scheme prediction populus simonii Net Photosynthetic Rate.
Prediction result
Since the geographical location of populus simonii distribution is different, different growing environments makes populus simonii have different characteristic, can quilt
It is divided into different sub- kinds.Therefore, it is more reasonably pre- to be carried out to Net Photosynthetic Rate to eliminate the difference of different sub- kinds
It surveys, carries out cluster analysis using averagely linkage method and PAM methods first when establishing prediction model, as a result as depicted in figs. 1 and 2.
From the point of view of cluster result, first kind populus simonii sample is from Shaanxi (110 degree of east longitude, 34 degree of north latitude) and Henan (east
Through 113 degree, 33 degree of north latitude), the two province geographical locations are close, and climate type belongs to temperate zone subtropical monsoon climate, this
The populus simonii phenotypic characteristic of one subclass is that blade area is big, and the median of blade area has reached 59.95cm2, the leaf factor
Median is 4.52.
Second class populus simonii sample is essentially from western part of China, including Shaanxi, Qinghai, Ningxia.This subclass populus simonii leaf
Piece area median is 25.13cm2, the median of the leaf factor is maximum, and it is 4.60 to be worth.
Third class populus simonii sample covers most provinces, essentially from Hebei, these central plain areas of Henan.It is this kind of small
The blade area of Ye Yang and the second class populus simonii are similar, and blade area median is 24.65cm2, but the leaf factor is also with preceding two
Class populus simonii gap is larger, and leaf factor median is only 4.32.
According to algorithm iteration stopping criterion, first kind populus simonii sample is utilizing gradient boosting algorithm iteration 1534 times
Afterwards, it is upper and lower 1.006 to float to error existing for the prediction of each sample Net Photosynthetic Rate.Second and third class populus simonii sample exists
Respectively after iteration 916 times, 231 times, it is respectively up and down 1.275, upper and lower 1.965 that prediction error, which is floated,.Predictablity rate is three
93.3%, 91.1%, 88.5% has been respectively reached in class populus simonii.
As can be seen that in the case of known leaf morphology feature, the model that the present invention is built is effectively improved net light
Close the precision of prediction of rate.
The above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (8)
1. the construction method of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature, includes the following steps：
1) the xylophyta blade of the same species of China's covering entire geographical distribution of xylophyta is chosen, obtains xylophyta leaf
Piece sample；
2) the leaf morphology feature of the xylophyta blade sample is measured, obtains the leaf morphology number of each xylophyta individual
According to；
The leaf morphology feature includes leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf perimeter and the leaf shape factor；
3) based on the leaf morphology data, using average linkage clustering method and PAM clustering methods to xylophyta individual into
Row Clustering obtains xylophyta individual phylogenetic group；
4) Net Photosynthetic Rate of xylophyta blade sample in the step 1) is measured, obtains the Net Photosynthetic Rate data of sample；
5) blade phenotypic data in each xylophyta individual phylogenetic group after being grouped using gradient boosting algorithm to the step 3)
The Net Photosynthetic Rate data structure model of the sample obtained with the step 4), obtains Net Photosynthetic Rate prediction model；
The Net Photosynthetic Rate prediction model is shown in formula (1)；
Wherein, M represents iterations；xiRepresent six-vector (x1i, x2i, x3i, x4i, x5i, x6i), i-th of populus simonii is corresponded to respectively
The area of sample, length, width, perimeter, length-width ratio and the leaf factor；RmjWhen representing the m times iteration, j-th of populus simonii sample
Set；cmjWhen representing the m times iteration, in j-th of populus simonii sample set, pass through minimum
Obtained constant；
There is no time restriction between the step 2) and step 4).
2. construction method according to claim 1, which is characterized in that construction method in the step 5), including following step
Suddenly：
A. initialization constants c so thatMinimum,For loss function, N
For sample number, yiNet Photosynthetic Rate value for i-th of woody sample；I represents vector, is a six-vector (x1i, x2i,
x3i, x4i, x5i, x6i), x1i, x2i, x3i, x4i, x5i, x6iThe leaf area of i-th of xylophyta sample, leaf length, leaf are corresponded to respectively
Width, leaf perimeter, ratio of length to breadth and the leaf shape factor；
B. it to iterations m=1,2...M, if i=1,2...N, calculatesIt is lost
The negative gradient of function, using obtained negative gradient as the estimation of residual error, fm(x) for the m times iteration function when, according to rmiStudy
One regression tree；M is the m times iteration, and i is i-th of sample；Function when f (x) is the m-1 times iteration, when first time iteration be
f0(x), i.e.,And f (xi) it is the leaf length by the leaf area of i-th of xylophyta sample, leaf width degree, leaf perimeter,
Ratio of length to breadth and the leaf shape factor substitute into function f (x)；
C. so thatMinimum acquires cmj, update regression tree
Wherein, λ ∈ (0.001,0.01)；M be the m time iteration, j be j-th of region, the blade sample of substantially several xylophytas
This set；RmjDuring for the m times iteration, all samples have been divided into J set, and j represents j-th of set；
D. the residual error critical value of iteration stopping is calculatedWherein n be sample number, s2For sample variance, s
For sample standard deviation, α is significance；When residual error is less than σIt is criticalWhen, otherwise iteration stopping, carries out next iteration until residual
Difference is less than σIt is critical, obtaining regression tree model is
3. construction method according to claim 1, which is characterized in that using the clustering method that averagely links in the step 3)
The method that Clustering is carried out to xylophyta individual, includes the following steps：
1. the phenotypic characteristic data of xylophyta sample are chosen as the variable needed；
2. the variable data is normalized, normalization variable data is obtained；
3. calculate Distance matrix D using the normalization variable dataij；
4. utilize hclust () function pair Distance matrix D of R languageijImplementation level clusters, and two closest classes are divided into one
Class, wherein between class distance are defined as：Point and the average distance at another class midpoint in one class.
4. construction method according to claim 1, which is characterized in that the PAM clustering methods include the following steps：
A, sample is set as X { x (1), x (2) ... .. }；
B. k cluster centre is randomly selected in the sample；
C. the distance to each cluster centre is calculated the sample point in addition to cluster centre, sample is referred to distance sample center
Nearest sample point obtains initial cluster；
D. to the distance sum of other sample points calculating in each initial cluster in addition to the point at class center to other all the points most
The minimum value is just realized primary cluster optimization by small value；
E. step d is repeated, until the position of cluster centre twice no longer changes, obtains cluster result.
5. construction method according to claim 1, which is characterized in that the quantity of xylophyta sample is in the step 1)
200~500.
6. construction method according to claim 1, which is characterized in that step 2) the middle period form factor calculation formula is such as
Shown in formula (2)；
7. construction method according to claim 1, which is characterized in that in the step 4) measure of Net Photosynthetic Rate when
Between for 9 points to 11 points of the morning, photosynthetic photon flux density is set in 1600 μm of olm during the measure-2·s-1, concentration sets
It is set to 400 μm of ol-1。
8. it is a kind of based on the prediction model that construction method claim 1~7 described obtains to the pre- of xylophyta Net Photosynthetic Rate
Survey method, includes the following steps：
A. the leaf morphology feature of xylophyta to be predicted is measured, obtains leaf morphology characteristic, the leaf morphology feature
Including leaf area, leaf length, leaf width degree, ratio of length to breadth, leaf perimeter and the leaf shape factor；
B. the leaf morphology data are inputted in the prediction model obtained in said program construction method, obtains predicting woody plant
The predicted value of object Net Photosynthetic Rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810119071.1A CN108229065B (en) | 2018-02-06 | 2018-02-06 | A kind of construction method and prediction technique of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810119071.1A CN108229065B (en) | 2018-02-06 | 2018-02-06 | A kind of construction method and prediction technique of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108229065A true CN108229065A (en) | 2018-06-29 |
| CN108229065B CN108229065B (en) | 2019-03-26 |
Family
ID=62671095
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810119071.1A Expired - Fee Related CN108229065B (en) | 2018-02-06 | 2018-02-06 | A kind of construction method and prediction technique of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108229065B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829943A (en) * | 2018-11-13 | 2019-05-31 | 上海烟草集团有限责任公司 | Blade construction detection method, system, medium and equipment based on machine vision |
| CN113159560A (en) * | 2021-04-15 | 2021-07-23 | 中国科学院地理科学与资源研究所 | Optimized configuration method for psammophyte industry |
| CN114814099A (en) * | 2022-04-25 | 2022-07-29 | 南京农业大学 | Photosynthesis prediction method based on grape leaf shape |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011223915A (en) * | 2010-04-19 | 2011-11-10 | Ohbayashi Corp | System and method for predicting growth of lawn |
| CN103234916A (en) * | 2013-04-06 | 2013-08-07 | 吉林大学 | Prediction method for net photosynthetic rate of population |
| CN105678405A (en) * | 2015-12-31 | 2016-06-15 | 西北农林科技大学 | Modeling method for cucumber GA-RBF photosynthetic rate prediction model integrated with stomatal conductance |
| CN105956380A (en) * | 2016-04-25 | 2016-09-21 | 东北林业大学 | Construction method of leaf area prediction model |
| CN107037090A (en) * | 2017-04-24 | 2017-08-11 | 江苏大学 | A kind of photosynthesis Forecasting Methodology changed based on blade tensity |
-
2018
- 2018-02-06 CN CN201810119071.1A patent/CN108229065B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011223915A (en) * | 2010-04-19 | 2011-11-10 | Ohbayashi Corp | System and method for predicting growth of lawn |
| CN103234916A (en) * | 2013-04-06 | 2013-08-07 | 吉林大学 | Prediction method for net photosynthetic rate of population |
| CN105678405A (en) * | 2015-12-31 | 2016-06-15 | 西北农林科技大学 | Modeling method for cucumber GA-RBF photosynthetic rate prediction model integrated with stomatal conductance |
| CN105956380A (en) * | 2016-04-25 | 2016-09-21 | 东北林业大学 | Construction method of leaf area prediction model |
| CN107037090A (en) * | 2017-04-24 | 2017-08-11 | 江苏大学 | A kind of photosynthesis Forecasting Methodology changed based on blade tensity |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829943A (en) * | 2018-11-13 | 2019-05-31 | 上海烟草集团有限责任公司 | Blade construction detection method, system, medium and equipment based on machine vision |
| CN109829943B (en) * | 2018-11-13 | 2023-02-10 | 上海烟草集团有限责任公司 | Blade structure detection method, system, medium and equipment based on machine vision |
| CN113159560A (en) * | 2021-04-15 | 2021-07-23 | 中国科学院地理科学与资源研究所 | Optimized configuration method for psammophyte industry |
| CN113159560B (en) * | 2021-04-15 | 2023-12-05 | 中国科学院地理科学与资源研究所 | Optimization configuration method for sandy plant industry |
| CN114814099A (en) * | 2022-04-25 | 2022-07-29 | 南京农业大学 | Photosynthesis prediction method based on grape leaf shape |
| CN114814099B (en) * | 2022-04-25 | 2023-09-12 | 南京农业大学 | Photosynthesis prediction method based on grape leaf shape |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108229065B (en) | 2019-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sinclair et al. | An analysis of errors in the calculation of energy flux densities above vegetation by a Bowen-ratio profile method | |
| Shipley et al. | Dry matter content as a measure of dry matter concentration in plants and their parts | |
| Overman et al. | Evaluation of regression models for above-ground biomass determination in Amazon rainforest | |
| CN108229065B (en) | A kind of construction method and prediction technique of the Net Photosynthetic Rate prediction model based on xylophyta leaf morphology feature | |
| CN108319984B (en) | The construction method and prediction technique of xylophyta leaf morphology feature and photosynthesis characteristics prediction model based on DNA methylation level | |
| Luo et al. | Portable canopy chamber measurements of evapotranspiration in corn, soybean, and reconstructed prairie | |
| Bai et al. | Modelling diurnal and seasonal hysteresis phenomena of canopy conductance in an oasis forest ecosystem | |
| Biscoe et al. | Barley and its environment. I. Theory and practice | |
| Yu et al. | A PSO-XGBoost Model for Estimating Daily Reference Evapotranspiration in the Solar Greenhouse. | |
| CN109214591B (en) | Method and system for predicting aboveground biomass of woody plant | |
| CN110334404A (en) | A kind of rapid dry accurate recognition methods of drought of Watershed Scale | |
| CN102903145A (en) | Method for three-dimensional reconstruction of plant population morphological structure | |
| Thakur et al. | Increase in light interception cost and metabolic mass component of leaves are coupled for efficient resource use in the high altitude vegetation | |
| CN106885892A (en) | The method and device split to forest ecosystem evapotranspiration | |
| CN101044823A (en) | Method for estimating crop energy utilization rate and predetermining the yield | |
| Meir et al. | Plant structure-function relationships and woody tissue respiration: upscaling to forests from laser-derived measurements | |
| Sugiura et al. | Concentrative nitrogen allocation to sun-lit branches and the effects on whole-plant growth under heterogeneous light environments | |
| Evans et al. | Inventory time-cost and statistical power: a case study of a Lao rattan | |
| Zhang et al. | Use of a regression method to partition sources of ecosystem respiration in an alpine meadow | |
| CN116705144A (en) | Method for measuring correlation between species distribution data and environmental factors | |
| CN116663779A (en) | Multi-depth fertility point-surface conversion method and device for cultivated land soil | |
| Zhang et al. | Comparison of the partitioning of evapotranspiration–numerical modeling with different isotopic models using various kinetic fractionation coefficients | |
| Zhong et al. | Organic carbon storage in soils of tropical and subtropical China | |
| MATSUDA et al. | Dry deposition velocity of sulfur dioxides on a red pine forest in Nagano, Japan | |
| CN108918578A (en) | A kind of Heat Tolerance of Wheat Cultivars brief appraisal method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190326 Termination date: 20210206 |