CN105654203A - Cucumber whole-course photosynthetic rate predicting model based on support vector machine, and establishing method - Google Patents

Abstract

The invention relates to a cucumber whole-course photosynthetic rate predicting model based on a support vector machine. Photosynthetic rate test data of cucumber seedlings is obtained by utilizing a multi-factor nested experiment; model training is carried out by adopting a LM training method; a cucumber whole-course photosynthetic rate model fused with multiple growing periods is established; the cucumber whole-course photosynthetic rate model is subjected to comparison verification with a single-growing-period photosynthetic rate model and a whole-growing-period cucumber photosynthetic rate model respectively by adopting an abnormal check manner; the result shows that the whole-course photosynthetic rate model, which is established by adding the growing period as the one-dimensional input quantity, can effectively cross a local flat region, have obvious superiority, and satisfy training requirements that the error is less than 0.0001; the determination coefficient of a predicted value and a practically measured value of the model is 0.993; the error is less than 6.253%; both the training effect and the model-fitting degree of the model are superior to the mixed-growing period model; the precision of the model is similar to that of the single-growing-period photosynthetic rate model; and thus, the model disclosed by the invention can provide theoretical basis and technical support for light environment adjustment and control of facility crops.

Description

The omnidistance photosynthetic rate predictive model of a kind of cucumber based on SVMs and establishment method

Technical field

The invention belongs to reading intelligent agriculture technical field, in particular to the omnidistance photosynthetic rate predictive model of a kind of cucumber based on SVMs and establishment method.

Background technology

Cucumber is one of main vegetables of China's cultivation, and it is inseparable that the quality and yield of cucumber carries out photosynthetic ability with it. Photosynthetic rate and chlorophyll content, temperature, CO₂Multiple factors such as concentration, intensity of illumination, relative humidity have remarkable relation. Wherein, chloroplast(id) is that green plants carries out photosynthetic basis organoid, and chlorophyll is the basic composition material of chloroplast(id), in photosynthesis of plant most important, its content is the important indicator of photosynthesis of plant ability, nutritional status and growth situation, effect of temperature makes activity, the stomatal conductance of Rubisco activating enzymes in object, CO₂Concentration directly affects the accumulation that crop carries out dark reaction speed and dry-matter, and intensity of illumination is photosynthetic direct driving force and motive force, and relative humidity affects leaf stomatal conductance etc., and between each factor, existence influences each other. Therefore, consider the omnidistance photosynthetic rate predictive model that multiple factor affects, sets up multiplefactor coupling, optimization cucumber luminous environment is had vital role.

External a lot of relevant scholar and research institution, by the further investigation to photosynthesis of plant, and establish a large amount of relevant inside greenhouse environmental Kuznets Curves models and the model of plant-growth based on this. The seventies, it is one of initial model setting up leaf photosynthesis model that Charles-Edwards proposes photosynthesis of plant physiological models, and wherein physiological models comprises photorespiration, Dark respiration and oxygen effect. On this correlative study basis, relevant scholar establishes multiple Photosynthetic model, and comprising models such as right angle hyperbolic model, non-right angle hyperbolic model and exponential relationships, but its model parameter not easily obtains, and brings certain difficulty to the application of model. Based on above-mentioned physiological models, ZipiaoYe etc. propose the photosynthetic rate model etc. based on electron transport and propose photosynthetic rate steady-state model, J.Z.XU etc. have carried out the research of photosynthesis model under different nitrogen, Y.LANG etc. utilize different blades, have carried out correlative study and the exploration of photosynthetic rate model.So, choose plant photosynthesis model of good performance and determine correlation parameter comparatively accurately for the environment of regulating plant growth and and crop cultivate and seem and be more necessary, but the domestic model in heliogreenhouse crop net photosynthesis also needs to update at present.

In recent years, numerous scholar has carried out correlative study setting up in photosynthetic rate model, and above-mentioned research all considers the association between different environmental factor, but there is the deficiencies such as degree of fitting is lower, fitting formula is complicated, error is bigger. And neural network has the advantage such as nonlinear mapping and adaptive learning ability, suitable matching and prediction complicated nonlinear system model, therefore photosynthetic rate modeling based on neural network has become research focus. Occur based on Hopfield network photosynthetic rate model, the greenhouse tomato leaf stomatal conductance model based on BP neural network, the tomato single leaf net photosynthesis effect rate prediction model in flowering period based on WSN in the recent period, above-mentioned research from different angles by Application of Neural Network in photosynthetic rate modeling, but all do not consider that different growing stages is on the impact of crop, not yet set up omnidistance cucumber photosynthetic rate predictive model, and it is relatively slow to there is training process, the deficiency that training error difference is bigger.

SVMs is a kind of new general-purpose machinery learning method under Corpus--based Method study theoretical frame. The problems such as the non-linear classification of the height in sample space and recurrence can be solved, it is a kind of effective ways processing non-linear classification and non-linear regression. Photosynthetic rate prediction comprises a large amount of non-linear factors. Traditional statistical prediction methods, requires the factor and exists significant linear relevant between forecasting object when setting up predictive model, and require between the factor to be linearly correlated with reach minimum. And cause photosynthetic rate to change the complicacy of all multiplefactors and non-linear, what determine between predictor to forecasting object is non-linear relevant, thus traditional model prediction method is difficult to solve the forecasting problem that essence is nonlinear relationship, and SVMs is that plant photosynthetic rate prediction provides a kind of feasible effective way.

Summary of the invention

In order to overcome the shortcoming of above-mentioned prior art, it is an object of the invention to provide the omnidistance photosynthetic rate predictive model of a kind of cucumber based on SVMs, design multiplefactor Nested simulation experiment, model construction of SVM is adopted after being processed by data normalization, set up the cucumber photosynthetic rate predictive model only distinguished as a dimension input factor for seedling phase, the phase of yielding positive results, full vegetative period and the vegetative period of adding respectively, omnidistance cucumber photosynthetic rate predictive model is set up, for basis is set up in the luminous environment regulation and control of industrialized agriculture by contrast verification.

In order to realize above-mentioned purpose, the technical solution used in the present invention is:

Based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, model formation is:Wherein, the photosynthetic rate that f (x) represents prediction is exported, input signal X'=(X₁'X₂'��X₅')^T, X₁'��X₂'��X₃'��X₄'��X₅' it is respectively temperature, CO₂Concentration, intensity of illumination, relative humidity and chlorophyll content, w is weight vector, and b is biased, and �� (x) is nonlinear mapping function, and l is that training set sample is to { (x_i,y_i), i=1,2,3 ..., the learning sample number in l}, x_iIt is the input column vector of the i-th learning sample,y_iFor the output value of correspondence, y_i�� R,Being that i �� d ties up real number field, d is column vector dimension, a_iAnd a_i ^*Optimum solution for following formula:

For kernel function, �� is width parameter, and �� is for stopping training error, and c is the punishment factor.

The establishment method of the omnidistance photosynthetic rate predictive model of the described cucumber based on SVMs, comprises the steps:

Step 1, obtains experimental data, and process is as follows:

Adopt feeding block seedlings raising, treat that cucumber seedling grows up to two leaves wholeheartedly, growing way cucumber seedling even, that the horizontal footpath of stem is between 0.6��0.8cm, within strain height 10cm is selected to test, choose healthy and strong cucumber seedling 150 strain as test sample, treat that cucumber is in the phase of yielding positive results, choose the test sample of plant 150 strain as the phase of yielding positive results of about 50 centimetres, Flower node distance tap;

Measure Net Photosynthetic Rate, process utilizes temperature control module setting 16,20,24,28,32 DEG C totally 5 thermogrades; Utilize CO₂Injection module setting carbonic acid gas volume ratio is 300,600,900,1200,1500 �� L/L totally 5 gradients; LED light source module is utilized to obtain 0,20,50,100,200,300,500,700,1000,1200,1500 ��m of ol/ (m²S) totally 11 photon flux density gradients, carrying out 275 groups of tests in a nesting relation altogether, often group test does repeated test on the 3 strain plant chosen at random, records leaf room relative humidity in test, and record tested chlorophyll content in leaf blades, thus formed with chlorophyll content, temperature, CO₂Concentration, intensity of illumination, relative humidity are input, and Net Photosynthetic Rate is the 1650 groups of experimental datas exported, i.e. seedling phase 825 groups, phase 825 groups of yielding positive results;

Step 2, Modling model

The experimental data that step 1 obtains is chosen learning sample and test sample book at random, then choosing SVMs kernel function type is RBF, determine that the optimizing of SVMs punishment factor c and Radial basis kernel function parameter g is interval by grid method, utilize genetic algorithm to parameter c and g based on the interval optimizing further of described optimizing until reach maximum iteration time, the optimizing result of output parameter c and g, builds the omnidistance photosynthetic rate predictive model of the cucumber based on SVMs.

In described step 2,

Experimental data normalized is obtained entirely organizing sampled data to [0.2,0.9] interval, chooses sample number at random in 4:1 ratio and build training sample set, test sample book collection.

In described step 2,

The optimizing interval table of punishment factor c and Radial basis kernel function parameter g illustrated as lgc �� [2,3], lgg �� [0,1].

In described step 2,

Adopt genetic algorithm to the interval optimizing further of the optimizing of punishment factor c and Radial basis kernel function parameter g, process is: setting genetic algorithm parameter, individual amount NIND=100, maximum genetic algebra MAXGEN=50, generation gap GGAP=0.95, crossover probability and variation probability be 0.8,0.25, c and g as variable, represent with 10 scale-of-two respectively; Objective function is model square error value MSE; Genetic operator operates: selects, intersect, make a variation; Calculate the slotting filial generation of laying equal stress on of filial generation target value function and obtain new population to parent; Reach the optimizing result that maximum iteration time exports c and g.

In described step 2,

The model built is:Wherein, input signal X'=(X₁'X₂'��X₅')^T, X₁'��X₂'��X₃'��X₄'��X₅' it is respectively temperature, CO₂Concentration, intensity of illumination, relative humidity and chlorophyll content, w is weight vector, and b is biased, and �� (x) is nonlinear mapping function.

After obtaining the omnidistance photosynthetic rate predictive model of the cucumber based on SVMs, input amendment data carry out photosynthetic efficiency prediction, if reaching accuracy requirement, then export and predict the outcome, otherwise return genetic algorithmic steps, to c and g based on the interval optimizing further of described optimizing.

Compared with prior art, the invention has the beneficial effects as follows:

1) model is built based on algorithm of support vector machine. SVMs topological framework is determined by support vector, and applies kernel function and nonlinear transformation maps the linear transformation in high dimension space, had both ensured that model had good generalization ability, and had solved again the problem of " dimension disaster ". Avoid tradition neural network to need to try to gather the problem determining network structure. In addition, modelling verification result shows, in the regression fit problem for small sample, the prediction effect of SVM model to be obviously better than BP neural network, and effectively avoids BP neural network Local Minimum problem.

2) the model parameter optimizing that grid method is combined with genetic algorithm. Traditional support vector machine parameter determines many employing empirical methods or K-fold cross validation method, owing to sampled data type is different, parameter value is chosen exists larger difference, therefore often cause that model error is relatively big or optimizing overlong time, herein by grid method pre-training Confirming model parameter Search Range, and then use genetic algorithm that above-mentioned scope is carried out optimizing. Empirical tests, under choosing same sample data at random, building model actual measurement value and analogue value square error by present method is 0.000409, the coefficient of determination is 0.9915, and adopt K-fold cross validation method structure model actual measurement value and analogue value square error to be 0.000860, the coefficient of determination is 0.9769, and prediction precision effectively improves.

3) Model Fusion time series, consider the change of Different growth phases plant photosynthetic rate, by adding one-dimensional growth phase variable, the photosynthetic rate value effectively having distinguished cucumber seedling phase and the phase of yielding positive results difference at different conditions, builds omnidistance photosynthetic rate predictive model. Prediction effect shows, and predictor and the measured value coefficient of determination are 0.993, and worst error is 3.571, and relative error is less than 6.253%, and prediction precision is higher than the predictive model of mixed growth phase.

The omnidistance photosynthetic rate predictive model that the present invention proposes can be the regulation and control of cucumber luminous environment and provides theoretical foundation, can expanded application set up in the photosynthetic optimization regulation-control model of Different Crop, to improve the photosynthetic capacity of chamber crop.

Accompanying drawing explanation

Fig. 1 is that the present invention is based on algorithm of support vector machine schema.

Fig. 2 is different IPs function Type model prediction effect figure of the present invention, wherein Fig. 2 (a) represents linear kernel function predictor, Fig. 2 (b) representative polynomial kernel function predictor, Fig. 2 (c) represents Radial basis kernel function predictor, and Fig. 2 (d) represents sigmoid kernel function predictor.

Fig. 3 be the present invention based on grid method Confirming model parameter Search Range, wherein Fig. 3 (a) represents the change of training set sample square error with c, g, and wherein Fig. 3 (b) represents the change of test set sample square error with c, g.

Fig. 4 is that the present invention is based on genetic algorithm optimization model parameter schema.

Fig. 5 is that the present invention is based on genetic algorithm optimization model parameter evolutionary process figure.

Fig. 6 is dependency diagram between photosynthetic rate measured value and the analogue value in modelling verification of the present invention. Wherein, Fig. 6 (a) is dependency diagram between photosynthetic rate measured value and the analogue value in the photosynthetic rate modelling verification of plant seedlings phase; Fig. 6 (b) is dependency diagram between photosynthetic rate measured value and the analogue value in the photosynthetic rate modelling verification of flowering of plant result phase; Fig. 6 (c) is dependency diagram between photosynthetic rate measured value and the analogue value in the full phase plant photosynthetic rate modelling verification of mixed growth;Fig. 6 (d) is dependency diagram between photosynthetic rate measured value and the analogue value in the full phase photosynthetic rate modelling verification of fusion growth.

Embodiment

Below in conjunction with drawings and Examples, embodiments of the present invention are described in detail.

The process of establishing of the omnidistance photosynthetic rate predictive model of a kind of cucumber based on neural network of the present invention is as follows:

1, test materials and method

This is tested and carries out in Xibei Univ. of Agricultural & Forest Science & Technology's scientific research greenhouse in April, 2014 to July. Supply examination cucumber variety to be " Chang Chun Mi Ci ", in culture dish, carry out vernalization by soaking swollen seed, in time being sprouted, carry out subzero treatment, in the dish of 50 holes (540mm280mm50mm) cave, adopt feeding block seedlings raising. Seedling medium is agricultural special seedling substrate. During seedling culture, keep liquid manure sufficient, treat that cucumber seedling grows up to two leaves wholeheartedly, select growing way cucumber seedling even, that the horizontal footpath of stem is between 0.6��0.8cm, within strain height 10cm to test. Choose healthy and strong cucumber seedling 150 strain as test sample. Between trial period, carry out normal field planting management, do not spray any agricultural chemicals and hormone, treat that cucumber is in the phase of yielding positive results, choose the test sample of plant 150 strain as the phase of yielding positive results of about 50 centimetres, Flower node distance tap.

The portable photosynthetic instrument of the Li-6400XT type adopting U.S. LI-COR company to produce measures Net Photosynthetic Rate, adopts the parameters such as the temperature around multiple submodule block control on demand blades that photosynthetic instrument matches, CO2 concentration, intensity of illumination in process of the test. Wherein, temperature control module setting 16,20,24,28,32 DEG C totally 5 thermogrades are utilized; Utilize CO₂Injection module setting carbonic acid gas volume ratio is 300,600,900,1200,1500 �� L/L totally 5 gradients; LED light source module is utilized to obtain 0,20,50,100,200,300,500,700,1000,1200,1500 ��m of ol/ (m²S) totally 11 photon flux density (Photofluxdensity, PFD) gradient, carry out 275 groups of tests in a nesting relation altogether, often group test does repeated test on the 3 strain plant chosen at random, test records leaf room relative humidity, and adopt the SPAD-502Plus type tested chlorophyll content in leaf blades of chlorophyll meter record of Konica company of Japan, thus formed with chlorophyll content, temperature, CO₂Concentration, intensity of illumination, relative humidity are the 1650 groups of experimental datas exported for input, Net Photosynthetic Rate, i.e. the seedling phase 825 groups, yield positive results the phase 825 groups.

2, method for establishing model

2.1 Support vector regression basic theories

Originally SVM is the classification problem (SVC) for solving two class samples in linear separability situation, and its core concept finds an optimal separating hyper plane, makes the classification margin maximization of two class samples. When SVM is applied to regression fit analysis, its basic thought is no longer find an optimal classification plane to make two class samples separately, but finds an optimal classification surface and make all learning sample minimum from the error of this optimal classification plane.

Do not lose generality, if the training set sample containing l learning sample is to being { (x_i,y_i), i=1,2,3 ..., l}, wherein,Being the input column vector of i-th learning sample, d is column vector dimension,It is that i �� d ties up real number field,y_i�� R is the output value of correspondence.

Be located in high-dimensional feature space set up linear regression function be

F (x)=w �� (x)+b (2-1)

Wherein x is input vector, and w is weight vector, and b is biased, and �� (x) is nonlinear mapping function.

The definition linear insensitive loss function of ��

The predictor that wherein f (x) returns for regression function; Y is corresponding true value, if the difference namely represented between predictor and true value is less than or equal to ��, then loss equals 0.

For linear regression problem, problem turns into seeking an optimal hyperlane so that under given precision (�� >=0) condition can free from errors matching y, namely all sample points are all not more than �� to the distance of optimal hyperlane; Consider the situation of permissible error, slack variable �� can be introduced_i, ��_i ^*>=0 its optimization problem transforms corresponding quadratic programming problem:

Wherein c is the punishment factor, and the sample punishment that training error is greater than �� by the more big expression of c is more big, and �� defines the error requirements of regression function, and the error of the more little expression regression function of �� is more little.

2-3 formula Solve problems can be converted into dual problem:

Wherein a_i, a_i ^*For (2-4) formula optimum solution.

Solve the problems referred to above to solve and can obtain optimum regression function and be:

Wherein K (x_i,x_j)=�� (x_i)��(x_j) it is kernel function.

2.2 SVMs kernel functions are chosen

The kernel function of SVM is by the feature space of Nonlinear separability sample conversion to linear separability, and the Optimal Separating Hyperplane that different Selection of kernel functions can make SVM model produce is different, produces bigger otherness, the performance of SVM model is had direct impact. Therefore, choosing of kernel function is the key affecting SVM predictive model. Conventional kernel function has: linear function, polynomial function, RBF, sigmoid function etc. This fa.m carries out model pre-training for above four kinds of kernel functions respectively, and selects square error and the coefficient of determination two as evaluation index. Model prediction result such as table 1, shown in Fig. 2.

Table 1 kernel function is on the impact of model performance

Table1Theimpactontheperformanceofthemodelkernel

As shown in Table 1, RBF and Polynomial kernel function have less experience error compared to linear function and sigmoid function, but polynomial function generalization ability is poor, and compared with Radial basis kernel function, polynomial function it needs to be determined that parameter with many, as shown in Figure 2, Radial basis kernel function measured value and analogue value fitting effect are best. Therefore, consider training effect and complexity, choose the kernel function of RBF as this paper SVM photosynthetic rate predictive model.

2.3 SVMs nuclear parameters are chosen

SVM, as the predictive model of a kind of Corpus--based Method scientific principle opinion, adopts its difficult point carrying out predicting to be the selection to model parameter. People is often according to experience in prediction, selects suitable parameter by repetition test. This can not ensure that model can converge to global minima, and the also nature that predicts the outcome cannot ensure optimum. For the actually operating personnel not having theory basis, select optimized parameter extremely difficult especially, which also limits the application of SVM model. SVMs parameter comprises punishment factor c, Radial basis kernel function parameter g, rank number p, stops training error �� etc.

Wherein penalty factor c is a coefficient going by user to specify, represent and a point wrong point is added how many punishment, when c is very big time, point wrong point will be less, but the situation of over-fitting may compare seriously, when c is very little time, a point wrong point may be a lot, and the model thus obtained can be not too correct.

The difference that nuclear parameter g chooses, can there is corresponding change in the form of function, and then cause the change of SVM model.

The present invention adopts grid method to be roughly selected by parameter area and gets. Choose RBF as model kernel function, select square error MSE as evaluation index, respectively punishment factor c and kernel functional parameter g is carried out optimizing, Search Range lgc �� [-5,5], lgg �� [-5,5], optimizing result Confirming model parameter c, the scope of g is lgc �� [2,3], lgg �� [0,1], as shown in Figure 3.

Utilizing genetic algorithm to carry out optimizing based on above-mentioned scope, it is determined that c, g parameter is specifically worth, as shown in Figure 4, genetic evolution process is as shown in Figure 5 for concrete steps.

2.4 based on the photosynthetic rate model construction of SVMs

Difference in vegetative period for cucumber adopts same modeling method to build together vertical four kinds of models, be respectively only for the cucumber seedling phase predictive model, only yield positive results the predictive model of phase, the photosynthetic rate predictive model of cucumber whole process and the difference in vegetative period is set up as a dimension input predictive model of cucumber whole process for cucumber. Input signal is X'=(X₁'X₂'��X₅')^T, X₁'��X₂'��X₃'��X₄'��X₅' it is respectively temperature, CO₂Concentration, intensity of illumination, relative humidity and chlorophyll content, the 4th kind of model adds vegetative period as a dimension input, and output signal all uses T_oRepresenting the photosynthetic rate that network calculations obtains, often group corresponding actual measurement photosynthetic rate is teacher signal T_d. Omnidistance cucumber seedling photosynthetic rate model T is set up by SVMs coaching method_d'(X')��

3 model training performance analysiss

Based on above-mentioned test sample collection, adopting algorithm of support vector machine to train, obtain four kinds of models, the cucumber predictive model that the seedling phase sets up, training error is 0.000446, it is resolved that coefficient is 0.9883;

The cucumber predictive model bloomed and set up vegetative period, training error is 0.000387, it is resolved that coefficient is 0.9900; Mixing the omnidistance cucumber predictive model in two kinds of vegetative period, training error is 0.0022, it is resolved that coefficient is 0.9419; Adding vegetative period as a dimension independent variable(s) factor, set up the cucumber predictive model merging two kinds of vegetative period, training error is 0.00021128, it is resolved that coefficient is 0.9943. Comparative analysis training result can find, trains model all to reach the error level of expectation stage by stage, and training set measured value and the analogue value have good facies relationship; Mix the omnidistance cucumber predictive model in two kinds of vegetative period, training error is bigger, it is poor that training set measured value and the analogue value have good facies relationship, and add the fusion forecasting model of vegetative period as a dimension independent variable(s) factor, training error and the coefficient of determination are all better than phasing model, and model performance reaches optimization.

Based on the above results, add the modelling effect set up vegetative period as a dimension factor remarkable, it is possible to for luminous environment regulation and control provide theoretical basis and technical support, simplify the operation of luminous environment equipment.

4 modelling verification results are analyzed

The test sample collection totally 1650 Ge Liang group obtained with multiplefactor Nested simulation experiment, sample is divided into training set and test set, wherein 700 for model training, remain 175 groups for forming test set, account for the 20% of total sample, adopt different verification method to carry out modelling verification, obtain photosynthetic rate measured value and predictor correlation analysis as shown in the figure. can find from Fig. 6, in Fig. 6 a, the coefficient of determination of SVM machine model measured value and predictor correlation analysis is 0.988, straight slope is 0.984, intercept is 0.2066, largest prediction error 1.4672, in Fig. 6 b, the coefficient of determination of a SVM model actual measurement value and predictor correlation analysis is 0.986, straight slope is 0.9864, intercept is 0.3202, largest prediction error 2.7186, in Fig. 6 c, the coefficient of determination of SVM model actual measurement value and predictor correlation analysis is 0.884, straight slope is 0.9661, intercept is 0.7343, largest prediction error 12.55, in Fig. 6 d, the coefficient of determination of SVM model actual measurement value and predictor correlation analysis is 0.993, straight slope is 0.9923, intercept is 0.05523, largest prediction error 3.575, consider that the linear lag of Modling model in vegetative period is obviously higher, fitting degree is better.

Test-results is carried out errot analysis known, consider that measured value and the analogue value maximum relative error of the omnidistance photosynthetic rate predictive model that vegetative period sets up are less than �� 6.253%, show that the model set up can carry out the photosynthetic rate model prediction in full vegetative period herein, have good precision.

Claims (7)

1. the omnidistance photosynthetic rate predictive model of the cucumber based on SVMs, it is characterised in that, model formation is: $f\left(x\right)=w\mathrm{\Φ}\left(x\right)+b=\underset{i=1}{\overset{l}{\Σ}}(a_i-{a_i}^{*})\·\exp (-\frac{\left|\right|x-x_i\left|\right|}{2{\mathrm{\σ}}^2})+b,$ Wherein, the photosynthetic rate that f (x) represents prediction is exported, input signal X'=(X₁'X₂'��X₅')^T, X₁'��X₂'��X₃'��X₄'��X₅' it is respectively temperature, CO₂Concentration, intensity of illumination, relative humidity and chlorophyll content, w is weight vector, and b is biased, and �� (x) is nonlinear mapping function, and l is that training set sample is to { (x_i,y_i), i=1,2,3 ..., the learning sample number in l}, x_iIt is the input column vector of the i-th learning sample,y_iFor the output value of correspondence, y_i�� R,Being that i �� d ties up real number field, d is column vector dimension, a_iAnd a_i ^*Optimum solution for following formula:

$\left\{\begin{array}{c}\underset{\mathrm{\α},{\mathrm{\α}}^{*}}{\max }\[-\frac{1}{2}\underset{i=1}{\overset{l}{\Σ}}\underset{j=1}{\overset{l}{\Σ}}\left({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*}\right)\left({\mathrm{\α}}_j-{\mathrm{\α}}_j^{*}\right)K\left(x_i,x_j\right)-\underset{i=1}{\overset{l}{\Σ}}\left({\mathrm{\α}}_i+{\mathrm{\α}}_i^{*}\right)\mathrm{\ϵ}+\underset{i=1}{\overset{l}{\Σ}}\left({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*}\right)y_i\]\\ s.t.\left\{\begin{array}{c}\underset{i=1}{\overset{l}{\Σ}}\left({\mathrm{\α}}_i-{\mathrm{\α}}_i^{*}\right)=0\\ 0\≤{\mathrm{\α}}_i\≤c\\ 0\≤{\mathrm{\α}}_i^{*}\≤c\end{array}\right.\end{array}\right.$

For kernel function, �� is width parameter, and �� is for stopping training error, and c is the punishment factor.

2. based on the establishment method of the omnidistance photosynthetic rate predictive model of the cucumber of SVMs described in claim 1, it is characterised in that, comprise the steps:

Step 1, obtains experimental data, and process is as follows:

Adopt feeding block seedlings raising, treat that cucumber seedling grows up to two leaves wholeheartedly, growing way cucumber seedling even, that the horizontal footpath of stem is between 0.6��0.8cm, within strain height 10cm is selected to test, choose healthy and strong cucumber seedling 150 strain as test sample, treat that cucumber is in the phase of yielding positive results, choose the test sample of plant 150 strain as the phase of yielding positive results of about 50 centimetres, Flower node distance tap;

Measure Net Photosynthetic Rate, process utilizes temperature control module setting 16,20,24,28,32 DEG C totally 5 thermogrades; Utilize CO₂Injection module setting carbonic acid gas volume ratio is 300,600,900,1200,1500 �� L/L totally 5 gradients; LED light source module is utilized to obtain 0,20,50,100,200,300,500,700,1000,1200,1500 ��m of ol/ (m²S) totally 11 photon flux density gradients, carrying out 275 groups of tests in a nesting relation altogether, often group test does repeated test on the 3 strain plant chosen at random, records leaf room relative humidity in test, and record tested chlorophyll content in leaf blades, thus formed with chlorophyll content, temperature, CO₂Concentration, intensity of illumination, relative humidity are input, and Net Photosynthetic Rate is the 1650 groups of experimental datas exported, i.e. seedling phase 825 groups, phase 825 groups of yielding positive results;

Step 2, Modling model

The experimental data that step 1 obtains is chosen learning sample and test sample book at random, then choosing SVMs kernel function type is RBF, determine that the optimizing of SVMs punishment factor c and Radial basis kernel function parameter g is interval by grid method, utilize genetic algorithm to parameter c and g based on the interval optimizing further of described optimizing until reach maximum iteration time, the optimizing result of output parameter c and g, builds the omnidistance photosynthetic rate predictive model of the cucumber based on SVMs.

3. according to claim 2 based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, establishment method, it is characterised in that, in described step 2,

Experimental data normalized is obtained entirely organizing sampled data to [0.2,0.9] interval, chooses sample number at random in 4:1 ratio and build training sample set, test sample book collection.

4. according to claim 2 based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, establishment method, it is characterised in that, in described step 2,

The optimizing interval table of punishment factor c and Radial basis kernel function parameter g illustrated as lgc �� [2,3], lgg �� [0,1].

5. according to claim 2 based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, establishment method, it is characterised in that, in described step 2,

Adopt genetic algorithm to the interval optimizing further of the optimizing of punishment factor c and Radial basis kernel function parameter g, process is: setting genetic algorithm parameter, individual amount NIND=100, maximum genetic algebra MAXGEN=50, generation gap GGAP=0.95, crossover probability and variation probability be 0.8,0.25, c and g as variable, represent with 10 scale-of-two respectively; Objective function is model square error value MSE; Genetic operator operates: selects, intersect, make a variation; Calculate the slotting filial generation of laying equal stress on of filial generation target value function and obtain new population to parent; Reach the optimizing result that maximum iteration time exports c and g.

6. according to claim 2 based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, establishment method, it is characterised in that, in described step 2,

The model built is: $f\left(x\right)=w\mathrm{\Φ}\left(x\right)+b=\underset{i=1}{\overset{l}{\Σ}}(a_i-{a_i}^{*})\·\exp (-\frac{\left|\right|x-x_i\left|\right|}{2{\mathrm{\σ}}^2})+b,$ Wherein, input signal X'=(X₁'X₂'��X₅')^T, X₁'��X₂'��X₃'��X₄'��X₅' it is respectively temperature, CO₂Concentration, intensity of illumination, relative humidity and chlorophyll content, w is weight vector, and b is biased, and �� (x) is nonlinear mapping function.

7. according to claim 2 based on the omnidistance photosynthetic rate predictive model of cucumber of SVMs, establishment method, it is characterized in that, after obtaining the omnidistance photosynthetic rate predictive model of the cucumber based on SVMs, input amendment data carry out photosynthetic efficiency prediction, if reaching accuracy requirement, then export and predict the outcome, otherwise return genetic algorithmic steps, to c and g based on the interval optimizing further of described optimizing.