CN106444378A - Plant culture method and system based on IoT (Internet of things) big data analysis - Google Patents

Abstract

The present invention provides a plant cultivation method and system based on the big data analysis of the Internet of Things, wherein the method includes: collecting plant types, soil humidity, soil pH value, light intensity, ambient temperature, ambient humidity, images, watering amount , fertilization amount, and fertilization type constitute the influencing factor matrix X, and upload it to the server; among them, the watering amount, fertilization amount, and fertilization type constitute decision variables; the Elman neural network is used in the server to establish the plant influencing factor matrix X and plant health The complex nonlinear relationship between the indices is used to obtain the plant cultivation model; the NSGA‑Ⅱ algorithm is used to optimize the plant cultivation model to obtain a set of optimal solutions of decision variables; the optimal solution of the decision variables is used as the recommended decision for plants X ^* is sent to the user's terminal device by the server for display; the user cultivates plants according to the recommended decision displayed by the terminal device. Utilizing the invention can determine the optimal plant cultivation scheme, and create a better living environment for the plants.

Description

Plant cultivation method and system based on big data analysis of Internet of Things

technical field

The invention relates to the field of plant intelligent cultivation, in particular to a plant cultivation method and system based on big data analysis of the Internet of Things.

Background technique

With the rapid development of the national economy, potted plants have entered thousands of households as a way to increase living comfort. But because most plant owners lack experience in planting plants, the plants grow in a sub-healthy environment for a long time. On the other hand, due to limited indoor space, plant owners will require plants to have different degrees of denseness according to their own conditions, so as to avoid space waste.

At present, the problem that needs to be solved urgently is to establish a comprehensive plant cultivation model and feed back the plant health indicators to users, so that users can make timely adjustments to the plant cultivation plan. The various factors affecting the health of plants often reflect a high degree of complexity and nonlinearity, and it is difficult to use conventional prediction and analysis methods.

Contents of the invention

The present invention provides a plant cultivation method and system based on big data analysis of the Internet of Things to solve the problem in the prior art that the plant growth condition deviates from the expected index due to the inability to provide a suitable growth environment for the plant during the plant cultivation process.

In order to solve the above problems, the present invention adopts the following technical solutions to achieve:

On the one hand, the plant cultivation method based on the big data analysis of the Internet of Things provided by the present invention includes:

Step S1: Collect plant types, soil moisture, soil pH, light intensity, ambient temperature, ambient humidity, images, watering amount, fertilization amount, and fertilization type to form an influencing factor matrix X, and upload it to the server; among them, watering The amount of water, the amount of fertilization and the type of fertilization constitute the decision variables;

Step S2: use the Elman neural network in the server to establish a complex nonlinear relationship between the plant influencing factor matrix X and the plant health index, and obtain a plant cultivation model;

Step S3: using the NSGA-II algorithm to optimize the plant cultivation model to obtain a set of optimal solutions for the decision variables;

Step S4: The optimal solution of the group of decision variables is used as the recommended decision X ^* of the plant and sent to the user's terminal device through the server for display;

Step S5: The user cultivates plants according to the recommended decision displayed on the terminal device.

On the other hand, the plant cultivation system based on the big data analysis of the Internet of Things provided by the present invention includes:

The data acquisition unit is used to collect plant types, growth periods, soil humidity, soil pH, light intensity, ambient temperature, ambient humidity, images, watering amount, fertilization amount, and fertilization type to form an influencing factor matrix X, and upload to the server; wherein, the amount of watering, the amount of fertilization and the type of fertilization constitute decision variables;

The plant cultivation model building unit is used to use the Elman neural network in the server to establish the complex nonlinear relationship between the matrix X of the various influencing factors of the plant and the plant health index, and obtain the plant cultivation model;

The decision variable optimal solution acquisition unit is used to optimize the plant cultivation model by using the NSGA-II algorithm to obtain a group of optimal solutions of the decision variables, and use this group of optimal solutions of the decision variables as the recommended decision X ^* for the plant;

The recommended decision delivery unit is configured to deliver the recommended decision X ^* of the plant to the user's terminal device for display through the server.

Compared with the prior art, the advantages of the plant cultivation method and system based on the big data analysis of the Internet of Things provided by the present invention are: using the Elman neural network to establish a plant cultivation model, and then using the NSGA-II algorithm to optimize the plant cultivation model, and determine the plant cultivation model. The optimal value of the amount of watering, fertilization, and fertilization types, and instant feedback to the user, so that the user can understand the current status of the plant anytime, anywhere, and realize intelligent cultivation.

Description of drawings

Fig. 1 is the schematic flow chart of the plant cultivation method based on big data analysis of the Internet of Things according to an embodiment of the present invention;

Fig. 2 is a health index prediction result diagram according to an embodiment of the present invention;

Fig. 3 is a health index prediction error diagram according to an embodiment of the present invention;

Fig. 4 is a schematic diagram of a user interface according to an embodiment of the present invention.

detailed description

FIG. 1 shows the flow of a plant cultivation method based on big data analysis of the Internet of Things according to an embodiment of the present invention.

As shown in Figure 1, the plant cultivation method based on the big data analysis of the Internet of Things of the present invention comprises:

Step S1: Collect plant species, growth period, soil humidity, soil pH, light intensity, ambient temperature, ambient humidity, images, watering amount, fertilization amount, and fertilization type to form an influencing factor matrix X, and upload it to the server; Among them, watering amount, fertilization amount and fertilization type constitute decision variables.

Through statistics, the variables that have the greatest impact on the plant health index y ₁ are: plant species x ₁ , growth period x ₂ , soil moisture x ₃ , soil pH x ₄ , light intensity x ₅ , ambient temperature x ₆ , and ambient humidity x ₇ , image x ₈ , watering amount x ₉ , fertilization amount x ₁₀ , fertilization type x ₁₁ , a total of 11 variables; among them, soil moisture x ₃ , soil pH x ₄ , light intensity x ₅ , ambient temperature x ₆ , Environmental humidity x ₇ and image x ₈ are measured by corresponding sensors. Plant species and growth periods are inherent attributes, which are input by users. Watering amount, fertilization amount, and fertilization type are decision variables.

The environmental temperature of plants x ₆ is obtained by collecting temperature sensors; the soil humidity of plants x ₃ and environmental humidity x ₇ are obtained by collecting humidity sensors; the light intensity of plants x ₅ is obtained by collecting illuminance sensors; the soil pH of plants x ₄ is obtained by soil The pH meter is collected; the sampling circuit is connected to the temperature sensor, humidity sensor, illuminance sensor, and soil pH meter respectively, and the ambient temperature, ambient humidity, and soil pH meter are respectively collected by the temperature sensor, humidity sensor, illuminance sensor, and soil pH meter. Humidity, light intensity, and soil pH are converted into digital signals.

The characteristic image of the plant at the current moment is collected by the camera, and the camera converts the image information into a digital signal.

In the present invention, the server is preferably a cloud server.

Step S2: Using the Elman neural network in the server to establish a complex nonlinear relationship between the plant influencing factor matrix X and the plant health index to obtain a plant cultivation model.

Set X _k ＝[x _k1 ,x _k2 ,L,x _kM ](k=1,2,L,S) as the input vector, N as the number of training samples,

is the weight vector between the input layer M and the hidden layer I at the gth iteration, W _JP (g) is the weight vector between the hidden layer J and the output layer P at the gth iteration, W _JC (g) is the weight vector Y _k (g)=[y _k1 (g), y _k2 (g), L, y _kP (g)] (k=1 ,2,L,S) is the actual output of the network at the gth iteration, d _k =[d _k1 ,d _k2 ,L,d _kP ](k=1,2,L,S) is the expected output, the number of iterations g is 500.

In the server, the Elman neural network is used to establish the complex nonlinear relationship between the matrix X of the plant's various influencing factors and the plant health index, and obtain the plant cultivation model, including:

Step S21: Initialize, set the initial value of the number of iterations g to 0, and assign W _MI (0), W _JP (0), W _JC (0) a random value in the interval (0,1) respectively;

Step S22: Randomly input samples X _k ;

Step S23: For the input sample X _k , forwardly calculate the actual output Y _k (g) of neurons in each layer of the Elman neural network;

Step S24: Calculate the error E(g) according to the expected output d _k and the actual output Y _k (g);

Step S25: judge whether the error E(g) is less than the preset error value, if it is greater than or equal to it, go to step S26, if it is smaller, go to step S29;

Step S26: Determine whether the number of iterations g+1 is greater than the maximum number of iterations, if it is greater, go to step S29, otherwise, go to step S27;

Step S27: Reversely calculate the local gradient δ of each layer of neurons in the Elman neural network for the input sample X _k ;

Step S28: Calculate the weight correction amount ΔW, and correct the weight; let g=g+1, jump to step S23;

Among them, ΔW _ij =η·δ _ij , η is learning efficiency; W _ij (g+1)=W _ij (g)+ΔW _ij (g);

Step S29: Determine whether the training of all samples is completed; if yes, complete the modeling; if not, go to step S22.

In the design of Elman neural network, the number of nodes in the hidden layer is the key to determine the quality of the Elman neural network model, and it is also the difficulty in the design of the Elman neural network. Here, the trial and error method is used to determine the number of nodes in the hidden layer.

In the formula, p is the number of neuron nodes in the hidden layer, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and k is a constant between 1-10. The setting parameters of the Elman neural network are shown in Table 2 below.

Table 2 Elman neural network setting parameters

objective function health index iterations 500 hidden layer transfer function Tansig output layer transfer function Purelin hidden layer nodes 15

Through the above process, the prediction effect of Elman neural network can be obtained as shown in Figures 2 and 3. The basis of intelligent plant cultivation is the establishment of models, and the accuracy of the models directly affects the output results. Through the analysis of Figures 2 and 3, it can be seen that the maximum prediction error of the health index is -3.5%, and the prediction accuracy of the model is high, which meets the modeling requirements.

Step S3: using the NSGA-II algorithm (Non-dominated Sorting Genetic Algorithm-II, non-dominated sorting genetic algorithm with elitist strategy) to optimize the plant breeding model to obtain a set of optimal solutions for decision variables.

Obtain a set of optimal solutions for decision variables, that is, obtain a set of optimal values for plant watering, fertilization, and fertilization types.

The steps of optimizing the plant cultivation model using the NSGA-II algorithm include:

Step S31: Initialize system parameters; wherein, the system parameters include population size N, maximum genetic algebra G, crossover probability P, and mutation probability Q.

Step S32: Merge the new population Q _{t generated in generation t} with its parent population P _t to form population R _t . The size of population R _t is 2N; if it is the first generation population, use the first generation population as population R _t .

Step S33: Perform non-dominated sorting on the population R _t to obtain a series of non-dominated sets Z _i , and calculate the crowding degree of each individual in the non-dominated set Z _i to generate a new parent population P _t+1 .

The concrete process of step S33 is as follows:

Step S331: Use the fitness function to judge the mutual domination relationship among all individuals in the population R _t ; where, D(i).n represents the number of individuals dominating the i-th individual, and D(i).p represents the number of individuals dominated by the i-th individual A set of individuals dominated by individuals; if individual i dominates j, put individual j into the D(i).p set, and add 1 to the value of D(j).n; operate sequentially to obtain all individuals D in the population R _t (i).n and D(i).p information.

Step S332: Put all the individuals whose D(i).n value is 0 in the population R _t , that is, the individuals of this type are not dominated by other individuals, put them into the first layer of the non-dominated layer, and set the D(i).n value to 1 into the second layer of the non-dominated layer, and operate sequentially until all individuals in the population R _t are put into different non-dominated layers; individuals in the same layer share the same virtual fitness value, and the level The smaller the value is, the lower the virtual fitness value is, and the better the individual in this layer is, the layers of the non-dominated layer are sorted in ascending order.

Step S333: Since all individuals in each layer share the same virtual fitness value, when it is necessary to select a better individual in the same layer, calculate its degree of congestion.

The initial value of the crowding degree i _d of each point is set to 0; for each target, the population R _t is non-dominated sorted, and the crowding degree of the two individuals at the boundary of the population R _t is infinite, and the Other individuals in the population R _t calculate the degree of crowding:

Among them, _id represents the congestion degree of point i, Indicates the jth objective function value of point i+1, Indicates the jth objective function value of point i-1.

Step S334: After fast non-dominated sorting and congestion degree calculation, each individual i in population R _t has two attributes: non-dominated rank i _rank determined by non-dominated sorting and congestion degree _id . According to these two attributes, a crowding degree comparison operator can be defined: individual i is compared with individual j, if the non-dominated rank of individual i is better than the non-dominated rank of individual j, that is, i _rank < j _rank , or , individual i has the same rank as individual j, and individual i has a longer crowding distance than individual j, that is, i _rank =j _rank and i _d >j _d , then individual i wins.

Step S335: Since the individuals of the child population and the individuals of the parent population P _t+1 are included in the population R _t , the individuals contained in the non-dominated set Z ₁ after non-dominated sorting are the best in R _t , so first put the non-dominated set Z ₁ into the parent population P _t+1 ; if the number of individuals in the parent population P _t+1 does not exceed the population size N, then put the next-level non-dominated set Z ₂ into the parent population Generation population P _t+1 , until the non-dominated set Z ₃ is put into the parent population P _t+1 , the number of individuals in the parent population P _t+1 exceeds the population size N, and the individuals in the non-dominated set Z ₃ are used Comparing with the crowding degree comparison operator, take the first {num(Z ₃ )-(num(P _t+1 )-N)} individuals, so that the number of individuals in the parent population P _t+1 reaches the population size N.

Step S34: Perform crossover and mutation basic genetic operations on the parent population P _t+1 to obtain the offspring population Q _t+1 .

The process of crossover genetic operation on the parent population _Pt+1 is:

Randomly match all individuals in the parent population P _t+1 into pairs, and generate a random number for each pair of individuals. If the random number of a pair of individuals is less than the crossover probability P, exchange the part between the pair of individuals chromosome.

The basic genetic operation process of mutation on the parent population P _t+1 is as follows:

For each individual in the parent population P _t+1 , generate a random number, if the random number of an individual is less than the mutation probability Q, then change the gene value of one or some loci of the individual to other genes value.

Step S35: Add 1 to the genetic algebra, determine whether the genetic algebra reaches the maximum genetic algebra G, if yes, output the current global optimal solution; if not, jump to step S32 to repeat the calculation until the genetic algebra reaches the maximum genetic algebra G.

Step S4: The optimal solution of the group of decision variables is sent to the user's terminal device for display through the server as the recommended decision X ^* of the plant.

Various sensors collect data every 2 hours and upload it to the server. The server receives the data and gives the current recommended watering amount, fertilization amount and fertilization type for the plant through the plant cultivation model.

Step S5: The user cultivates plants according to the recommended decision displayed on the terminal device.

The user can open the intelligent plant cultivation interface (as shown in Figure 4) on the terminal device, and the interface displays the brief information of the plant. The brief information of the plant includes the image of the plant and the current health index. The user can set the ideal health index of the plant on the interface , Ideal, the server sends the recommended watering amount, fertilization amount, and fertilization type, and the user can complete automatic watering and fertilization through remote operations on the mobile phone.

The current health index of the plant is obtained by optimizing the plant breeding model based on the NSGA-Ⅱ algorithm, and the current health index of the plant corresponds to a set of optimal solutions of the decision variables.

The plant cultivation method based on the big data analysis of the Internet of Things provided by the present invention first uses hardware such as sensors and cameras to collect plant index parameters, plant images, watering amount, fertilization amount, and fertilization type, and then uploads the collected data to The server stores it, uses the Elman neural network to establish the complex nonlinear relationship between the influencing factor matrix X and the plant health index in the server, obtains the plant cultivation model, uses the NSGA-Ⅱ algorithm to optimize the plant cultivation model, and obtains the decision variables. A set of optimal values, and send this set of optimal solutions to the user's PC or APP terminal as a recommended decision. Finally, the user can decide the watering amount, fertilization amount, and fertilization type of the plant according to the recommended decision to realize remote automatic cultivation . The method can determine the optimal plant cultivation scheme and create a better living environment for the plants.

Corresponding to the above method, the present invention also provides a plant cultivation system based on big data analysis of the Internet of Things.

The plant cultivation system based on the big data analysis of the Internet of Things provided by the present invention includes:

The data acquisition unit is used to collect plant types, growth periods, soil humidity, soil pH, light intensity, ambient temperature, ambient humidity, images, watering amount, fertilization amount, and fertilization type to form an influencing factor matrix X, and upload to the server; where watering amount, fertilization amount, and fertilization type constitute the decision variables. For the process of collecting data by the data collection unit, refer to the above step S1.

The plant cultivation model building unit is used to establish the complex nonlinear relationship between the plant influencing factor matrix X and the plant health index by using the Elman neural network in the server to obtain the plant cultivation model. For the specific process of establishing the plant cultivation model by the plant cultivation model building unit, refer to the above step S2.

The decision variable optimal solution acquisition unit is used to optimize the plant cultivation model by using the NSGA-II algorithm, obtain a group of optimal solutions of the decision variables, and use the group of optimal solutions of the decision variables as the recommended decision X ^* for the plant. For the specific process of obtaining the optimal solution of the decision variable by the unit for obtaining the optimal solution of the decision variable, refer to the above step S3.

The recommended decision delivery unit is configured to deliver the recommended decision X ^* of the plant to the user's terminal device for display through the server.

The user cultivates the plants according to the recommended decision displayed on the terminal device.

It should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the above-mentioned examples. Those skilled in the art may make changes, modifications, additions or replacements within the scope of the present invention. It should also belong to the protection scope of the present invention.

Claims (8)

1. a kind of based on Internet of Things big data analysis plant cultivating method, it is characterised in that comprise the steps：

Step S1：The species of herborization, growth period, soil moisture, soil pH value, intensity of illumination, ambient temperature, environmental wet Degree, image, irrigation amount, dose, Fertilizer Type simultaneously constitute influence factor matrix X, and upload onto the server；Wherein, described pour The water yield, the dose and the Fertilizer Type constitute decision variable；

Step S2：Referred to plant health using each influence factor's matrix X of Elman neural network plant in the server Complex nonlinear relation between number, obtains plant cultivation model；

Step S3：Using II algorithm of NSGA-, the plant cultivation model is optimized, obtains one group of the decision variable most Excellent solution；

Step S4：Using the group optimal solution of the decision variable as the plant recommendation decision-making X^*By under the server The terminal unit for being sent to user is shown；

Step S5：The user cultivates the plant according to the recommendation decision-making that the terminal unit shows.

2. according to claim 1 based on Internet of Things big data analysis plant cultivating method, it is characterised in that the plant Thing cultivates X in model_k=[x_k1,x_k2,L,x_kM] (k=1,2, L, S) be input vector, S for training sample number, W_MI(g) For weighted vector during the g time iteration between input layer M and hidden layer I, W_JP(g) be the g time iteration when hidden layer J and output layer P it Between weighted vector, W_JCG () is hidden layer J and the weighted vector that accepts between layer C, Y during the g time iteration_k(g)=[y_k1(g),y_k2 (g),L,y_kP(g)] (k=1,2, L, S) be reality output during the g time iteration, d_k=[d_k1,d_k2,L,d_kP] (k=1,2, L, S) it is desired output；And,

The step of setting up the plant cultivation model includes：

Step S21：Initialization, if iterationses g initial value is 0, to be assigned to W respectively_MI(0)、W_JP(0)、W_JC(0) one (0,1) interval Random value；

Step S22：Stochastic inputs sample X_k；

Step S23：To input sample X_k, reality output Y of per layer of neuron of Elman neutral net described in forward calculation_k(g)；

Step S24：According to desired output d_kWith reality output Y_k(g), calculation error E (g)；

Step S25：Whether error in judgement E (g) is less than default error amount, if greater than or be equal to, enter step S26, if It is less than, then enters step S29；

Step S26：Judge that iterationses g+1, whether more than maximum iteration time, if it does, step S29 is entered, otherwise, enters Enter step S27；

Step S27：To input sample X_kThe partial gradient δ of per layer of neuron of Elman neutral net described in backwards calculation；

Step S28：Modified weight amount Δ W is calculated, and revises weights；G=g+1 is made, jumps to step S23；

Wherein, Δ W_ij=η δ_ij, η is the learning efficiency；W_ij(g+1)=W_ij(g)+ΔW_ij(g)；

Step S29：Judge whether to complete the training of all samples；If it is, completing modeling；If not, jumping to step S22.

3. according to claim 1 based on Internet of Things big data analysis plant cultivating method, it is characterised in that utilize The step of II algorithm of NSGA- is optimized to the plant cultivation model, including：

Step S31：Initialization system parameter；Wherein, the systematic parameter includes population scale N, maximum genetic algebra G, intersection Probability P and mutation probability Q；

Step S32：The new population Q that produce t generation_tWith its parent population P_tMerge composition population R_t, population R_tSize be 2N； If first generation population, then using first generation population as the population R_t；

Step S33：To the population R_tNon-dominated ranking is carried out, obtains a series of non-dominant collection Z_i, and calculate the non-dominant Collection Z_iIn each individual crowding, produce new parent population P_t+1；

Step S34：To the parent population P_t+1Carry out intersecting, the basic genetic that makes a variation operation obtains progeny population Q_t+1；

Step S35：Genetic algebra adds 1, judges whether genetic algebra reaches the maximum genetic algebra G, if it is, output is current Globally optimal solution；If not, jump to step S32 double counting is carried out, until genetic algebra reaches the maximum genetic algebra G Till.

4. according to claim 3 based on Internet of Things big data analysis plant cultivating method, it is characterised in that step S33 includes：

Step S331：The population R is judged using fitness function_tIn all individuality between mutual dominance relation；Wherein, D I () .n represents the individual amount of i-th individuality of domination, D (i) .p represents by the individual collections of i-th individual domination；If individuality i Domination j, then be put into D (i) .p set by individual j, and the value of D (j) .n adds 1；Operate successively, obtain all individuality D (i) .n and D The information of (i) .p；

Step S332：By the population R_tIn all D (i) .n values for 0 individuality, be put into the ground floor of non-dominant layer, by the kind Group R_tIn all D (i) .n values be put into the second layer of non-dominant layer for 1 individuality, until by the population R_tIn all individualities be put into Till different non-dominant layers, the number of plies of non-dominant layer is ranked up by order from small to large；

Step S333：For each target, to the population R_tNon-dominated ranking is carried out, makes the population R_tTwo of border The crowding of body is infinite, to the population R_tMiddle others individuality carries out the calculating of crowding：

$i_d=\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left(\right|{f_j}^{i+1}-{f_j}^{i-1}\left|\right)$

Wherein, i_dRepresent the crowding of i point,Represent j-th target function value of i+1 point,Represent j-th of i-1 point Target function value；

Step S334：Non-dominant sequence i that non-dominated ranking is determined_rankWith crowding i_dAs the population R_tIn per each and every one Two attributes of body i, define crowding comparison operator：Individual i is compared with individual j, if the non-dominant layer residing for individuality i Better than non-dominant layer, i.e. i residing for individuality j_rank＜ j_rank, or individuality i and individual j has an identical grade, and individuality i than The crowding distance of body j is long, i.e. i_rank=j_rankAnd i_d＞ j_d, then individuality i triumph；

Step S335：By non-dominant collection Z₁It is put into the parent population P_t+1；If the parent population P_t+1Individual amount do not surpass Go out the population scale N, then by the non-dominant collection Z of next stage₂It is put into the parent population P_t+1, until by non-dominant collection Z₃It is put into The parent population P_t+1When, the parent population P_t+1Individual amount exceed the population scale N, to the non-dominant collection Z₃ In individuality be compared using the crowding comparison operator, { num (Z before taking₃)-(num(P_t+1)-N) individuality, make institute State parent population P_t+1Individual amount reach the population scale N.

5. according to claim 3 based on Internet of Things big data analysis plant cultivating method, it is characterised in that to described Parent population P_t+1The process for carrying out criss-cross inheritance operation is：

By the parent population P_t+1It is right that interior all individualities are mixed at random, to individual per a pair, generates a random number, if The random number of certain a pair of individuality is less than the crossover probability P, then exchange this to the chromosome dyad between individuality.

6. according to claim 3 based on Internet of Things big data analysis plant cultivating method, it is characterised in that to described Parent population P_t+1Enter row variation basic genetic operation process be：

To the parent population P_t+1In each is individual, generate a random number, if certain individual random number is less than described Mutation probability Q, then it is other genic values to change the genic value on some or certain some locus of the individuality.

7. according to any one of claim 1-6 based on Internet of Things big data analysis plant cultivating method, its feature It is,

The ambient temperature of the plant is gathered using temperature sensor；

Ambient humidity and the soil moisture of the plant are gathered using humidity sensor；

The intensity of illumination of the plant is gathered using illuminance sensor；

P in soil H-number using the above-mentioned plant of P in soil H meter collection；

Feature of the plant in current time is gathered using photographic head, and convert image information into digital signal；And,

Using sample circuit respectively with the temperature sensor, the humidity sensor, the illuminance sensor, the soil PH meter is attached, and the temperature sensor, the humidity sensor, the illuminance sensor, P in soil H are counted The ambient temperature that collects respectively, ambient humidity, soil moisture, intensity of illumination, P in soil H-number are converted into digital signal.

8. a kind of based on Internet of Things big data analysis plant cultivation system, including：

Data acquisition unit, for the species of herborization, growth period, soil moisture, soil pH value, intensity of illumination, environment Temperature, ambient humidity, image, irrigation amount, dose, Fertilizer Type simultaneously constitute influence factor matrix X, and upload onto the server； Wherein, the irrigation amount, the dose and the Fertilizer Type constitute decision variable；

Unit set up by plant cultivation model, in the server using Elman neural network plant respectively affect because Complex nonlinear relation between prime matrix X and plant health index, obtains plant cultivation model；

Decision variable optimal solution acquiring unit, for being optimized to the plant cultivation model using II algorithm of NSGA-, is obtained One group of optimal solution of the decision variable, and using the group optimal solution of the decision variable as the plant recommendation decision-making X^*；

Recommend decision-making issuance unit, for by the server by the recommendation decision-making X of the plant^*It is issued to the terminal of user Equipment is shown.