Method for predicting growth height of plant

Abstract

The invention discloses a method for predicting the growth height of a plant. The method comprises the steps that environment characteristic data of a planting environment where a to-be-predicted plant is located is acquired, and the environment characteristic data is used for representing influence factors, used for influencing the growth height of the to-be-predicted plant, in the planting environment; a historical growth data set of at least one grown plant in the planting environment is obtained, at least one historical growth data set is obtained, the at least one grown plant comprises a to-be-predicted plant, and the historical growth data set is used for representing the historical growth height of the grown plant; generating an initial predicted growth height of the to-be-predicted plant based on the environment characteristic data and the historical growth data set; the at least one initial predicted growth height is converted, a target predicted growth height is obtained, and the accuracy of the target predicted growth height is higher than that of the initial predicted growth height. The technical problem that the prediction accuracy of the plant growth height is low is solved.

Description

Technical Field

[0001] The invention relates to the field of power systems, and in particular to a method for predicting plant growth height. Background Art

[0002] At present, predicting the growth height of plants can effectively determine the time point when plants may pose a hidden danger to transmission lines, thereby improving the efficiency of transmission line operation and maintenance inspections and reducing the maintenance costs of transmission lines.

[0003] In the related art, the growth height of a plant is usually predicted by obtaining a tree growth prediction model through a large amount of training data, and the height of the plant is predicted using the tree growth prediction model. However, the model trained by this method is only for prediction of certain specific types of plants. In actual use, there is a technical problem of low prediction accuracy of plant growth height.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiment of the present invention provides a method for predicting the growth height of a plant, so as to at least solve the technical problem of low prediction accuracy of the growth height of a plant.

[0006] According to one aspect of an embodiment of the present invention, a method for predicting the growth height of a plant is provided, and the method may include: obtaining environmental characteristic data of a planting environment in which a plant to be predicted is located, wherein the environmental characteristic data is used to characterize influencing factors in the planting environment that affect the growth height of the plant to be predicted; respectively obtaining historical growth data sets of at least one grown plant in the planting environment to obtain at least one historical growth data set, wherein at least one grown plant includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the grown plant; generating an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set; converting at least one initial predicted growth height to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height.

[0007] Optionally, based on the environmental characteristic data and the historical growth data set, an initial predicted growth height of the plant to be predicted is generated, including: mapping the environmental characteristic data into a latent space to obtain a first latent variable; based on the first latent variable and the historical growth data set, obtaining a latent variable distribution corresponding to the historical growth data set; and predicting at least one latent variable distribution corresponding to at least one historical growth data set to obtain at least one initial predicted growth height.

[0008] Optionally, based on the first latent variable and the historical growth data set, the latent variable distribution corresponding to the historical growth data set is obtained, including: controlling the distribution sampling network, transforming the first latent variable and the historical growth data set to obtain a first variance of the historical growth data set and a first mean of the historical growth data set; based on the first variance and the first mean, determining the latent variable distribution.

[0009] Optionally, at least one latent variable distribution corresponding to at least one historical growth data set is predicted respectively to obtain at least one initial predicted growth height, including: sampling the latent variable distribution to obtain multiple second latent variables, wherein the second latent variables are used to characterize the growth of the plant to be predicted in the planting environment; predicting the multiple second latent variables respectively to obtain multiple predicted growth heights; and determining the average of the multiple predicted growth heights as the initial predicted growth height.

[0010] Optionally, the method may also include: determining a target planting environment whose environmental similarity with the planting environment is greater than a similarity threshold; constructing first training data of the planting environment and second training data of the target planting environment, wherein the first training data at least includes a first environmental characteristic data sample of the planting environment and a first growth data sample of the plants grown in the planting environment; the second training data at least includes a second environmental characteristic data sample of the target planting environment and a second growth data sample of the plants grown in the target planting environment; and using the first training data and the second training data to train a distributed sampling network.

[0011] Optionally, the first training data and the second training data are used to train a distribution sampling network, including: inputting the first environmental feature data sample and the second environmental feature data sample into the first sub-distribution sampling network for analysis, respectively, to obtain the third latent variable of the first environmental feature data sample mapped to the latent space, and the fourth latent variable of the second environmental feature data sample mapped to the latent space; inputting the third latent variable, the first growth data sample, the fourth latent variable and the second growth data sample into the second sub-distribution sampling network for analysis, to obtain the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment; based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, construct a first sub-target equation; based on the first sub-target equation, determine the target model parameters of the sub-distribution sampling network, wherein the sub-distribution sampling network includes the first sub-distribution sampling network and the second sub-distribution sampling network; using the target model parameters, update the initial model parameters of the sub-distribution sampling network to obtain the distribution sampling network.

[0012] Optionally, based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, a first sub-goal equation is constructed, including: sampling the latent variable distribution of the planting environment to obtain multiple fifth latent variables; sampling the latent variable distribution of the target planting environment to obtain multiple sixth latent variables; based on multiple fifth latent variables and multiple sixth latent variables, constructing the first sub-goal equation.

[0013] Optionally, multiple second latent variables are predicted respectively to obtain multiple predicted growth heights, including: controlling the prediction model to predict the second latent variable to obtain the predicted growth height, wherein the prediction model is trained based on the first training data and the second training data.

[0014] Optionally, the method may also include: mapping the first environmental feature data sample to the latent space to obtain multiple seventh latent variables; controlling the sub-prediction model of the target planting environment to predict multiple sixth latent variables corresponding to the target planting environment to obtain a first predicted growth height, and controlling the sub-prediction model of the planting environment to predict the seventh latent variable to obtain a second predicted growth height; using the first predicted growth height and the second predicted growth height to construct a second sub-target equation; based on the second sub-target equation, determining the target model parameters of the sub-prediction model of the target planting environment, and the target model parameters of the sub-prediction model of the planting environment; using the target model parameters, respectively updating the initial model parameters of the sub-prediction model of the target planting environment, and the initial model parameters of the sub-prediction model of the planting environment, to obtain a prediction model, wherein the prediction model includes an adjusted sub-prediction model of the target planting environment, and an adjusted sub-prediction model of the planting environment.

[0015] Optionally, the method may also include: controlling the sub-reconstruction model corresponding to the target planting environment to reconstruct the sixth latent variable to obtain a first reconstruction result, and controlling the sub-reconstruction model corresponding to the planting environment to reconstruct the seventh latent variable to obtain a second reconstruction result; based on the first reconstruction result and the second reconstruction result, constructing a third sub-target equation.

[0016] Optionally, based on the first sub-goal equation, target model parameters of the sub-distribution sampling network are determined, including: constructing a target equation based on the first sub-goal equation, the second sub-goal equation and the third sub-goal equation; and using the target equation to determine the target model parameters of the sub-distribution sampling network.

[0017] Optionally, based on the second sub-target equation, target model parameters of the sub-prediction model of the target planting environment, and target model parameters of the sub-prediction model of the planting environment are determined, including: constructing a target equation based on the first sub-target equation, the second sub-target equation and the third sub-target equation; using the target equation to determine the target model parameters of the sub-prediction model of the target planting environment, and target model parameters of the sub-prediction model of the planting environment.

[0018] Optionally, the method may further include: determining, based on the target predicted growth height, a time at which the plant to be predicted will have an impact on a power transmission line in the planting environment.

[0019] According to another aspect of an embodiment of the present invention, a device for predicting the growth height of a plant is also provided, and the device may include: a first acquisition unit, used to acquire environmental characteristic data of a planting environment in which a plant to be predicted is located, wherein the environmental characteristic data is used to characterize influencing factors in the planting environment that affect the growth height of the plant to be predicted; a second acquisition unit, used to respectively acquire historical growth data sets of at least one grown plant in the planting environment to obtain at least one historical growth data set, wherein at least one grown plant includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the grown plant; a generation unit, used to generate an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set; a conversion unit, used to convert at least one initial predicted growth height to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height.

[0020] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the plant growth height prediction method of the embodiment of the present invention.

[0021] According to another aspect of an embodiment of the present invention, a processor is provided. The processor is used to run a program. When the program is run by the processor, the method for predicting the growth height of a plant according to an embodiment of the present invention is executed.

[0022] According to another aspect of an embodiment of the present invention, a program product is further provided. The program product includes computer instructions. When the computer instructions are executed by a processor, the method for predicting the growth height of a plant according to an embodiment of the present invention is implemented.

[0023] According to another aspect of the embodiments of the present invention, a vehicle is provided. The vehicle can be used to execute the method for predicting the growth height of a plant according to the embodiments of the present invention.

[0024] In an embodiment of the present invention, environmental characteristic data of the planting environment where the plant to be predicted is located is obtained, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted; historical growth data sets of at least one plant that has grown in the planting environment are respectively obtained to obtain at least one historical growth data set, wherein at least one plant that has grown includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the plant that has grown; based on the environmental characteristic data and the historical growth data set, an initial predicted growth height of the plant to be predicted is generated; at least one initial predicted growth height is converted to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height. That is, in an embodiment of the present invention, at least one historical growth data set and environmental characteristic data are used to obtain at least one initial predicted growth height corresponding to at least one historical growth data set, the initial predicted growth height and the historical growth data set are in one-to-one correspondence, and the obtained at least one initial predicted growth height is converted to obtain a target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 is a flow chart of a method for predicting the growth height of a plant according to an embodiment of the present invention;

[0027] Figure 2 is a schematic diagram of a latent variable extraction module according to an embodiment of the present invention;

[0028] Figure 3 is a schematic diagram of a discriminator module according to an embodiment of the present invention;

[0029] Figure 4 is a schematic diagram of a prediction model module according to an embodiment of the present invention;

[0030] Figure 5 is a schematic diagram of a reconstruction model module according to an embodiment of the present invention;

[0031] Figure 6 is a schematic diagram of a plant growth height prediction device according to an embodiment of the present invention;

[0032] Figure 7 It is a block diagram of an electronic device for predicting the growth height of a plant according to an embodiment of the present application. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] According to an embodiment of the present invention, an embodiment of a method for predicting the growth height of a plant is provided. The steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] In this embodiment, a method for predicting the growth height of a plant is proposed. The method can utilize at least one historical growth data set and environmental characteristic data to obtain at least one initial predicted growth height corresponding to at least one historical growth data set. The initial predicted growth height corresponds to the historical growth data set one-to-one. The at least one initial predicted growth height obtained is converted to obtain a target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0037] Figure 1 FIG. 1 is a flow chart of a method for predicting the growth height of a plant according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0038] Step S102, obtaining environmental characteristic data of the planting environment where the plant to be predicted is located, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted.

[0039] In the technical solution provided in the above step S102 of the present invention, the above-mentioned plants to be predicted can be vegetables, trees and other plants that will be affected by the planting environment, such as bamboo, fir, pine, eucalyptus, corn, etc. It should be noted that this is only an example, and there is no specific limitation on the type of plants to be predicted. The planting environment can be used to characterize the target domain for planting the plants to be predicted, and can include the target area. The environmental characteristic data can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the influencing factors in the planting environment that affect the growth height of the plants to be predicted, for example, it can include meteorological data such as temperature, humidity, wind speed, etc., and can also include the longitude and latitude information of the planting environment, the planting density of the planting environment and other data, which can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the influencing factors in the planting environment that affect the growth height of the plants to be predicted, for example, it can include meteorological data such as temperature, humidity, wind speed, etc., and can also include the latitude and longitude information of the planting environment, the planting density of the planting environment and other data, which can be used to characterize the environmental characteristics at multiple times, and ... environmental characteristics at multiple times, and can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the environmental characteristics at multiple times, and can be used to characterize the environmental * It should be noted that this is only an example and there is no specific restriction on the content and form of expression of the environmental characteristic data.

[0040] Optionally, the environmental characteristics of the planting environment can be collected by information collection equipment such as a hygrometer and a thermometer to obtain environmental characteristic data. The environmental characteristic data can be used to characterize the influencing factors such as temperature and humidity that affect the growth height of the plant to be predicted over a period of time. It should be noted that this is only an example, and there is no specific limitation on the method of collecting environmental characteristic data.

[0041] Optionally, the planting environment may be an area where a power transmission line is located, or may be a target domain. The plants to be predicted may be plants growing in the planting environment. When the plants to be predicted grow to a certain height, they may have an impact on the power transmission line. Therefore, in this embodiment, the environmental characteristic data of the planting environment may be used to determine the height to which the plants to be predicted may grow under the planting environment. Therefore, the environmental characteristic data of the planting environment where the plants to be predicted are located is obtained, and the environmental characteristic data is used to predict the growth height of the plants to be predicted.

[0042] Step S104, respectively acquiring a historical growth data set of at least one plant grown in the planting environment to obtain at least one historical growth data set, wherein the at least one plant grown includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the plant grown.

[0043] In the technical solution provided in step S104 of the present invention, the historical growth data set can be used to determine the historical growth height of the plant that has grown. It can be used to characterize the growth height of a plant at different times in history. It can be used to characterize the number of days recorded. The grown plants can be plants that have grown in the planting environment, and can include plants to be predicted.

[0044] Optionally, the historical growth conditions of at least one plant grown in the planting environment may be collected in advance to obtain a historical growth data set of the plant grown over a period of time.

[0045] For example, we can choose n t The target value of the training samples in the target domain in

[0046] Step S106, generating an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set.

[0047] In the technical solution provided in the above step S106 of the present invention, the environmental characteristic data and the historical growth data set are processed to generate the initial predicted growth height of the plant to be predicted.

[0048] For example, the environmental characteristic data and the historical growth data set are input into the prediction model to obtain the initial predicted growth height. It should be noted that the method for obtaining the initial predicted growth height here is only for illustration and is not specifically limited here.

[0049] Step S108, converting at least one initial predicted growth height to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height.

[0050] In the technical solution provided in step S108 of the present invention, an initial predicted growth height determined based on a historical growth data set is obtained to obtain at least one initial predicted growth height corresponding to at least one historical growth data set. The at least one initial predicted growth height is converted to obtain a target predicted growth height. The accuracy of the target predicted growth height obtained at this time is higher than the accuracy of the initial predicted growth height.

[0051] Optionally, select the nth t Historical growth dataset of target domain in New environmental feature data x * and nth t indivual Processing is performed to obtain an initial predicted growth height. At least one initial predicted growth height can be obtained by the above method. The at least one initial predicted growth height can be converted according to the following formula to obtain a target predicted growth height:

[0052]

[0053] in, Can be used to characterize the initial predicted growth height. * It can be used to characterize the predicted growth height of the target.

[0054] Through the above-mentioned step S102 and step S108 of the present invention, the environmental characteristic data of the planting environment where the plant to be predicted is located is obtained, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted; the historical growth data set of at least one plant that has grown in the planting environment is obtained respectively to obtain at least one historical growth data set, wherein at least one plant that has grown includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the plant that has grown; based on the environmental characteristic data and the historical growth data set, the initial predicted growth height of the plant to be predicted is generated; the at least one initial predicted growth height is converted to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height. That is to say, in the embodiment of the present invention, at least one historical growth data set and environmental characteristic data are used to obtain at least one initial predicted growth height corresponding to at least one historical growth data set, the initial predicted growth height and the historical growth data set are in one-to-one correspondence, and the obtained at least one initial predicted growth height is converted to obtain the target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0055] The above method of this embodiment is further introduced below.

[0056] As an optional implementation, step S106 generates an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set, including: mapping the environmental characteristic data into a latent space to obtain a first latent variable; based on the first latent variable and the historical growth data set, obtaining a latent variable distribution corresponding to the historical growth data set; and predicting at least one latent variable distribution corresponding to at least one historical growth data set to obtain at least one initial predicted growth height.

[0057] In this embodiment, the environmental characteristic data can be mapped into the latent space to obtain a first latent variable, and the latent variable distribution corresponding to the historical growth data set can be obtained based on the first latent variable and the historical growth data set; at least one latent variable distribution is predicted respectively to obtain at least one initial predicted growth height. The latent variable distribution can be expressed as It indicates that, Can be used to represent the mean, It can be used to represent variance. The first latent variable can be z x express.

[0058] Optionally, the environmental characteristic data is mapped to a first latent variable, the first latent variable and the historical growth data set can be sampled, the latent variable distribution corresponding to the historical growth data set can be obtained, and the latent variable distribution can be processed to predict an initial predicted growth height.

[0059] As an optional implementation, based on the first latent variable and the historical growth data set, the latent variable distribution corresponding to the historical growth data set is obtained, including: controlling the distribution sampling network, transforming the first latent variable and the historical growth data set, obtaining the first variance of the historical growth data set, and the first mean of the historical growth data set; based on the first variance and the first mean, determining the latent variable distribution.

[0060] In this embodiment, the above-mentioned distribution sampling network can be a latent variable extraction module, which can include a joint distribution sampling network and an independent distribution sampling network. The first latent variable and the historical growth data set can be input into the distribution sampling network to obtain the first variance and the first mean output by the distribution sampling network. The latent variable distribution corresponding to the historical growth data set can be determined based on the first variance and the first mean. It should be noted that in this embodiment, there is no specific restriction on the method of determining the latent variable. Except for the Gaussian distribution, any method of determining the latent variable distribution by the first variance and the first mean should be within the scope of protection of this application.

[0061] Optionally, the joint distribution sampling network in the trained distribution sampling network is retrieved, and the first latent variable and the historical growth data set are input into the joint distribution sampling network to obtain the first variance and the first mean, and then based on the first variance and the first mean, the latent variable distribution belonging to the target domain can be obtained. One historical growth data set corresponds to one latent variable distribution.

[0062] As an optional implementation, at least one latent variable distribution corresponding to at least one historical growth data set is predicted to obtain at least one initial predicted growth height, including: sampling the latent variable distribution to obtain multiple second latent variables, wherein the second latent variables are used to characterize the growth of the plant to be predicted in the planting environment; predicting the multiple second latent variables to obtain multiple predicted growth heights; and determining the average of the multiple predicted growth heights as the initial predicted growth height.

[0063] In this embodiment, the independent distribution sampling network in the distribution sampling network can be controlled to sample the latent variable distribution to obtain multiple second latent variables, or after obtaining the latent variable distribution, multiple second latent variables can be obtained based on the Bayesian rule and importance sampling, for example, S second latent variables can be sampled, and the multiple second latent variables can be represented by z'. Based on the second latent variable, the initial predicted growth height can be predicted, wherein the number of the initial predicted growth height is the same as the number of the second latent variables.

[0064] Optionally, select the nth t Historical growth dataset of target domain in The first latent variable z can be x and nth t indivual Input to the joint distribution sampling network C2 to get the mean and variance in, It can be used to represent the model parameters of C2, and then the latent variable distribution belonging to the target domain can be obtained Based on the Bayesian rule and using importance sampling, S second latent variables can be obtained: Further prediction models can be obtained Based on all the predicted values ​​of the S latent variables z′, multiple predicted growth heights are obtained, and the mean of the sum of the multiple predicted growth heights is determined, that is, This mean value can be determined as the initial predicted growth height. Based on all target values The mean of the sum of all predicted values, and the final target predicted growth height y is obtained based on the following formula * :

[0065]

[0066] For example, suppose the historical growth dataset in the target domain is Among them, the historical growth data set is Y t ={2,3,5,6,8}, where When n t =2, the corresponding Multiple second latent variables z can be x and Input to the joint distribution sampling network C2 to get the mean and variance Then we get the latent variable distribution belonging to the target domain Based on the Bayesian rule and using importance sampling, S latent variables are obtained Leverage the predictive model of the target domain Based on What you get Predict each latent variable and get the prediction result All prediction results can be summed up to get Through the above steps, we can obtain Based on all target values The mean of the sum of all predicted values, and the final predicted value y is obtained based on the following formula * ;

[0067]

[0068] It should be noted that the number and size of the above-mentioned figures are only for illustration and can be selected according to actual conditions. No specific limitation is made here.

[0069] In this embodiment, environmental characteristic data may be acquired, and the environmental characteristic data and the historical growth data set may be processed using a trained prediction model and a distribution sampling network to obtain a target predicted growth height.

[0070] In this embodiment, the optimal parameter θ * Take out In order to update the trained prediction model and distribution sampling network. Further, the environmental characteristic data of the planting environment of the plant to be predicted can be obtained, and the newly obtained environmental characteristic data (x * ) is input into the trained independent distribution sampling network C1 to obtain independent latent variables The independent latent variable may be a first latent variable.

[0071] Furthermore, at least one historical growth data set can be obtained to obtain the target value of the training sample in the target domain. The first latent variable and the historical growth data set can be input into the joint distribution sampling network C2 to obtain the first mean and the first variance Then we get the latent variable distribution belonging to the target domain, that is, we get the latent variable distribution corresponding to the historical growth data set Based on the Bayesian rule and using importance sampling, the latent variable distribution can be sampled to obtain S latent variables That is, multiple second latent variables can be used to obtain a prediction model based on the target domain. The mean of the sum of all predicted values ​​based on S latent variables z′, that is, Get the initial predicted growth height.

[0072] According to the above steps, the initial predicted growth height corresponding to at least one historical growth data set can be obtained respectively, and at least one initial predicted growth height can be obtained. Based on all target values The mean of the sum of all predicted values, further, the final target predicted growth height y can be obtained based on the following formula: * ;

[0073]

[0074] As an optional implementation, the method may also include: determining a target planting environment whose environmental similarity with the planting environment is greater than a similarity threshold; constructing first training data of the planting environment and second training data of the target planting environment, wherein the first training data at least includes a first environmental characteristic data sample of the planting environment and a first growth data sample of the plant grown in the planting environment; the second training data at least includes a second environmental characteristic data sample of the target planting environment and a second growth data sample of the plant grown in the target planting environment; and using the first training data and the second training data to train a distributed sampling network.

[0075] In this embodiment, the planting environment of the plant to be predicted (i.e., the target domain) and the target planting environment (i.e., the source domain) whose environmental similarity with the planting environment is greater than the similarity threshold can be determined. Further, first training data of the planting environment and second training data of the target planting environment can be constructed. Further, using the first training data and the second training data, a distribution sampling network can be trained. The first training data can be the training data of the target domain, which can be obtained by (X t Y t ) indicates that it may at least include a first environmental feature data sample of the planting environment and a first growth data sample of the plant grown in the planting environment. The second training data may be training data of the source domain, which may be represented by (X s , Y s ) indicates that it may at least include a second environmental characteristic data sample of the target planting environment and a second growth data sample of the plants grown in the target planting environment. It should be noted that the above-mentioned first training data and second training data are presented in the form of examples only and are not specifically limited here.

[0076] Optionally, historical growth information data of trees in the source domain can be collected from the power grid database in, It can be used to characterize the number of days recorded. The historical growth information data can include data of trees such as bamboo, fir, pine, eucalyptus, etc., so as to obtain the second growth data sample in the second training data. Similarly, the historical growth information data of trees in the target domain can be collected. in, can be used to represent the number of days recorded. The tree type of trees in the target domain can be different from the tree type of trees in the source domain, and, The dataset collected in the source domain can complement the dataset in the target domain.

[0077] Optionally, the meteorological data A corresponding to the second growth data sample in the source domain can be collected by using a collection device, such as a thermometer, a hygrometer, etc. s ={tp s ,h s , ws s , p s , r s},in It can be used to represent daily temperature. Can be used to characterize daily humidity, It can be used to represent the daily wind speed. It can be used to represent the daily surface air pressure. It can be used to characterize daily rainfall; and can collect the latitude and longitude information of trees in the source domain {lo s , la s Soil type s , and the planting density pd s .

[0078] Similarly, the meteorological data A corresponding to the first growth data sample in the target domain can be collected t ={tp t ,h t ws t , p t , r t},in It can be used to represent daily temperature. Can be used to characterize daily humidity, It can be used to represent the daily wind speed. It can be used to represent the daily surface air pressure. It can be used to characterize daily rainfall; it can collect the latitude and longitude information of trees in the target domain {lot, la t Soil type t , and the planting density pd t .

[0079] Further, based on the collected data, a first environment feature data sample, a second environment feature data sample, a first growth data sample, and a second growth data sample can be constructed to obtain a first training data (X t Y t ) and the second training data (X s , Y s ).

[0080] For example, construct a source domain training set, that is, the second training data (Xs Y s ).in The i-th input feature (ie, the second environment feature data sample) can be The target value corresponding to the input feature (ie, the second growth data sample) can be Similarly, construct the target domain training set, that is, the first training data (X t , Y t ),in The i-th input feature (i.e., the first environmental feature data sample) is The corresponding target value (i.e., the first growth data sample) is It should be noted that the above method of constructing the first training data and the second training data and the content included in the first environmental feature data sample are only for illustrative purposes and are not specifically limited here.

[0081] Optionally, a distributed sampling network may be trained using the constructed first training data and the second training data. The method for training the distributed sampling network is further described below.

[0082] As an optional implementation, the first training data and the second training data are used to train a distribution sampling network, including: respectively inputting the first environmental feature data sample and the second environmental feature data sample into the first sub-distribution sampling network for analysis to obtain the third latent variable of the first environmental feature data sample mapped to the latent space, and the fourth latent variable of the second environmental feature data sample mapped to the latent space; inputting the third latent variable, the first growth data sample, the fourth latent variable and the second growth data sample into the second sub-distribution sampling network for analysis to obtain the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment; based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, construct a first sub-target equation; based on the first sub-target equation, determine the target model parameters of the sub-distribution sampling network, wherein the sub-distribution sampling network includes the first sub-distribution sampling network and the second sub-distribution sampling network; using the target model parameters, update the initial model parameters of the sub-distribution sampling network to obtain a distribution sampling network.

[0083] In this embodiment, the first environment feature data samples (that is, the input feature X of the target domain) can be respectively t ) and the second environment feature data sample (that is, the input feature X of the source domain s), input into the first sub-distribution sampling network, obtain the third latent variable of the first environmental feature data sample mapped to the latent space, and the fourth latent variable of the second environmental feature data sample mapped to the latent space. Furthermore, the third latent variable, the first growth data sample, the fourth latent variable and the second growth data sample can be input into the second sub-distribution sampling network, and the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment can be obtained; based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, the first sub-goal equation can be constructed; based on the first sub-goal equation, the target model parameters of the sub-distribution sampling network can be determined, and the initial model parameters of the sub-distribution sampling network can be updated using the target model parameters to obtain the distribution sampling network.

[0084] In this embodiment, the latent variable extraction module may be a sub-distribution sampling network, which may include: a sub-joint distribution sampling network C2 and a sub-independent distribution sampling network C1. The trained sub-distribution sampling network may be a distribution sampling network, which may include a trained sub-joint distribution sampling network (i.e., a joint distribution sampling network) and a trained sub-independent distribution sampling network (i.e., an independent distribution sampling network).

[0085] In this embodiment, the first environment feature data sample and the second environment feature data sample (that is, the input feature X of the source domain) can be respectively s ), input into the first sub-distribution sampling network, obtain the third latent variable of the first environment feature data sample mapped to the latent space, and the fourth latent variable of the second environment feature data sample mapped to the latent space. Further, the third latent variable, the first growth data sample, the fourth latent variable and the second growth data sample can be input into the second sub-distribution sampling network to obtain the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment.

[0086] Optionally, the sub-distribution sampling network can be a pre-built latent variable extraction module, which can include a first sub-distribution sampling network and a second sub-distribution sampling network. The first sub-distribution sampling network can be a sub-independent distribution sampling network, which can be used to determine the latent variable. The second sub-distribution sampling network can be a sub-joint distribution sampling network, which can be used to determine the distribution of the latent variable. The third latent variable can be an independent latent variable corresponding to the input feature of the target domain. The fourth latent variable mentioned above can be an independent latent variable corresponding to the input feature of the source domain The latent variable distribution of the above planting environment can be the latent variable distribution belonging to the target domain The latent variable distribution of the target planting environment can be the latent variable distribution belonging to the source domain

[0087] For example, the first sub-distribution sampling network can be constructed in advance. The input feature X of the source domain can be s Input into the first sub-distribution sampling network C1 to obtain the independent latent variables corresponding to the input features of the source domain Similarly, the input feature X of the target domain can be t Input into the first sub-distribution sampling network C1 to obtain the independent latent variables corresponding to the input features of the target domain Wherein, φ1 can be a parameter of the first sub-distribution sampling network, and the first sub-distribution sampling network can include but is not limited to a fully connected neural network, a convolutional neural network, etc. There is no specific restriction on the type of the first sub-distribution sampling network. Furthermore, the second sub-distribution sampling network can be constructed in advance. The independent latent variable and the target value Y in the source domain s (That is, the second growth data sample) is input into the second sub-distribution sampling network C2 to obtain the mean and variance corresponding to the source domain Then we get the hidden variable distribution belonging to the source domain Similarly, the independent latent variables can be and the target value Y of the target domain t (That is, the first growth data sample) is input into the second sub-distribution sampling network C2 to obtain the mean and variance corresponding to the target domain Then we get the latent variable distribution belonging to the target domain Wherein, φ2 may be a parameter of the second sub-joint distribution sampling network, which may include but is not limited to LSTM neural network, recurrent neural network, gated recurrent neural network, and variants of recurrent neural network. It should be noted that the type of the second sub-joint distribution sampling network is not specifically limited here.

[0088] As an optional implementation, based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, a first sub-goal equation is constructed, including: sampling the latent variable distribution of the planting environment to obtain multiple fifth latent variables; sampling the latent variable distribution of the target planting environment to obtain multiple sixth latent variables; based on multiple fifth latent variables and multiple sixth latent variables, constructing the first sub-goal equation.

[0089] In this embodiment, based on the constructed discriminator module, a first sub-goal equation can be constructed based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment; based on the first sub-goal equation, the target model parameters of the sub-distribution sampling network can be determined, and the initial model parameters of the sub-distribution sampling network can be updated using the target model parameters to obtain a distribution sampling network.

[0090] In this embodiment, the latent variable distribution of the planting environment can be sampled to obtain multiple fifth latent variables; the latent variable distribution of the target planting environment can be sampled to obtain multiple sixth latent variables; based on the multiple fifth latent variables and the multiple sixth latent variables, the first sub-goal equation is constructed. Among them, the multiple fifth latent variables can be latent variables of the target domain after data enhancement, which can be used The above-mentioned multiple sixth latent variables can be latent variables of the source domain obtained by sampling, which can be expressed as express.

[0091] Alternatively, the latent variable distribution based on the target planting environment can be obtained using Bayesian rule and importance sampling Hidden variables That is, multiple sixth latent variables can be based on the latent variable distribution of the target domain in order to achieve data enhancement of the latent variables of the target domain. Using Bayesian rule and importance sampling, we can get Hidden variables That is, multiple fifth latent variables are obtained. Further, the pre-established discriminator module D (which can be referred to as the discriminator) can be used to discriminate the input latent variables and Whether it is from the source domain or the target domain, its objective equation (that is, the first sub-objective equation) is as follows:

[0092]

[0093] Among them, θ D It can be used to represent the parameters of the discriminator, φ = {φ1, φ2}, and the discriminator can be a fully connected neural network or a convolutional neural network. It should be noted that this is only an example and there is no specific restriction on the type of the discriminator.

[0094] For example, based on the latent variable distribution of the source domain Using Bayesian rule and importance sampling, we can get Hidden variables Latent variable distribution based on target domain Using Bayesian rule and importance sampling, we can get Hidden variables Use the discriminator module D to discriminate the latent variables of the input and Is it from the source domain or the target domain, thus obtaining the first sub-goal equation:

[0095]

[0096] As an optional implementation, multiple second latent variables are predicted separately to obtain multiple predicted growth heights, including: controlling the prediction model to predict the second latent variable to obtain the predicted growth height, wherein the prediction model is trained based on the first training data and the second training data.

[0097] In this embodiment, the above prediction model may be a prediction model, which may include a source domain prediction model. and target domain prediction model It can be used to predict latent variables to obtain predicted growth height.

[0098] As an optional implementation, the method may also include: mapping the first environmental feature data sample to the latent space to obtain multiple seventh latent variables; controlling the sub-prediction model of the target planting environment to predict multiple sixth latent variables corresponding to the target planting environment to obtain a first predicted growth height, and controlling the sub-prediction model of the planting environment to predict the seventh latent variable to obtain a second predicted growth height; constructing a second sub-target equation using the first predicted growth height and the second predicted growth height; based on the second sub-target equation, determining the target model parameters of the sub-prediction model of the target planting environment, and the target model parameters of the sub-prediction model of the planting environment; using the target model parameters, respectively updating the initial model parameters of the sub-prediction model of the target planting environment, and the initial model parameters of the sub-prediction model of the planting environment, to obtain a prediction model, wherein the prediction model includes an adjusted sub-prediction model of the target planting environment, and an adjusted sub-prediction model of the planting environment.

[0099] In this embodiment, the first environmental characteristic data sample can be mapped to the latent space to obtain multiple seventh latent variables; the sub-prediction model of the target planting environment is controlled to predict multiple sixth latent variables corresponding to the target planting environment to obtain a first predicted growth height, and the sub-prediction model of the planting environment is controlled to predict the seventh latent variable to obtain a second predicted growth height; the first predicted growth height and the second predicted growth height are used to construct a second sub-target equation; based on the second sub-target equation, the target model parameters of the sub-prediction model of the target planting environment and the target model parameters of the sub-prediction model of the planting environment are determined; using the target model parameters, the initial model parameters of the sub-prediction model of the target planting environment and the initial model parameters of the sub-prediction model of the planting environment can be updated respectively to obtain a prediction model, wherein the prediction model includes an adjusted sub-prediction model of the target planting environment and an adjusted sub-prediction model of the planting environment. The above-mentioned seventh latent variable can be used express.

[0100] Optionally, based on the latent variable distribution of the target domain It can be obtained by using Bayes' rule and importance sampling Hidden variables (That is, the seventh latent variable). The sub-prediction model of the source domain can be constructed in advance and the sub-prediction model of the target domain Use the sampled z s (i.e., the sixth latent variable) and z′ t The plant height in the source domain and the plant height in the target domain are predicted to obtain a first predicted growth height and a second predicted growth height. and the sub-prediction model of the target domain It can be a fully connected neural network. The activation function can be a linear rectification function (ReLU). It should be noted that the type of activation function here is only an example and is not specifically limited.

[0101] Furthermore, based on the first predicted growth height and the second predicted growth height, a regression error objective equation of the sub-prediction model can be constructed, that is, a second sub-objective equation. The second sub-objective equation can be used to improve the prediction accuracy of the tree height and can be expressed by the following formula:

[0102]

[0103] in, and Can be the sub-prediction model of the source domain respectively. and the sub-prediction model of the target domain Parameters.

[0104] Optionally, in order to perform domain adaptation on target samples near the regression boundary to improve the prediction accuracy in the target domain, this embodiment may also construct a prediction model adaptation error target equation as follows:

[0105]

[0106] Optionally, the target model parameters of the sub-prediction model can be determined using the second sub-target equation and the adaptive error target equation. The initial model parameters of the sub-prediction model are updated using the target model parameters to obtain a trained prediction model, which can predict the second latent variable to obtain a predicted growth height.

[0107] As an optional implementation, the method may also include: controlling the sub-reconstruction model corresponding to the target planting environment to reconstruct the sixth latent variable to obtain a first reconstruction result, and controlling the sub-reconstruction model corresponding to the planting environment to reconstruct the seventh latent variable to obtain a second reconstruction result; based on the first reconstruction result and the second reconstruction result, constructing a third sub-target equation.

[0108] In this embodiment, the sub-reconstruction model corresponding to the target planting environment can be controlled to reconstruct the sixth latent variable to obtain a first reconstruction result, and the sub-reconstruction model corresponding to the planting environment can be controlled to reconstruct the seventh latent variable to obtain a second reconstruction result; based on the first reconstruction result and the second reconstruction result, a third sub-objective equation is constructed, and the accuracy of sampling in the joint distribution sampling network can be improved through the third sub-objective equation. Among them,

[0109] Optionally, in order to improve the sampling accuracy of the independent distribution sampling network C1 and the joint distribution sampling network C2 and reduce the distance between the original data and the generated data, a sub-reconstruction model of the source domain is also constructed and target domain sub-reconstruction model Use the sampled z s and z′ t Reconstruct data in the source domain and the target domain. and the sub-reconstruction model of the target domain It can be a fully connected neural network, and the activation function can be Linear. It should be noted that there is no specific restriction on the type of the sub-reconstruction model and the type of the activation function.

[0110] Optionally, based on the first reconstruction structure and the second reconstruction result, a target equation of the reconstruction model (i.e., the third sub-target equation) can be constructed. By using the third sub-target equation, the sampling accuracy of the independent distribution sampling network C1 and the joint distribution sampling network C2 can be improved. The third sub-target equation is as follows:

[0111]

[0112] in, and The model can be reconstructed for the source domain respectively and target domain reconstruction model Parameters.

[0113] As an optional implementation, based on the first sub-goal equation, the target model parameters of the sub-distribution sampling network are determined, including: constructing a target equation based on the first sub-goal equation, the second sub-goal equation and the third sub-goal equation; using the target equation, determining the target model parameters of the sub-distribution sampling network.

[0114] As an optional implementation, based on the second sub-target equation, the target model parameters of the sub-prediction model of the target planting environment, and the target model parameters of the sub-prediction model of the planting environment are determined, including: constructing a target equation based on the first sub-target equation, the second sub-target equation and the third sub-target equation; using the target equation to determine the target model parameters of the sub-prediction model of the target planting environment, and the target model parameters of the sub-prediction model of the planting environment.

[0115] In this embodiment, the target equation can be constructed based on the first sub-target equation, the second sub-target equation and the third sub-target equation; using the target equation, the target model parameters of the sub-prediction model of the target planting environment, the target model parameters of the sub-prediction model of the planting environment, and the target model parameters of the sub-distribution sampling network can be determined.

[0116] In this embodiment, based on the constructed first sub-goal equation, the second sub-goal equation and the third sub-goal equation, the final goal equation can be obtained:

[0117]

[0118] in,

[0119] In this embodiment, the target model parameters of the sub-prediction model of the target planting environment, the target model parameters of the sub-prediction model of the planting environment, and the target model parameters of the sub-distribution sampling network can be determined using the above target equation.

[0120] Optionally, in order to prevent overfitting during model training, the number of iterations during training can be limited. That is, the maximum step size T and the batch size can be set in advance. The model parameters θ can be set and fixed in advance. Based on the mini-batch gradient descent method and the adaptive learning rate optimization algorithm (for example, the Adam optimization algorithm), the gradient of the discriminator D can be obtained. Update θ D . The gradient of the discriminator D is as follows:

[0121]

[0122] Furthermore, it is possible to fix Based on the mini-batch gradient descent method, the gradients of the remaining models can be obtained And update θ. The gradient As shown below:

[0123]

[0124] Repeat T times or The optimal parameters can be obtained and θ * . Among them, ∈=1e-6 can be a preset error. θ * The target model parameters that can be used include the target planting environment's sub-prediction model's target model parameters, the planting environment's sub-prediction model's target model parameters, and the sub-distribution sampling network's target model parameters.

[0125] Optionally, according to θ * The sub-prediction model of the target planting environment, the sub-prediction model of the planting environment, and the initial model parameters of the sub-distribution sampling network can be updated to obtain a trained prediction model and distribution sampling network.

[0126] As an optional implementation, the method may further include: determining, based on the target predicted growth height, a time when the predicted plant will have an impact on the power transmission line in the planting environment.

[0127] In this embodiment, the target predicted growth height may be the growth height of the plant to be predicted at at least one moment in the future. Therefore, based on the target predicted growth height, it can be determined at what moment the plant to be predicted will have an impact on the power transmission line in the planting environment.

[0128] Optionally, the time when the predicted plant will affect the power transmission line in the planting environment can be displayed on the mobile terminal or display to remind the user. It should be noted that there is no specific limitation on the reminder method, and reminders can also be made through text messages, information push, etc.

[0129] In this embodiment, a detailed prediction is performed for fast-growing tree species, thereby predicting the time point when hidden dangers of power transmission lines will be formed.

[0130] In this embodiment, at least one historical growth data set and environmental characteristic data are used to obtain at least one initial predicted growth height corresponding to the at least one historical growth data set. The initial predicted growth height corresponds to the historical growth data set one-to-one. The at least one initial predicted growth height is converted to obtain a target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0131] At present, the classic tree growth prediction models are usually mathematical models and nonlinear mixed effect models, but the solutions of these two models are quite complicated and it is difficult to describe the nonlinear relationship between tree growth and age.

[0132] In related technologies, with the development of artificial intelligence technology, some scholars have used the good nonlinear fitting ability of artificial intelligence models to solve problems that the above models cannot solve. However, artificial intelligence models are data-driven models, and accurate prediction of growth height depends on a large amount of training data. At present, the collection of tree growth data is only for certain specific tree species, and the resolution of the data is monthly or annual. It is difficult to make detailed predictions for fast-growing tree species, and then predict the time node for the formation of transmission line hazards. There is a technical problem of low prediction accuracy of plant growth height.

[0133] In an embodiment of the present invention, at least one historical growth data set and environmental characteristic data are used to obtain at least one initial predicted growth height corresponding to the at least one historical growth data set. The initial predicted growth height corresponds to the historical growth data set one-to-one. The at least one initial predicted growth height obtained is converted to obtain a target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0134] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.

[0135] An accurate tree growth prediction model can accurately analyze and predict the time nodes when trees may pose hidden dangers to transmission lines, improve the efficiency of transmission line operation and maintenance inspections, and greatly reduce the maintenance cost of transmission lines. Therefore, studying an accurate tree growth prediction model is of great significance for the operation and maintenance of transmission lines.

[0136] At present, the classic tree growth prediction models are usually mathematical models and nonlinear mixed effect models, but the solutions of these two models are quite complicated and it is difficult to describe the nonlinear relationship between tree growth and age.

[0137] In related technologies, with the development of artificial intelligence technology, some scholars have used the good nonlinear fitting ability of artificial intelligence models to solve problems that the above models cannot solve. However, artificial intelligence models are data-driven models, and accurate prediction of growth height depends on a large amount of training data. At present, the collection of tree growth data is only for certain specific tree species, and the resolution of the data is monthly or annual. It is difficult to make detailed predictions for fast-growing tree species, and then predict the time node for the formation of transmission line hazards. There is a technical problem of low prediction accuracy of plant growth height.

[0138] In this embodiment, in order to solve the above problems, a tree growth prediction model based on latent space knowledge transfer is proposed, and the transfer learning technology is used to accurately predict tree growth. The method pre-constructs a tree growth prediction model training set; establishes a latent variable extraction module; establishes a discriminator module and constructs a target equation; establishes a prediction model module and constructs a target equation; establishes a reconstruction model module and constructs a target equation; uses the above target equation to construct the final target equation, and trains the established model based on the gradient descent method (for example, mini-batch) to obtain a trained tree growth prediction model. The trained tree growth prediction model is used to predict the input features to determine the target predicted growth height of the plant to be predicted in the planting environment, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0139] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments; the size of the numbers in the embodiments is only for illustration and is not specifically limited here. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention and are not specifically limited here.

[0140] As an optional implementation manner, first training data and second training data are constructed.

[0141] In this embodiment, the planting environment of the plant to be predicted (i.e., the target domain) and the target planting environment (i.e., the source domain) whose environmental similarity with the planting environment is greater than a similarity threshold can be determined. Further, first training data of the planting environment and second training data of the target planting environment can be constructed. Among them, the above-mentioned first training data can be the training data of the target domain, which can at least include the first environmental characteristic data sample of the planting environment and the first growth data sample of the plant grown in the planting environment. The above-mentioned second training data can be the training data of the source domain, which can at least include the second environmental characteristic data sample of the target planting environment and the second growth data sample of the plant grown in the target planting environment.

[0142] In this embodiment, the first training data and the second training data are constructed, and the model in the source domain and the model corresponding to the target domain are obtained by training. A model trained in the source domain can be migrated to a target domain with missing labels, so that the model has better generalization ability in the target domain, thereby improving the accuracy of the model in predicting the plant growth height. It should be noted that the domain adaptation method is not specifically limited here.

[0143] Optionally, historical growth information data of trees in the source domain can be collected from the power grid database in, It can be used to characterize the number of days recorded. The historical growth information data can include data of trees such as bamboo, fir, pine, eucalyptus, etc., so as to obtain the second growth data sample in the second training data. Similarly, the historical growth information data of trees in the target domain can be collected. in, can be used to represent the number of days recorded. The tree type of trees in the target domain can be different from the tree type of trees in the source domain, and, The dataset collected in the source domain can complement the dataset in the target domain.

[0144] Optionally, the meteorological data corresponding to the second growth data sample in the source domain can be collected by using a collection device, such as a thermometer, a hygrometer, etc. s ,h s ws s , p s , r s},in It can be used to represent daily temperature. Can be used to characterize daily humidity, It can be used to represent the daily wind speed. It can be used to represent the daily surface air pressure. It can be used to characterize daily rainfall; and can collect the latitude and longitude information of trees in the source domain {lo s , la s Soil type s , and the planting density pd s .

[0145] Similarly, the meteorological data A corresponding to the first growth data sample in the target domain can be collected t ={tp t ,h t ws t , p t , r t},in It can be used to represent daily temperature. Can be used to characterize daily humidity, It can be used to represent the daily wind speed. It can be used to represent the daily surface air pressure. It can be used to characterize daily rainfall; it can collect the latitude and longitude information of the trees in the target domain {lo t , la t Soil type t , and the planting density pdt .

[0146] Further, based on the collected data, a first environment feature data sample, a second environment feature data sample, a first growth data sample, and a second growth data sample can be constructed to obtain a first training data (X t , Y t ) and the second training data (X s ,Y s ).

[0147] For example, construct a source domain training set, that is, the second training data (X s , Y s ).in The i-th input feature (ie, the second environment feature data sample) can be The target value corresponding to the input feature (ie, the second growth data sample) can be Similarly, construct the target domain training set, that is, the first training data (X t , Y t ),in The i-th input feature (i.e., the first environmental feature data sample) is The corresponding target value (i.e., the first growth data sample) is

[0148] As an optional implementation, a latent variable extraction module is constructed.

[0149] In this embodiment, the latent variable extraction module may be a sub-distribution sampling network, which may include: a sub-joint distribution sampling network C2 and a sub-independent distribution sampling network C1. The trained sub-distribution sampling network may be a distribution sampling network, which may include a trained sub-joint distribution sampling network (i.e., a joint distribution sampling network) and a trained sub-independent distribution sampling network (i.e., an independent distribution sampling network).

[0150] In this embodiment, the first environment feature data samples (that is, the input feature X of the target domain) can be respectively t ) and the second environment feature data sample (that is, the input feature X of the source domain s ), input into the first sub-distribution sampling network, obtain the third latent variable of the first environment feature data sample mapped to the latent space, and the fourth latent variable of the second environment feature data sample mapped to the latent space. Further, the third latent variable, the first growth data sample, the fourth latent variable and the second growth data sample can be input into the second sub-distribution sampling network to obtain the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment.

[0151] Optionally, the sub-distribution sampling network can be a pre-built latent variable extraction module, which can include a first sub-distribution sampling network and a second sub-distribution sampling network. The first sub-distribution sampling network can be a sub-independent distribution sampling network, which can be used to determine the latent variable. The second sub-distribution sampling network can be a sub-joint distribution sampling network, which can be used to determine the distribution of the latent variable. The third latent variable can be an independent latent variable corresponding to the input feature of the target domain. The fourth latent variable mentioned above can be an independent latent variable corresponding to the input feature of the source domain The latent variable distribution of the above planting environment can be the latent variable distribution belonging to the target domain The latent variable distribution of the target planting environment can be the latent variable distribution belonging to the source domain

[0152] For example, the first sub-distribution sampling network can be constructed in advance. The input feature X of the source domain can be s Input into the first sub-distribution sampling network C1 to obtain the independent latent variables corresponding to the input features of the source domain Similarly, the input feature X of the target domain can be t Input into the first sub-distribution sampling network C1 to obtain the independent latent variables corresponding to the input features of the target domain Wherein, φ1 can be a parameter of the first sub-distribution sampling network, and the first sub-distribution sampling network can include but is not limited to a fully connected neural network, a convolutional neural network, etc. There is no specific restriction on the type of the first sub-distribution sampling network. Furthermore, the second sub-distribution sampling network can be constructed in advance. The independent latent variable and the target value Y in the source domain s (That is, the second growth data sample) is input into the second sub-distribution sampling network C2 to obtain the mean and variance corresponding to the source domain Then we get the hidden variable distribution belonging to the source domain Similarly, the independent latent variables can be and the target value Y of the target domain t (That is, the first growth data sample) is input into the second sub-distribution sampling network C2 to obtain the mean and variance corresponding to the target domain Then we get the latent variable distribution belonging to the target domain Wherein, φ2 may be a parameter of the second sub-joint distribution sampling network, which may include but is not limited to LSTM neural network, recurrent neural network, gated recurrent neural network, and variants of recurrent neural network. It should be noted that the type of the second sub-joint distribution sampling network is not specifically limited here.

[0153] For example, Figure 2 is a schematic diagram of a latent variable extraction module according to an embodiment of the present invention, such as Figure 2 As shown, the second environment feature data sample of the source domain and the first environment feature data sample of the target domain are obtained, and the second environment feature data sample 201 and the first environment feature data sample 202 can be input into the first sub-independent distribution sampling network 203 to obtain the third latent variable and the fourth latent variable. The third latent variable, the fourth latent variable, the first growth data sample 207 and the second growth data sample 208 can be input into the second sub-independent distribution sampling network 204 to obtain the latent variable distribution 205 of the planting environment and the latent variable distribution 206 of the target planting environment.

[0154] As an optional implementation, a discriminator can be built and the first sub-objective equation can be constructed.

[0155] In this embodiment, a discriminator module D can be established to construct a first sub-goal equation, wherein the first sub-goal equation can be used to discriminate the source of the input latent variable.

[0156] In this embodiment, based on the constructed discriminator module, a first sub-goal equation can be constructed based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment; based on the first sub-goal equation, the target model parameters of the sub-distribution sampling network can be determined, and the initial model parameters of the sub-distribution sampling network can be updated using the target model parameters to obtain a distribution sampling network.

[0157] In this embodiment, the latent variable distribution of the planting environment can be sampled to obtain multiple fifth latent variables; the latent variable distribution of the target planting environment can be sampled to obtain multiple sixth latent variables; based on the multiple fifth latent variables and the multiple sixth latent variables, the first sub-goal equation is constructed. Among them, the multiple fifth latent variables can be latent variables of the target domain after data enhancement, which can be used The above-mentioned multiple sixth latent variables can be latent variables of the source domain obtained by sampling, which can be expressed as express.

[0158] Optionally, based on the latent variable distribution of the target planting environment, the Bayesian rule and importance sampling can be used to obtain

[0159] arrive Hidden variables That is, multiple sixth latent variables can be based on the latent variable distribution of the target domain in order to achieve data enhancement of the latent variables of the target domain. Using Bayesian rule and importance sampling, we can get Hidden variables That is, multiple fifth latent variables are obtained. Further, the pre-established discriminator module D (which can be referred to as the discriminator) can be used to discriminate the input latent variables and Whether it is from the source domain or the target domain, its objective equation (that is, the first sub-objective equation) is as follows:

[0160]

[0161] Among them, θ D It can be used to represent the parameters of the discriminator, φ = {φ1, φ2}, and the discriminator can be a fully connected neural network or a convolutional neural network. It should be noted that this is only an example and there is no specific restriction on the type of the discriminator.

[0162] For example, Figure 3 is a schematic diagram of a discriminator module according to an embodiment of the present invention, such as Figure 3 As shown, the latent variable distribution 302 based on the source domain Using Bayesian rule and importance sampling, we can get Hidden variables Hidden variable distribution based on target domain 303 Using Bayesian rule and importance sampling, we can get Hidden variables Using the discriminator module 301, discriminate the latent variables of the input and Whether it is from the source domain or the target domain can be judged by the first sub-goal equation:

[0163]

[0164] As an optional implementation, a sub-prediction model can be pre-constructed and a second sub-objective equation can be constructed.

[0165] In this embodiment, the first environmental characteristic data sample can be mapped to the latent space to obtain multiple seventh latent variables; the sub-prediction model of the target planting environment is controlled to predict multiple sixth latent variables corresponding to the target planting environment to obtain a first predicted growth height, and the sub-prediction model of the planting environment is controlled to predict the seventh latent variable to obtain a second predicted growth height; the first predicted growth height and the second predicted growth height are used to construct a second sub-target equation; based on the second sub-target equation, the target model parameters of the sub-prediction model of the target planting environment and the target model parameters of the sub-prediction model of the planting environment are determined; using the target model parameters, the initial model parameters of the sub-prediction model of the target planting environment and the initial model parameters of the sub-prediction model of the planting environment can be updated respectively to obtain a prediction model, wherein the prediction model includes an adjusted sub-prediction model of the target planting environment and an adjusted sub-prediction model of the planting environment. The above-mentioned seventh latent variable can be used express.

[0166] Optionally, Figure 4is a schematic diagram of a prediction model module according to an embodiment of the present invention, such as Figure 4 As shown, the latent variable distribution based on the target domain is 403 It can be obtained by using Bayes' rule and importance sampling Hidden variables (That is, the seventh latent variable). The sub-prediction model of the source domain can be constructed in advance and the sub-prediction model of the target domain We use the z obtained by sampling the latent variable distribution 404 s (i.e., the sixth latent variable) and z′ t The plant height in the source domain and the plant height in the target domain are predicted, that is, the seventh latent variable can be input into the sub-prediction model 401 of the target domain (that is, the sub-prediction model of the planting environment) to obtain the first predicted growth height, and the sixth latent variable can be input into the sub-prediction model 402 of the source domain (that is, the sub-prediction model of the target planting environment) to obtain the second predicted growth height. and the sub-prediction model of the target domain Domain adaptation can be performed between them.

[0167] Furthermore, based on the first predicted growth height and the second predicted growth height, a regression error objective equation of the sub-prediction model can be constructed, that is, a second sub-objective equation. The second sub-objective equation can be used to improve the prediction accuracy of the tree height and can be expressed by the following formula:

[0168]

[0169] in, and Can be used as sub-prediction models for the source domain and the sub-prediction model of the target domain Parameters.

[0170] Optionally, in order to perform domain adaptation on target samples near the regression boundary to improve the prediction accuracy in the target domain, this embodiment may also construct a prediction model adaptation error target equation as follows:

[0171]

[0172] Optionally, the target model parameters of the sub-prediction model can be determined using the second sub-target equation and the adaptive error target equation. The initial model parameters of the sub-prediction model are updated using the target model parameters to obtain a trained prediction model, which can predict the second latent variable to obtain a predicted growth height.

[0173] As an optional implementation, a sub-reconstruction model may be constructed in advance, and a target equation may be constructed.

[0174] In this embodiment, the sub-reconstruction model corresponding to the target planting environment can be controlled to reconstruct the sixth latent variable to obtain a first reconstruction result, and the sub-reconstruction model corresponding to the planting environment can be controlled to reconstruct the seventh latent variable to obtain a second reconstruction result; based on the first reconstruction result and the second reconstruction result, a third sub-objective equation is constructed, and the third sub-objective equation can be used to improve the sampling accuracy of the joint distribution sampling network.

[0175] Optionally, in order to improve the sampling accuracy of the independent distribution sampling network C1 and the joint distribution sampling network C2 and reduce the distance between the original data and the generated data, a sub-reconstruction model of the source domain is also constructed and target domain sub-reconstruction model Figure 5 is a schematic diagram of a reconstruction model module according to an embodiment of the present invention, such as Figure 5 As shown, the hidden variable distribution of the source domain and the hidden variable distribution of the target domain are sampled to obtain z s and z′ t Data reconstruction is performed on the source domain and the target domain, that is, the sixth latent variable can be reconstructed using the sub-reconstruction model 501 of the source domain, and the seventh latent variable can be reconstructed using the sub-reconstruction model 502 of the target domain. and the sub-reconstruction model of the target domain It can be a fully connected neural network, and the activation function can be a linear function. It should be noted that the type of the sub-reconstruction model and the type of the activation function are not specifically limited here.

[0176] Optionally, based on the first reconstruction structure and the second reconstruction result, a target equation of the reconstruction model (i.e., the third sub-target equation) can be constructed. By using the third sub-target equation, the sampling accuracy of the independent distribution sampling network C1 and the joint distribution sampling network C2 can be improved. The third sub-target equation is as follows:

[0177]

[0178] in, and The model can be reconstructed for the source domain respectively and target domain reconstruction model Parameters.

[0179] As an optional implementation, the target equation can be constructed based on the first sub-target equation, the second sub-target equation and the third sub-target equation.

[0180] In this embodiment, based on the constructed first sub-goal equation, the second sub-goal equation and the third sub-goal equation, the final goal equation (L) can be obtained:

[0181]

[0182] in,

[0183] In this embodiment, the target model parameters of the sub-prediction model of the target planting environment, the target model parameters of the sub-prediction model of the planting environment, and the target model parameters of the sub-distribution sampling network can be determined using the above target equation.

[0184] Optionally, in order to prevent overfitting during model training, the number of iterations during training can be limited. That is, the maximum step size T and the batch size can be set in advance. The model parameters θ can be set and fixed in advance. Based on the mini-batch gradient descent method and the adaptive learning rate optimization algorithm (for example, the Adam optimization algorithm), the gradient of the discriminator D can be obtained. Update θ D . The gradient of the discriminator D is as follows:

[0185]

[0186] Furthermore, it is possible to fix Based on the mini-batch gradient descent method, the gradients of the remaining models can be obtained And update θ. The gradient As shown below:

[0187]

[0188] Repeat T times or The optimal parameters can be obtained and θ * . Among them, ∈=1e-6 can be a preset error. θ * The target model parameters that can be used include the target planting environment's sub-prediction model's target model parameters, the planting environment's sub-prediction model's target model parameters, and the sub-distribution sampling network's target model parameters.

[0189] Optionally, according to θ * The sub-prediction model of the target planting environment, the sub-prediction model of the planting environment, and the initial model parameters of the sub-distribution sampling network can be updated to obtain a trained prediction model and distribution sampling network.

[0190] As an optional implementation method, environmental characteristic data is obtained, and the environmental characteristic data is processed using a trained prediction model and a distribution sampling network to obtain a target predicted growth height.

[0191] In this embodiment, the optimal parameter θ * Take out In order to update the trained prediction model and distribution sampling network. Further, the environmental characteristic data of the planting environment of the plant to be predicted can be obtained, and the newly obtained environmental characteristic data (x * ) is input into the trained independent distribution sampling network C1 to obtain independent latent variables The independent latent variable may be a first latent variable.

[0192] Furthermore, at least one historical growth data set can be obtained to obtain the target value of the training sample in the target domain. The first latent variable and the historical growth data set (for example, the nth t Target domain training sample target value ), input into the joint distribution sampling network C2, and get the nth t Target domain training sample target value The corresponding first mean and first variance Then we get the latent variable distribution belonging to the target domain, that is, we get the latent variable distribution corresponding to the historical growth data set Based on the Bayesian rule and using importance sampling, the latent variable distribution can be sampled to obtain S latent variables That is, multiple second latent variables. The prediction model based on the target domain can be obtained Based on S latent variables The mean of the sum of all predicted values ​​of Get the initial predicted growth height.

[0193] According to the above steps, the initial predicted growth height corresponding to at least one historical growth data set can be obtained respectively, and at least one initial predicted growth height can be obtained. Based on all target values The mean of the sum of all predicted values, further, the final target predicted growth height y can be obtained based on the following formula: * ;

[0194]

[0195] Optionally, select the nth t Historical growth dataset of target domain in The first latent variable and the nth t indivual Input to the joint distribution sampling network C2 to get the mean and variance Then we get the latent variable distribution belonging to the target domain Based on the Bayesian rule and importance sampling, S second latent variables are obtained Further, we can obtain the prediction model based on The mean of the sum of all predicted values ​​based on S latent variables z′, that is, Obtained through the above steps Based on all target values The mean of the sum of all predicted values, and the final predicted value y is obtained based on the following formula * ;

[0196]

[0197] In an embodiment of the present invention, at least one historical growth data set and environmental characteristic data are used to obtain at least one initial predicted growth height corresponding to the at least one historical growth data set. The initial predicted growth height corresponds to the historical growth data set one-to-one. The at least one initial predicted growth height obtained is converted to obtain a target predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0198] According to an embodiment of the present invention, a plant growth height prediction device is also provided. It should be noted that the plant growth height prediction device of this embodiment can be used to execute the plant growth height prediction method of embodiment 1 of the present invention.

[0199] Figure 6 Schematic diagram of a plant growth height prediction device according to an embodiment of the present invention. Figure 6 As shown, the plant growth height prediction device 60 may include: a first acquisition unit 602 , a second acquisition unit 604 , a generation unit 608 and a conversion unit 610 .

[0200] The first acquisition unit 602 is used to acquire environmental characteristic data of the planting environment where the plant to be predicted is located, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted.

[0201] The second acquisition unit 604 is used to respectively acquire a historical growth data set of at least one grown plant in the planting environment to obtain at least one historical growth data set, wherein the at least one grown plant includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the grown plant.

[0202] The generating unit 608 is used to generate an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set.

[0203] The conversion unit 610 is used to convert at least one initial predicted growth height to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height.

[0204] The plant growth height prediction device of this embodiment obtains environmental characteristic data of the planting environment in which the plant to be predicted is located through a first acquisition unit, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted; obtains historical growth data sets of at least one grown plant in the planting environment through a second acquisition unit to obtain at least one historical growth data set, wherein at least one grown plant includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the grown plant; generates an initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set through a generation unit; converts at least one initial predicted growth height through a conversion unit to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height, thereby achieving the technical effect of improving the prediction accuracy of the plant growth height and solving the technical problem of low prediction accuracy of the plant growth height.

[0205] The embodiment of the present invention can provide a computer terminal, which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal can also be replaced by a terminal device such as a mobile terminal.

[0206] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0207] In this embodiment, the computer terminal can execute the above steps in the method for predicting the growth height of a plant.

[0208] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0209] According to an embodiment of the present invention, a computer-readable storage medium is further provided. The computer-readable storage medium includes a stored program, wherein the program executes the method for predicting the growth height of a plant in Example 1.

[0210] Optionally, in this embodiment, the computer-readable storage medium may be located in any one of the computer terminals in a computer terminal group in a computer network, or in any one of the mobile terminals in a mobile terminal group.

[0211] Optionally, in this embodiment, the computer-readable storage medium is configured to store program codes for executing the steps of the above-mentioned method for predicting the growth height of a plant.

[0212] According to an embodiment of the present invention, a processor is further provided, and the processor is used to run a program, wherein when the program is run by the processor, the method for predicting the growth height of a plant in Example 1 is executed.

[0213] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0214] In this embodiment, the computer terminal can execute the program code of the above steps of the method for predicting the growth height of a plant.

[0215] Wherein, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the multilingual translation method and device in the embodiment of the present invention, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned multilingual translation method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0216] According to an embodiment of the present invention, a computer program product is further provided. The computer program product includes computer instructions, wherein the computer instructions implement the method for predicting the growth height of a plant in Example 1 when executed by a processor.

[0217] An embodiment of the present application may provide an electronic device, which may include a memory and a processor.

[0218] Figure 7It is a block diagram of an electronic device according to a method for predicting the growth height of a plant according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only used as examples, and are not intended to limit the implementation of the present application described herein and / or required.

[0219] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 to a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0220] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 704, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0221] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as a verification method for data. For example, in some embodiments, the verification method for data may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the verification method for data described above may be executed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the data verification method in any other appropriate manner (for example, by means of firmware).

[0222] According to an embodiment of the present application, a method for predicting the growth height of a plant is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0223] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0224] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0225] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0226] To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or an LCD (liquid crystal display, monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a path ball) through which the user can provide input to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0227] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a Local Area Network (LAN), a Wide Area Network (WAN), and the Internet.

[0228] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0229] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0230] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0231] In the several embodiments provided by the present invention, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units can be a logical function division, and there can be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0232] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0233] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0234] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0235] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting the growth height of a plant, It is characterized in that include: Acquire environmental characteristic data of the planting environment where the plant to be predicted is located, wherein the environmental characteristic data is used to characterize the influencing factors in the planting environment that affect the growth height of the plant to be predicted; Respectively acquiring a historical growth data set of at least one plant that has grown in the planting environment to obtain at least one historical growth data set, wherein the at least one plant that has grown includes the plant to be predicted, and the historical growth data set is used to characterize the historical growth height of the plant that has grown; Based on the environmental characteristic data and the historical growth data set, generating an initial predicted growth height of the plant to be predicted; At least one of the initial predicted growth heights is converted to obtain a target predicted growth height, wherein the accuracy of the target predicted growth height is higher than the accuracy of the initial predicted growth height.

2. The method according to claim 1, It is characterized in that The step of generating the initial predicted growth height of the plant to be predicted based on the environmental characteristic data and the historical growth data set includes: Mapping the environmental feature data into a latent space to obtain a first latent variable; Based on the first latent variable and the historical growth data set, obtaining a latent variable distribution corresponding to the historical growth data set; Predict at least one latent variable distribution corresponding to at least one historical growth data set to obtain at least one initial predicted growth height.

3. The method according to claim 2, It is characterized in that The obtaining, based on the first latent variable and the historical growth data set, a latent variable distribution corresponding to the historical growth data set comprises: Controlling the distribution sampling network, transforming the first latent variable and the historical growth data set to obtain a first variance of the historical growth data set and a first mean of the historical growth data set; The latent variable distribution is determined based on the first variance and the first mean.

4. The method according to claim 2, It is characterized in that The step of respectively predicting at least one latent variable distribution corresponding to at least one historical growth data set to obtain at least one initial predicted growth height includes: Sampling the latent variable distribution to obtain a plurality of second latent variables, wherein the second latent variables are used to characterize the growth condition of the plant to be predicted under the planting environment; Predicting a plurality of the second latent variables respectively to obtain a plurality of predicted growth heights; The average of the multiple predicted growth heights is determined as the initial predicted growth height.

5. The method according to claim 4, It is characterized in that The method further comprises: Determine a target planting environment whose environmental similarity to the planting environment is greater than a similarity threshold; Constructing first training data of the planting environment and second training data of the target planting environment, wherein the first training data at least includes a first environmental characteristic data sample of the planting environment and a first growth data sample of the plant grown in the planting environment; the second training data at least includes a second environmental characteristic data sample of the target planting environment and a second growth data sample of the plant grown in the target planting environment; The distribution sampling network is trained using the first training data and the second training data.

6. The method according to claim 5, It is characterized in that The step of training the distributed sampling network using the first training data and the second training data includes: Inputting the first environmental feature data sample and the second environmental feature data sample into the first sub-distribution sampling network for analysis respectively, obtaining a third latent variable of the latent space mapped from the first environmental feature data sample and a fourth latent variable of the latent space mapped from the second environmental feature data sample; Inputting the third latent variable, the first growth data sample, the fourth latent variable, and the second growth data sample into a second sub-distribution sampling network for analysis, to obtain the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment; Based on the latent variable distribution of the planting environment and the latent variable distribution of the target planting environment, a first sub-objective equation is constructed; Based on the first sub-objective equation, determining a target model parameter of a sub-distribution sampling network, wherein the sub-distribution sampling network includes the first sub-distribution sampling network and the second sub-distribution sampling network; The target model parameters are used to update the initial model parameters of the sub-distribution sampling network to obtain a distribution sampling network.

7. The method according to claim 6, It is characterized in that The latent variable distribution based on the planting environment, And the hidden variable distribution of the target planting environment, the first sub-goal equation is constructed, including: Sampling the latent variable distribution of the planting environment to obtain a plurality of fifth latent variables; Sampling the latent variable distribution of the target planting environment to obtain a plurality of sixth latent variables; Based on the multiple fifth latent variables and the multiple sixth latent variables, the first sub-objective equation is constructed.

8. The method according to claim 6, It is characterized in that The step of respectively predicting the plurality of second latent variables to obtain a plurality of predicted growth heights includes: The prediction model is controlled to predict the second latent variable to obtain the predicted growth height, wherein the prediction model is trained based on the first training data and the second training data.

9. The method according to claim 8, It is characterized in that The method further comprises: Mapping the first environmental feature data sample to the latent space to obtain a plurality of seventh latent variables; Controlling the sub-prediction model of the target planting environment to predict a plurality of sixth latent variables corresponding to the target planting environment to obtain a first predicted growth height, and controlling the sub-prediction model corresponding to the planting environment to predict the seventh latent variable to obtain a second predicted growth height; Using the first predicted growth height and the second predicted growth height, construct a second sub-objective equation; Determining target model parameters of the sub-prediction model of the target planting environment and target model parameters of the sub-prediction model of the planting environment based on the second sub-target equation; Using the target model parameters, the initial model parameters of the sub-prediction model of the target planting environment and the initial model parameters of the sub-prediction model of the planting environment are updated respectively to obtain a prediction model, wherein the prediction model includes the adjusted sub-prediction model of the target planting environment and the adjusted sub-prediction model of the planting environment.

10. The method according to claim 9, It is characterized in that The method further comprises: Controlling the sub-reconstruction model corresponding to the target planting environment to reconstruct the sixth latent variable to obtain a first reconstruction result, and controlling the sub-reconstruction model corresponding to the planting environment to reconstruct the seventh latent variable to obtain a second reconstruction result; Based on the first reconstruction result and the second reconstruction result, a third sub-objective equation is constructed.

11. The method according to claim 10, It is characterized in that The step of determining target model parameters of the sub-distribution sampling network based on the first sub-target equation includes: Based on the first sub-goal equation, the second sub-goal equation and the third sub-goal equation, a goal equation is constructed; The target model parameters of the sub-distribution sampling network are determined using the target equation.

12. The method according to claim 10, It is characterized in that The step of determining the target model parameters of the sub-prediction model of the target planting environment and the target model parameters of the sub-prediction model of the planting environment based on the second sub-target equation includes: Based on the first sub-goal equation, the second sub-goal equation and the third sub-goal equation, a goal equation is constructed; The target model parameters of the sub-prediction model of the target planting environment and the target model parameters of the sub-prediction model of the planting environment are determined by using the target equation.

13. The method according to claim 1, It is characterized in that The method further comprises: Based on the target predicted growth height, the moment when the predicted plant affects the power transmission line in the planting environment is determined.

Images