A method, device, equipment and storage medium for predicting surface vegetation phenology information

Abstract

The present invention relates to the field of geographic information technology, and discloses a method, device, equipment and storage medium for predicting surface vegetation phenology information. The present invention more comprehensively reflects actual surface vegetation phenology information by fusing a historical remote sensing vegetation phenology information set and a historical multi-source climate environment information set. Furthermore, a surface vegetation phenology information prediction model based on an improved deep belief network is established in combination with a determined initial key factor set including geographic information elements, climate environment elements, remote sensing surface cover elements and model assimilation elements to predict surface vegetation phenology information, thereby improving the prediction accuracy of surface vegetation phenology information.

Description

Technical Field

[0001] The present invention relates to the field of geographic information technology, and in particular to a method, device, equipment and storage medium for predicting surface vegetation phenological information. Background Art

[0002] Phenology refers to the periodic changes in the growth, development, and activities of plants and animals under the influence of climate, environment, and other factors. As an important branch of phenology research, vegetation phenology mainly studies the natural phenomena of regular changes in germination, branching, leafing, flowering, fruiting, leaf fall, and dormancy in plants during their growth process of a year due to seasonal changes in climate and the influence of human activities. Vegetation phenology can regulate the physical properties of organisms in terrestrial ecosystems and their sensitivity to ecological and environmental changes. Changes in vegetation phenology usually cause changes in land surface albedo, surface energy balance, and air mobility, which in turn affect regional and even global carbon and water cycles, climate environment, and vegetation primary net productivity. In recent years, interannual changes in vegetation phenology have been widely used as a diagnostic indicator of the impact of climate change on ecosystem characteristics and processes, and have played a huge role in the research of ecological climate response, terrestrial energy balance, and carbon budget.

[0003] In the past few decades, satellite remote sensing has been proven to be a powerful supplement to surface phenology observations, and can be used to monitor and characterize the magnitude, trend, and characteristics of surface vegetation phenology changes from regional to global scales. However, since the extraction of remotely sensed vegetation phenology information is easily affected by observation errors, spatial scale, and interannual climate change, there are still some differences between remotely sensed vegetation phenology and measured surface vegetation phenology information, resulting in low accuracy in the prediction results of surface vegetation phenology information. Summary of the invention

[0004] In view of this, the present invention provides a method, device, equipment and storage medium for predicting surface vegetation phenological information to solve the problem that the extraction of remotely sensed vegetation phenological information is easily affected by observation errors, spatial scale and inter-annual climate change, and there is still a certain difference between remotely sensed vegetation phenology and the measured information of surface vegetation phenology, resulting in low accuracy of the prediction results of surface vegetation phenological information.

[0005] In a first aspect, the present invention provides a method for predicting surface vegetation phenology information, the method comprising:

[0006] Obtain historical remote sensing vegetation phenology information sets and historical multi-source climate and environmental information sets; based on the historical remote sensing vegetation phenology information sets and historical multi-source climate and environmental information sets, determine the initial key factor set including geographic information elements, climate and environmental elements, remote sensing surface cover elements and model assimilation elements, and the initial key factor set reflects the impact on vegetation phenology; based on the historical remote sensing vegetation phenology information sets, historical multi-source climate and environmental information sets and the initial key factor set, establish a surface vegetation phenology information prediction model based on an improved deep belief network; use the surface vegetation phenology information prediction model to predict the surface vegetation phenology information and obtain the surface vegetation phenology information prediction results.

[0007] The surface vegetation phenological information prediction method provided by the present invention more comprehensively reflects the actual surface vegetation phenological information by fusing the historical remote sensing vegetation phenological information set and the historical multi-source climate and environmental information set. Furthermore, a surface vegetation phenological information prediction model based on an improved deep belief network is established in combination with the determined initial key factor set including geographic information elements, climate and environmental elements, remote sensing surface cover elements and model assimilation elements to predict the surface vegetation phenological information, thereby improving the prediction accuracy of the surface vegetation phenological information.

[0008] In an optional implementation, based on a historical remote sensing vegetation phenology information set, a historical multi-source climate environment information set and an initial key factor set, a surface vegetation phenology information prediction model based on an improved deep belief network is established, including:

[0009] Based on the historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the initial key factor set, the preset deep confidence network is improved to obtain the target deep confidence network; based on the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set, after the target deep confidence network is trained, a surface vegetation phenology information prediction model based on the improved deep confidence network is established.

[0010] The surface vegetation phenology information prediction method provided by the present invention improves the preset deep belief network by combining the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, thereby improving the training effect of the target deep belief network. Furthermore, the improved target deep belief network is used to establish a corresponding surface vegetation phenology information prediction model, thereby improving the accuracy and stability of the model.

[0011] In an optional implementation, based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, the preset deep belief network is improved to obtain the target deep belief network, including:

[0012] Perform correlation analysis on the initial key factor set to obtain the target key factor set; input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the preset deep confidence network for training and determine the number of network layers and nodes; determine the target deep confidence network based on the number of network layers, the number of nodes and the preset deep confidence network.

[0013] The surface vegetation phenology information prediction method provided by the present invention can improve the training accuracy of the preset deep confidence network by analyzing the correlation of the initial key factor set reflecting the impact on vegetation phenology, and selecting the target key factor set with higher correlation to input into the preset deep confidence network for training. Furthermore, the number of network layers and nodes is determined by the training effect, and then the target deep confidence network with the best training effect can be determined according to the number of network layers and nodes, which provides support for subsequent improvement of the accuracy and stability of the model.

[0014] In an optional implementation, the preset deep belief network is composed of a multi-layer restricted Boltzmann machine, each layer of the restricted Boltzmann machine includes a display input layer and a hidden layer; the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set are input into the preset deep belief network for training and the number of network layers and nodes is determined, including:

[0015] The historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the target key factor set are input into the display input layer of the first restricted Boltzmann machine of the preset deep belief network; in the display input layer, the historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the target key factor set are used to determine the hidden layer of the first restricted Boltzmann machine; in the hidden layer, feature extraction is performed based on the historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the target key factor set to obtain feature extraction results; the feature extraction results are input into the display input layer of the restricted Boltzmann machine of the next layer, and training is repeated iteratively to obtain multiple training results; based on the multiple training results, the number of network layers and nodes is determined.

[0016] The surface vegetation phenology information prediction method provided by the present invention inputs the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the multi-layer restricted Boltzmann machine of the preset deep belief network for repeated iterations, and according to the multiple training results obtained, the number of network layers and the number of nodes of the preset deep belief network when the training effect is best can be determined.

[0017] In a second aspect, the present invention provides a device for predicting surface vegetation phenology information, the device comprising:

[0018] The acquisition module is used to obtain the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set; the determination module is used to determine the initial key factor set including geographic information elements, climate and environment elements, remote sensing surface cover elements and model assimilation elements based on the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set, and the initial key factor set reflects the impact on vegetation phenology; the establishment module is used to establish a surface vegetation phenology information prediction model based on the improved deep belief network based on the historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the initial key factor set; the prediction module is used to predict the surface vegetation phenology information using the surface vegetation phenology information prediction model to obtain the surface vegetation phenology information prediction result.

[0019] In an optional implementation, a module is established, including:

[0020] The improvement submodule is used to improve the preset deep confidence network based on the historical remote sensing vegetation phenology information set, the historical multi-source climate and environment information set and the initial key factor set to obtain the target deep confidence network; the establishment submodule is used to establish a surface vegetation phenology information prediction model based on the improved deep confidence network after training the target deep confidence network based on the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set.

[0021] In an optional implementation, the improving submodule includes:

[0022] The analysis unit is used to perform correlation analysis on the initial key factor set to obtain the target key factor set; the first determination unit is used to input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the preset deep confidence network for training and determine the number of network layers and nodes; the second determination unit is used to determine the target deep confidence network based on the number of network layers, the number of nodes and the preset deep confidence network.

[0023] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for predicting surface vegetation phenology information of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for predicting surface vegetation phenology information of the first aspect or any corresponding embodiment thereof.

[0025] In a fifth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for predicting surface vegetation phenology information of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 is a schematic diagram of the flow of a method for predicting surface vegetation phenological information according to an embodiment of the present invention;

[0028] Figure 2 is a schematic flow chart of another method for predicting surface vegetation phenological information according to an embodiment of the present invention;

[0029] Figure 3 is a target deep belief network according to an embodiment of the present invention;

[0030] Figure 4 is a schematic diagram of training accuracy results of a target deep belief network according to an embodiment of the present invention;

[0031] Figure 5A is a schematic diagram of a model prediction result based on reinforcement learning according to an embodiment of the present invention;

[0032] Figure 5B is a schematic diagram of prediction results of a multivariate regression model MLR according to an embodiment of the present invention;

[0033] Figure 5C is a schematic diagram of prediction results of the process model Chill&Force according to an embodiment of the present invention;

[0034] Figure 5D is a schematic diagram of model prediction results of a support vector machine (SVM) according to an embodiment of the present invention;

[0035] Figure 5E is a schematic diagram of model prediction results of a deep belief network DBN without adding an SVM model according to an embodiment of the present invention;

[0036] Fig. 5F is a schematic diagram of model prediction results of a surface vegetation phenology information prediction model based on an improved deep belief network according to an embodiment of the present invention;

[0037] Figure 6is a structural block diagram of a device for predicting surface vegetation phenology information according to an embodiment of the present invention;

[0038] Figure 7 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0040] The embodiment of the present invention provides a method for predicting surface vegetation phenological information, which more comprehensively reflects the actual surface vegetation phenological information by fusing a historical remote sensing vegetation phenological information set and a historical multi-source climate and environmental information set, and then combines a determined initial key factor set including geographic information elements, climate and environmental elements, remote sensing surface cover elements, and model assimilation elements to establish a surface vegetation phenological information prediction model based on an improved deep belief network, so as to achieve the effect of improving the prediction accuracy of surface vegetation phenological information.

[0041] According to an embodiment of the present invention, an embodiment of a method for predicting surface vegetation phenological information 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.

[0042] In this embodiment, a method for predicting surface vegetation phenology information is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 1 is a flow chart of a method for predicting surface vegetation phenology information according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0043] Step S101, obtaining a historical remote sensing vegetation phenology information set and a historical multi-source climate environment information set.

[0044] Specifically, the historical remote sensing vegetation phenology information set represents relevant information on the periodic changes of vegetation phenology under the influence of climate, environment and other factors obtained through satellite remote sensing observations.

[0045] Furthermore, the historical multi-source climate environment information set may include information such as temperature, longitude, rainfall, sudden temperature, accumulated temperature, etc.

[0046] Step S102, based on the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set, determine the initial key factor set including geographic information elements, climate environment elements, remote sensing surface cover elements and model assimilation elements.

[0047] Among them, the initial key factor set reflects the impact on vegetation phenology and can include multiple elements such as geographic information elements GEO, climate environment elements CLI, remote sensing surface cover elements RS and assimilation model elements REANALYSIS.

[0048] Furthermore, the geographic information element GEO is composed of longitude, latitude and elevation; the climate environment element CLI can include average temperature, rainfall, sudden temperature accumulation and accumulated temperature accumulation; the remote sensing surface cover element RS can be composed of forest area proportion, grassland area proportion and urban area proportion; the assimilation model element REANALYSIS can include soil moisture content and water evaporation.

[0049] Specifically, by combining the acquired historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set, the corresponding initial key factor set including geographic information elements, climate and environment elements, remote sensing surface cover elements and model assimilation elements can be determined.

[0050] Step S103, based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, a surface vegetation phenology information prediction model based on the improved deep belief network is established.

[0051] Specifically, the classic Deep Belief Network (DBN) can be regarded as a simple multi-layer unsupervised combination network, which is usually composed of a multi-layer restricted Boltzmann Machine (RBM) and a layer of back propagation (BP) neural network. Its essence is a specially constructed neural network. The training process of the deep belief network can be regarded as an effective unsupervised hierarchical method. As a Bayesian probability generation model, the deep belief network has not been used to predict surface vegetation phenology information.

[0052] Furthermore, this embodiment utilizes the advantages of deep belief network point and surface data fusion and feature learning, combines the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, and establishes a multi-source data fusion model based on improved deep belief network and support vector machine to predict the starting period of surface vegetation phenology.

[0053] Step S104, predicting the surface vegetation phenology information using the surface vegetation phenology information prediction model to obtain the surface vegetation phenology information prediction result.

[0054] Specifically, the established surface vegetation phenology information prediction model can realize real-time prediction of surface vegetation phenology information, thereby improving the prediction accuracy of surface vegetation phenology information.

[0055] The surface vegetation phenological information prediction method provided in this embodiment more comprehensively reflects the actual surface vegetation phenological information by fusing the historical remote sensing vegetation phenological information set and the historical multi-source climate and environmental information set. Furthermore, a surface vegetation phenological information prediction model based on an improved deep belief network is established in combination with the determined initial key factor set including geographic information elements, climate and environmental elements, remote sensing surface cover elements and model assimilation elements to predict the surface vegetation phenological information, thereby improving the prediction accuracy of the surface vegetation phenological information.

[0056] In this embodiment, a method for predicting surface vegetation phenology information is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 2 is a flow chart of a method for predicting surface vegetation phenology information according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0057] Step S201, obtain historical remote sensing vegetation phenology information set and historical multi-source climate environment information set. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0058] Step S202: Based on the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set, determine the initial key factor set including geographic information elements, climate environment elements, remote sensing land cover elements and model assimilation elements. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0059] Step S203, based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, a surface vegetation phenology information prediction model based on the improved deep belief network is established.

[0060] Specifically, the above step S203 includes:

[0061] Step S2031, based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, the preset deep belief network is improved to obtain the target deep belief network.

[0062] Specifically, by improving the preset deep belief network by combining the obtained historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, the training effect of the target deep belief network can be improved.

[0063] In some optional implementations, the above step S2031 includes:

[0064] Step a1: perform correlation analysis on the initial key factor set to obtain the target key factor set.

[0065] Step a2, input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the preset deep belief network for training and determine the number of network layers and nodes.

[0066] Specifically, the correlations among multiple elements included in the initial key factor set are analyzed, and factors with higher correlations are selected as the target key factor set.

[0067] Furthermore, the preset deep belief network can be composed of multiple RBMs, and each RBM layer also contains two layers of network, namely, the visible input layer and the hidden layer.

[0068] In this embodiment, the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set are input into a preset deep belief network for training, and the corresponding number of network layers and nodes are determined according to the training effect.

[0069] In some optional implementations, the above step a2 includes:

[0070] Step a21, inputting the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the display input layer of the first layer restricted Boltzmann machine of the preset deep belief network.

[0071] Step a22, at the display input layer, the hidden layer of the first layer of restricted Boltzmann machine is determined using the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set.

[0072] Step a23, in the hidden layer, feature extraction is performed based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set to obtain feature extraction results.

[0073] Step a24, input the feature extraction result into the display input layer of the restricted Boltzmann machine of the next layer, repeatedly iterate the training, and obtain multiple training results.

[0074] Step a25, determining the number of network layers and nodes based on multiple training results.

[0075] Specifically, in the preset deep belief network, the explicit input layer and hidden layer of each RBM layer are used to input training data, and the hidden layer is used for feature extraction. The training of the deep belief network in the RBM layer is progressive layer by layer. In each layer, the hidden layer is inferred by the input data, and then the output result of the hidden layer is input to the next layer, that is, the explicit input layer in the next layer of RBM is the implicit output layer in the previous layer of RBM. The training process of RBM is actually the process of solving the maximum probability distribution of training samples, and the core factor for solving the sample distribution lies in the matrix composed of network weights.

[0076] Furthermore, assuming that there are visible neurons and hidden neurons in an RBM, and neurons can be connected without direction between two layers, while neurons in the same layer are not connected, then the weight change ΔW of the RBM during training is ij It can be calculated using the following relationship (1):

[0077]

[0078] Where: ε represents the learning rate; Represents the expectation of the distribution of training samples; represents the expected partial derivative of the reconstructed model.

[0079] Therefore, according to the above process, the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set are sequentially input into each layer of the RBM for training, and the number of network layers and nodes when the training effect is the best are taken as the number of network layers and nodes of the final deep belief network.

[0080] In the training process, the logistic regression model is selected as the activation function in each layer of RBM. At the same time, the mean square error (MSE) indicator is used as the loss function, as shown in the following relationship (2), to optimize the prediction of the start period of surface vegetation phenology sos and the start period of surface vegetation phenology The differences:

[0081]

[0082] Furthermore, in the present implementation, during the above training process, it was found that the best training effect was achieved when a two-layer RBM network (with 10 and 5 nodes, respectively) was established, that is, the number of network layers and the number of nodes were 10 and 5, respectively.

[0083] Step a3, determining the target deep belief network based on the number of network layers, the number of nodes and the preset deep belief network.

[0084] Specifically, the number of network layers and nodes of the preset deep belief network is modified to the final number of network layers and nodes determined to obtain the corresponding target deep belief network, such as Figure 3 shown.

[0085] Furthermore, the target deep belief network can not only avoid the overfitting phenomenon caused by using a multi-layer network structure, but also accelerate the network training speed.

[0086] Step S2032, based on the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set, after training with the target deep belief network, a surface vegetation phenology information prediction model based on the improved deep belief network is established.

[0087] Specifically, the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set are input into the improved good target deep belief network for training, and the corresponding surface vegetation phenology information prediction model can be trained.

[0088] During the training process, about 70% of the samples are randomly selected for training in each network iteration, and the remaining 30% of the samples are used to test the rationality of the model.

[0089] Furthermore, during the training process, the loss function shown in the above equation (2) is also regularized to prevent overfitting of the model training. The model training results are shown in Figure 4 shown.

[0090] Furthermore, the deep belief network can quickly converge the training results through the unsupervised RBM network, and then fine-tune the network through the traditional global learning algorithm, so that the model converges to the local optimum. The surface vegetation phenology information prediction model based on the improved deep belief network established in this embodiment can be expressed as the following relationship (3):

[0091]

[0092] In the formula: Model represents the nonlinear estimation function required for model construction; SOSobs represents the measured data of the starting period of surface vegetation phenology; SOSlsp represents the starting period of remote sensing vegetation phenology; Temp represents temperature; Lon represents longitude; Lat represents latitude; Ele represents elevation; Prcp represents rainfall; Force represents shock temperature; Ch represents accumulated temperature; Soilwat represents soil moisture content; Evap represents water evaporation; Forest represents the proportion of forest area; Grass represents the proportion of grassland area.

[0093] Furthermore, in an example, in order to verify the prediction accuracy of the surface vegetation phenology information prediction model based on the improved deep belief network proposed in the above embodiment of the present invention, in addition to the remote sensing vegetation phenology start period, the reinforcement learning model LSP, the multivariate regression model MLR, the process model Chill&Force, the support vector machine SVM and the deep belief network DBN without adding the SVM model were selected as auxiliary models to participate in the experimental comparison. In the process of accuracy evaluation, five indicators, CC, MAE, SAM, RMSE and ERGAS, were used. The model prediction results are as follows: FIG. 5A to FIG. 5F As shown. Wherein, Obesrvations represents observations; Prediction represents prediction results; Proposed represents the surface vegetation phenology information prediction model based on the improved deep belief network proposed in the embodiment of the present invention; Lsp represents reinforcement learning.

[0094] Step S204: Use the surface vegetation phenology information prediction model to predict the surface vegetation phenology information and obtain the surface vegetation phenology information prediction result. Figure 1 Step S104 of the illustrated embodiment will not be described in detail here.

[0095] The surface vegetation phenology information prediction method provided in this embodiment is to input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the multi-layer restricted Boltzmann machine of the preset deep belief network for repeated iterations, and the network layers and nodes of the preset deep belief network with the best training effect can be determined based on the multiple training results obtained, and then the target deep belief network with the best training effect can be determined based on the network layers and nodes, thereby improving the training effect and accuracy of the target deep belief network. Finally, the corresponding surface vegetation phenology information prediction model is established using the improved target deep belief network, thereby improving the accuracy and stability of the model.

[0096] In one example, in order to better understand the actual contribution of geographical and climatic environmental factors in the surface vegetation phenology information prediction model proposed in the above embodiment of the present invention, in addition to retaining the key element remote sensing vegetation phenology start period SOS, the remaining factors are gradually introduced into the surface vegetation phenology information prediction model proposed in the above embodiment of the present invention in different combinations, and the prediction results of the surface vegetation phenology start period are calculated respectively. The remaining factors can be mainly divided into four categories: 1. Geographic information element GEO, composed of longitude, latitude and elevation; 2. Climate environment element CLI, including average temperature, rainfall, shock temperature accumulation and accumulated temperature accumulation; 3. Remote sensing surface cover element RS, composed of forest area proportion, grassland area proportion and urban area proportion; 4. Assimilation model element REANALYSIS, including soil moisture content and water evaporation. The experimental results are shown in Table 1.

[0097] Table 1. Model prediction accuracy under different combinations of factors

[0098] Combination of elements CC RMSE MAE SOS 0.827 8.003 6.050 SOS+GEO 0.804 8.007 6.054 SOS+RS 0.872 8.001 6.048 SOS+CLI 0.838 8.034 6.077 SOS+REANALYSIS 0.814 8.002 6.048 SOS+GEO+RS 0.880 7.996 6.065 SOS+GEO+CLI 0.825 9.214 7.190 SOS+GEO+REANALYSIS 0.818 8.002 6.049 SOS+RS+CLI 0.862 7.996 6.045 SOS+RS+REANALYSIS 0.860 7.998 6.050 SOS+CLI+REANALYSIS 0.830 8.677 6.700 SOS+GEO+RS+CLI 0.851 8.609 6.537 SOS+GEO+RS+REANALYSIS 0.880 7.996 6.065 SOS+RS+CLI+REANALYSIS 0.869 7.999 6.046 SOS+GEO+CLI+REANALYSIS 0.832 8.004 6.050 ALL 0.843 8.005 6.051

[0099] Therefore, after comprehensively considering the data quality, the correlation between elements and the model training efficiency, SOS+GEO+RS can be selected as the optimal combination of model elements to ensure high-precision vegetation phenology prediction results, that is, SOS+GEO+RS is used as the target key factor set.

[0100] In this embodiment, a surface vegetation phenology information prediction device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0101] This embodiment provides a device for predicting surface vegetation phenology information. Figure 6 As shown, the device comprises:

[0102] The acquisition module 601 is used to acquire a historical remote sensing vegetation phenology information set and a historical multi-source climate environment information set.

[0103] Determination module 602 is used to determine the initial key factor set including geographic information elements, climate and environment elements, remote sensing surface cover elements and model assimilation elements based on the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set, and the initial key factor set reflects the impact on vegetation phenology.

[0104] Establishing module 603, for establishing a surface vegetation phenology information prediction model based on an improved deep belief network based on a historical remote sensing vegetation phenology information set, a historical multi-source climate environment information set and an initial key factor set.

[0105] The prediction module 604 is used to predict the surface vegetation phenology information using the surface vegetation phenology information prediction model to obtain the surface vegetation phenology information prediction result.

[0106] In some optional implementations, the establishing module 603 includes:

[0107] The improvement submodule is used to improve the preset deep confidence network based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set to obtain the target deep confidence network.

[0108] A submodule is established to establish a surface vegetation phenology information prediction model based on an improved deep confidence network based on historical remote sensing vegetation phenology information sets and historical multi-source climate environment information sets after training with a target deep confidence network.

[0109] In some optional implementations, the improved submodule includes:

[0110] An analysis unit, used for performing correlation analysis on the key factor set to obtain a target key factor set;

[0111] The first determination unit is used to input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into a preset deep belief network for training and determine the number of network layers and nodes;

[0112] The second determining unit is used to determine the target deep confidence network based on the number of network layers, the number of nodes and the preset deep confidence network.

[0113] In some optional embodiments, the preset deep belief network is composed of multiple layers of restricted Boltzmann machines, each layer of restricted Boltzmann machines includes a display input layer and a hidden layer; the first determination unit includes:

[0114] The input subunit is used to input the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the display input layer of the first layer restricted Boltzmann machine of the preset deep belief network.

[0115] The first determination subunit is used to determine the hidden layer of the first restricted Boltzmann machine in the display input layer by using the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set.

[0116] The feature extraction subunit is used to extract features in the hidden layer based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set to obtain feature extraction results.

[0117] The iterative training subunit is used to input the feature extraction result into the display input layer of the restricted Boltzmann machine of the next layer, and repeatedly iterate the training to obtain multiple training results.

[0118] The second determination subunit is used to determine the number of network layers and nodes based on multiple training results.

[0119] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0120] The surface vegetation phenological information prediction device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0121] The embodiment of the present invention also provides a computer device having the above Figure 6 The surface vegetation phenological information prediction device shown.

[0122] See also Figure 7 , Figure 7 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 7 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 7 A processor 10 is taken as an example.

[0123] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0124] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0125] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device 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.

[0126] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0127] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0128] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0129] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0130] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for predicting surface vegetation phenology information, characterized in that: The method comprises: Obtain historical remote sensing vegetation phenology information sets and historical multi-source climate and environment information sets; Based on the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set, determining an initial key factor set including geographic information elements, climate environment elements, remote sensing land cover elements and model assimilation elements, wherein the initial key factor set reflects the impact on vegetation phenology; Based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, a surface vegetation phenology information prediction model based on an improved deep belief network is established; Predicting the surface vegetation phenology information using the surface vegetation phenology information prediction model to obtain a surface vegetation phenology information prediction result; Among them, based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set, a surface vegetation phenology information prediction model based on an improved deep belief network is established, including: Performing correlation analysis on the initial key factor set to obtain a target key factor set; Inputting the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set into the display input layer of the first layer of restricted Boltzmann machine of the preset deep belief network, wherein the preset deep belief network is composed of multiple layers of restricted Boltzmann machines, and each layer of restricted Boltzmann machine includes a display input layer and a hidden layer; In the display input layer, the hidden layer of the first layer restricted Boltzmann machine is determined using the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set; In the hidden layer, feature extraction is performed based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the target key factor set to obtain a feature extraction result; Inputting the feature extraction result into the display input layer of the restricted Boltzmann machine of the next layer, repeatedly iterating the training, and obtaining multiple training results; Based on the multiple training results, determine the number of network layers and the number of nodes; Determining a target deep confidence network based on the number of network layers, the number of nodes, and the preset deep confidence network; Based on the historical remote sensing vegetation phenology information set and the historical multi-source climate environment information set, after the target deep belief network training, the surface vegetation phenology information prediction model based on the improved deep belief network is established.

2. A device for predicting surface vegetation phenology information, characterized in that: Used to execute the method for predicting surface vegetation phenology information according to claim 1; the device comprises: An acquisition module is used to obtain historical remote sensing vegetation phenology information sets and historical multi-source climate environment information sets; A determination module, for determining an initial key factor set including geographic information elements, climate and environment elements, remote sensing land cover elements and model assimilation elements based on the historical remote sensing vegetation phenology information set and the historical multi-source climate and environment information set, wherein the initial key factor set reflects the impact on vegetation phenology; Establishing a module for establishing a surface vegetation phenology information prediction model based on an improved deep belief network based on the historical remote sensing vegetation phenology information set, the historical multi-source climate environment information set and the initial key factor set; The prediction module is used to predict the surface vegetation phenology information using the surface vegetation phenology information prediction model to obtain the surface vegetation phenology information prediction result.

3. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for predicting surface vegetation phenology information according to claim 1 by executing the computer instructions.

4. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for predicting surface vegetation phenological information according to claim 1.

5. A computer program product, characterized in that It includes computer instructions, and the computer instructions are used to enable a computer to execute the method for predicting surface vegetation phenological information as described in claim 1.

Images