JP2021197084A - Server, forest data utilization system, forest data utilization method and program - Google Patents

Abstract

To provide a utilization technique for forest data that can be applied to a variety of large forests in all regions of Japan and that can be quantitatively evaluated in terms of a value of a forest.SOLUTION: A server includes: a storage unit that stores, in a region ID for identifying each region on a map, data such as a forest condition, a weather condition, a ground condition, an artificial condition, an economic condition, a system condition in the area and forest space sensing data, in association with one another; a calculation unit that, on the basis of the data stored in the storage unit, calculates an estimated value, concerning a quantitative value of a forest in the region on the map; an acquisition unit that acquires an actual measurement value of the quantitative value of the forest in the region; the storage unit that stores as learning data, the estimated value calculated by the calculation unit and the actual measurement value acquired by the acquisition unit; a model creating unit that, on the basis of the learning data, creates a learning model in which a relation between the estimated value and the actual measurement value is learned; an evaluation result providing unit that, on the basis of, the created learning model, calculates and provides the quantitative value of the forest in the specified region.SELECTED DRAWING: Figure 1

Description

The present invention relates to a technique for quantitatively grasping the value of a forest by using forest data by artificial intelligence (AI).

In the past, there have been case studies of software and databases limited to a specific area for forests, but there has never been an example of building a platform and database applicable to the entire Japan with a unified theory and scheme. There was a need for a system that could link with the input / output data of users in all areas of the forest, which occupies 70% of the total.

Also, in AI analysis, problems can be solved in a closed and formulated world such as Go and Shogi, but problems are not closed when elements intervene infinitely, so a solution is created in that. It was difficult. There was a need for a system that could evolve this AI analysis to recognize rules without deciding on its own category, or to determine what data was missing and what data was needed for better performance. ..
Furthermore, there is no example of AI with physicality that can be applied to diverse and vast forests throughout Japan, and unlike conventional specialized AI, AI with physicality itself acquires big data. By automating the system, it has been required to establish a general-purpose AI that enables autonomous improvement in a wide area.

As a related technology, it is a damage classification calculation method for pine worms that calculates the number of red pine trees by damage classification using the red pine forest image data created based on the spectral reflection characteristics of the tree species included in the photographed image data. A step to create a red pine forest tree apex image data, a step to create a red pine forest canopy image data, a step to create a damage classification image data by tree canopy, a step to create a damage classification image data by tree apex, and a red pine tree. By having a step to calculate the number of trees for each damage category and creating a damage rate classification map, it is possible to create materials showing the damage status of pine worms in Japanese red pine forests at low cost in a short time and with high accuracy. (See Patent Document 1).

JP-A-2017-163934

However, the technology described in Patent Document 1 utilizes AI applicable to diverse and vast forests throughout Japan, and quantitatively evaluates the value of forests while linking with input / output data of users in all regions. There was a problem that it was difficult to evaluate.

An object of the present invention is to provide a technique for utilizing forest data that can quantitatively evaluate the value of a forest, which can be applied to various and vast forests throughout Japan.

The first server of the present invention is
A storage unit that stores forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition, and forest space sensing data in the area ID that identifies each area on the map in association with each other. ,
A calculation unit that calculates a predicted value based on the data stored in the storage unit for the quantitative value of the forest in the area on the map.
An acquisition unit that acquires the measured value of the quantitative value of the forest in the region, and
The storage unit that stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit as learning data, and the storage unit.
A model generation unit that generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
It is provided with an evaluation result providing unit that calculates and provides the quantitative value of the forest in the designated area based on the generated learning model.

Further, the second server of the present invention is
Relationship between the measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions, and forest space sensing data data in each area on the map and the measured values of the quantitative value of forests in the area. A model generator that generates a learning model that has been trained in
The quantitative value of the forest in the designated area of the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition and the forest space sensing data to be evaluated. It is provided with an evaluation result providing unit that is calculated and provided based on the generated learning model from the measured values of the data.

Further, the first forest data utilization system of the present invention is
A forest data utilization system in which one or more user terminals, one or more information collection devices, and a server are connected via a predetermined network.
The server
A storage unit that stores forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition, and forest space sensing data in the area ID that identifies each area on the map in association with each other. ,
A calculation unit that calculates a predicted value based on the data stored in the storage unit for the quantitative value of the forest in the area on the map.
An acquisition unit that acquires an actually measured value of the quantitative value of the forest in the region from at least one of the user terminal and the information collecting device.
The storage unit that stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit as learning data, and the storage unit.
A model generation unit that generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
A communication unit that receives an inquiry about the quantitative value of the forest in the designated area on the map from the user terminal.
It is provided with an evaluation result providing unit that calculates the quantitative value of the forest in the designated area on the map based on the generated learning model and provides it to the user terminal.

Further, the second forest data utilization system of the present invention is
A forest data utilization system in which one or more user terminals, one or more information collection devices, and a server are connected via a predetermined network.
The server
Measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions, and forest space sensing data in each area on the map from at least one of the user terminal and the information collecting device, and the above. An acquisition unit that acquires the measured value of the quantitative value of the forest in the area, and
A model generator that generates a learning model that learns the relationship between the measured value of the data and the measured value of the quantitative value of the forest.
A communication unit that receives an inquiry about the quantitative value of the forest in the designated area on the map from the user terminal.
The quantitative value of the forest in the designated area on the map is evaluated by the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition and the forest to be evaluated. It is provided with an evaluation result providing unit that calculates and provides the spatial sensing data based on the generated learning model from the measured values of the data.

Further, the first method of using forest data of the present invention is
It is a method of using forest data using one or more user terminals connected via a predetermined network, and one or more information collecting devices and servers.
Regarding the quantitative value of the forest in each area on the map, the calculation unit of the server uses the area ID for identifying each area as the forest condition, meteorological condition, geological condition, artificial condition, and economic condition in each area. , The process of calculating the predicted value based on the data stored in the storage unit that stores the data of the system conditions and the forest space sensing data in association with each other.
A step in which the acquisition unit of the server acquires an actually measured value of the quantitative value of the forest in the region from at least one of the user terminal and the information collecting device.
A step in which the server stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit in the storage unit as learning data.
A process in which the model generation unit of the server generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
A step of the user terminal designating the specific area on the map and transmitting the quantitative value inquiry of the forest to the server.
The process in which the communication unit of the server receives the inquiry from the user terminal, and
A step in which the evaluation result providing unit of the server calculates the quantitative value of the forest in the designated area on the map based on the generated learning model and provides it to the user terminal. ,
The user terminal has a step of displaying the quantitative value of the forest in the designated area on the map provided by the server.

In addition, the second method of using forest data of the present invention is
It is a method of using forest data using one or more user terminals connected via a predetermined network, and one or more information collecting devices and servers.
From at least one of the user terminal and the information gathering device, the acquisition unit of the server can obtain forest conditions, meteorological conditions, geological conditions, artificial conditions, economic conditions, institutional conditions, and forest space sensing data in each area on the map. The step of acquiring the measured value of the data and the measured value of the quantitative value of the forest in the region, and
A process in which the model generation unit of the server generates a learning model that learns the relationship between the measured value of the data and the measured value of the quantitative value of the forest.
A step of the user terminal designating the specific area on the map and transmitting the quantitative value inquiry of the forest to the server.
The process in which the communication unit of the server receives the inquiry from the user terminal, and
The evaluation result providing unit of the server evaluates the quantitative value of the forest in the designated area on the map by the forest condition, the meteorological condition, the ground condition, the artificial condition, and the above. A process of calculating and providing based on the learning model generated from the measured values of the data of the economic conditions, the institutional conditions, and the forest space sensing data.
The user terminal has a step of displaying the quantitative value of the forest in the designated area on the map provided by the server.

Further, the first program of the present invention is
For information processing equipment
According to the calculation unit, regarding the quantitative value of the forest in each area on the map, the area ID that identifies the area is the forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition and The process of calculating the predicted value based on the data stored in the storage unit that stores the forest space sensing data in association with each other.
The process of acquiring the measured value of the quantitative value of the forest in the area by the acquisition unit, and
The predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit stored in the storage unit by the model generation unit are used as learning data, and the relationship between the predicted value and the measured value is obtained. The process of generating the learned learning model and
The evaluation result providing unit executes a process of calculating and providing the quantitative value of the forest in the designated area based on the generated learning model.

Further, the second program of the present invention is
For information processing equipment
By the model generation unit, the measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions and forest space sensing data in each area on the map, and the quantitative value of the forest in the area. Processing to generate a learning model that learned the relationship with the measured value,
The evaluation result providing unit determines the quantitative value of the forest in the designated area as the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition, and the evaluation target. The process of calculating and providing the forest space sensing data based on the generated learning model from the measured values of the data is executed.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide a technique for using forest data that can quantitatively evaluate the value of a forest, which can be applied to various and vast forests throughout Japan.

It is a figure which shows the structure of the forest data utilization system which concerns on embodiment of this invention. It is a functional block diagram of a user terminal and a server constituting the forest data utilization system which concerns on embodiment of this invention. It is a flowchart explaining the flow of the process which provides the quantitative value of a forest based on the learning model of the forest data utilization system which concerns on embodiment of this invention. It is a figure which shows the output example in the forest data utilization system which concerns on embodiment of this invention. It is a conceptual diagram explaining an example of the forest data utilization system which concerns on embodiment of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

〔System configuration〕
The configuration of the forest data utilization system according to the embodiment of the present invention will be described with reference to FIG.

As shown in FIG. 1, the forest data utilization system 10 includes one or more user terminals 100, one or more information collection devices 200, and a server 300.
Each is connected via a predetermined network such as the Internet. The predetermined network may be a LAN (Local Area Network) or the like, and may be wired or wireless.

The user terminal 100 is an information processing device including storage means such as ROM (Read Only Memory) and RAM (Random Access Memory) controlled by a CPU (Central Processing Unite), and is provided with known input / output means. For example, smartphones, small personal computers, mobile terminals and the like. The service (software) provided by the server 300 may be available by the cloud (SaaS / ASP).

The information collecting device 200 is a device provided with a sensor (camera mechanism, voice recorder, etc.) for sensing, and is, for example, a drone, a space satellite, a forestry machine attached to a sensor, or the like. The collected information may be transmitted to the user terminal 100 or the server 300, or an instruction / command from the user terminal 100 or the server 300 may be received and operated according to the instruction / command to collect the information.

The server 300 is an information processing apparatus including a storage means such as a ROM (Read Only Memory) and a RAM (Random Access Memory) controlled by a CPU (Central Processing Unite) and a known input / output means. For example, workstations, high-performance personal computers, etc.

Each device constituting the forest data utilization system 10 is mainly an information processing device, and various known techniques may be applied to these detailed configurations, and description other than the above will be omitted here.

FIG. 2 shows the functional blocks of the user terminal 100 and the server 300 that constitute the forest data utilization system 10 according to the present embodiment.

Referring to FIG. 2, the user terminal 100 includes a communication unit 110, an input unit 120, an output unit 130, a storage unit 140, a sensor unit 150, and a control unit 160. Each part may be realized by a known input / output I / F, a CPU, or the like, or by an HDD or the like.

The communication unit 110 has a function of communicating with each device connected to a predetermined network.

The input unit 120 has a function of receiving an input based on a user operation, for example, a touch panel, a microphone, or the like. The sensing data detected by the sensor unit 150, which will be described later, may be input from the user. For example, the user can check the map information displayed on the touch panel and tap any point on the map to specify a specific point on the map, and the point on the map (latitude and longitude). The area ID information is acquired by referring to a predetermined table or the like in which the minute section including the information) and the corresponding area ID are associated with each other (either the user terminal 100 side or the server 300 side performs the processing). It may be.) It may be.

The output unit 130 has a function of outputting various data such as image data and audio data by, for example, a display mechanism such as a display or an audio output mechanism such as a speaker.

The storage unit 140 has a function of storing a program for operating the information processing device and various data.

The sensor unit 150 may acquire the location position of the user terminal 100, which is various devices for detecting various states of the user terminal 100, by the GPS function.

The control unit 160 has a function of executing various types of information processing. Data transmission / reception processing with the server device 300 is executed.

Referring to FIG. 2, the server 300 includes a communication unit 301, an acquisition unit 302, a calculation unit 303, a learning data storage unit 304, a model generation unit 305, an evaluation result providing unit 306, and an output unit 307. It has a storage unit 308. Each part may be realized by a known input / output I / F, a CPU, or the like, or by an HDD or the like. Some functions of the server 300 may be realized as predetermined software, and the software may be cloud-based.

The communication unit 301 has a function of communicating with each device connected to a predetermined network. The communication unit 301 receives from the user terminal 100 an inquiry or the like that specifies a point on the map to be evaluated for the quantitative value of the forest.

The acquisition unit 302 has a function of acquiring observation data (actual data) actually measured from the user terminal 100 and the information collecting device 200 via the communication unit 301. The observation data (actual data) includes the actual data of the parameter part of the calculation and the actual data of the quantitative value of the forest obtained as a result of the calculation.

The calculation unit 303 has a function of calculating the estimated value before correction. Forecast estimation data (before correction) is calculated by a known method. For example, the profit amount is calculated by deducting the value of the cost item from the value of the profit item, or a predetermined forecast that predicts the growth rate (r: annual rate) of the tree according to the thinning pattern when calculating the profit. Formula (r = F (t) + ρB (N, D), t: tree age (or forest age), N: number density, D: thickness (or average thickness), ρ: correction coefficient by thinning pattern), etc. It is possible to calculate in consideration. The calculation may be performed using data already acquired throughout Japan as initial data stored in the database, or observation / collected actual data added to the database from the user terminal 100 or the information collecting device 200.

The learning data storage unit 304 stores the estimated data calculated by the calculation unit 303 and the actually measured actual data acquired by the acquisition unit 302 as learning data (teacher data) in the storage unit 308.
Further, the learning data storage unit 304 is the actual data actually measured by the acquisition unit 302, which is the actual data of the parameter part of the calculation, and the actual data of the calculation result (quantitative value of the forest). It may be stored in the storage unit 308 as learning data (teacher data) in pairs with the data.

The model generation unit 305 uses an artificial intelligence (AI) module to learn the relationship between the estimated data calculated by the calculation unit 303 and the actual measured data acquired by the acquisition unit 302 based on the learning data. Generate and store in the storage unit 308. Specifically, the model generation unit 305 machine-learns the relationship between the calculated estimated data and the actually measured data by regression analysis or the like. As a method of regression, a known method such as linear regression, polynomial regression or logistic regression can be used, and therefore detailed description thereof will be omitted.
Further, the model generation unit 305 is the actual data actually measured by the acquisition unit 302 based on the learning data (teacher data), which is the actual data of the parameter part of the calculation, and the calculation result (forest). It may be possible to generate a learning model that learns the relationship with the actual data (quantitative value of) and store it in the storage unit 308. Specifically, the model generation unit 305 may perform machine learning by deep learning (deep learning) using a neural network having a multi-layer structure based on a large amount of learning data.
Further, the model generation unit 305 may generate a learning model by other known machine learning or the like.

The evaluation result providing unit 306 has a function of calculating and providing the result of quantitative evaluation of the value such as the economic value of the forest. Specifically, the estimated value before correction calculated by the calculation unit 303 is corrected by the learning model based on the area ID corresponding to the specific point on the map inquired by the user terminal 100, and the evaluation result data (correction). It is provided as a later estimated value / predicted value). It should be noted that the estimated value (before correction) such as the growth rate of the tree calculated by the calculation unit 303 may be provided without correction as it is.
Further, the evaluation result providing unit 306 is actually measured data to be evaluated by the acquisition unit 302 based on the area ID corresponding to the specific point to be evaluated on the map inquired by the user terminal 100. Even if it is to provide evaluation result data (estimated value / predicted value) calculated by a learning model by deep learning using a neural network based on the actual data of the parameter part of the calculation. good.

The output unit 307 has, for example, a function of outputting various data such as image data and audio data by a display mechanism such as a display or an audio output mechanism such as a speaker.

The storage unit 308 has a function of storing a program for operating the information processing device and various data. The storage unit 308 may be realized as a storage unit by encryption by a blockchain. The database has a predetermined table or the like in which various data are associated with and stored in the area ID or the like of the minute section. The database will be described later. Further, a predetermined table or the like in which the minute section corresponding to the point (latitude and longitude information) on the map and the above area ID are associated with each other may be stored.

[Processing operation of learning model]
Next, two basic (general-purpose) processing operations for providing the quantitative value of the forest based on the learning model of the forest data utilization system 10 according to the present embodiment are shown in FIG. 3 (a). This will be described with reference to the flowchart of (b).

First, referring to FIG. 3A, first, the communication unit 301 of the server 300 receives an inquiry of a predetermined value of the forest at a specific point from the user terminal 100 (step S301). It may be possible to specify a specific point on the map by tapping the screen from any point on the Japanese map displayed on the user terminal 100. Of course, it may be specified by selecting a specific area from a pull-down list or the like. Then, the area ID information is acquired by referring to a predetermined table or the like in which the point (latitude and longitude information) on the map and the corresponding area ID are associated with each other.

Next, the evaluation result providing unit 306 causes the calculation unit 303 to solve the evaluation problem of the value of the forest (the type of value specified by the user) in the area ID corresponding to the point designated by the user, and before the correction. The estimated value of is calculated (step S302).

Next, the evaluation result providing unit 306 corrects the calculated estimated value by the learning model generated by the model generation unit 305 (step S303).

Then, the evaluation result providing unit 306 provides the corrected value as the evaluation result data (corrected estimated value / predicted value) of the value of the forest (step S304).

Next, referring to FIG. 3B, as in step S301 above, first, the communication unit 301 of the server 300 receives an inquiry about the predetermined value of the forest at a specific point from the user terminal 100 (step S301). ´).

Next, the evaluation result providing unit 306 is a parameter part of the calculation acquired by the acquisition unit 302 for the evaluation problem of the forest value (value of the kind specified by the user) in the area ID corresponding to the point designated by the user. Based on the actual data of, the estimated value is calculated by solving with a learning model or the like by deep learning using the neural network generated by the model generation unit 305 (step S302').

Then, the evaluation result providing unit 306 provides the calculated value as the evaluation result data (estimated value / predicted value) of the value of the forest (step S303').
The two basic (general-purpose) processing operations related to the learning model for providing the quantitative value of the forest have been described above. Even if it is not explicitly stated, the processing operation of the learning model can be applied in parallel or in an alternative manner in either of the two, and therefore the repeated description will be omitted in individual parts.

[Database]
In order to make the UI (user interface) based on the map of the whole of Japan, the forests of the whole of Japan are divided into areas of "10m x 10m" to "50m x 50m" in a mesh shape, and the minute points are 100 million to 2.5 billion. Positioned as a collection of partitions, it forms a database platform for big data.
Map information is composed of mesh units separated into rectangles at regular latitude / longitude intervals. Further, each mesh is composed of a plurality of layers having different scales separated by a predetermined unit. In the case of Japan, the mesh may be, for example, adopting the standard of the standard area mesh defined by the Ministry of Internal Affairs and Communications. The standard area mesh is composed of a primary mesh, a secondary mesh, and a tertiary mesh in the order of about 1/10 of the area ratio. Further, as the mesh, a divided area mesh subdivided from the primary to tertiary mesh may be adopted as a mesh unit. When the map information is divided into mesh units, the map information has an area ID and corresponding latitude and longitude information, respectively.
Area IDs for identifying small partitions as each area are assigned to the rows of the table stored in the database.
The data stored in the columns of the table in the database is used to quantitatively evaluate the value of the forest, that is, there is some correlation with the value of the forest in view of common technical knowledge, and the value of the forest. Data that can be used as parameters for quantitative evaluation (calculation) of, for example, forest conditions (tree species, number (density), thickness (trunk diameter), height (tree height)) in each region. , Tree age, tree position, etc.), Meteorological conditions (monthly temperature, precipitation, sunshine time, snowfall, wind speed, wind direction, maximum instantaneous wind speed, etc.), geological conditions (latitude and longitude, altitude, terrain slope, position, curvature) , Valley density, water collection area estimation, disaster history (mountain collapse history), etc.), artificial conditions (childcare / thinning system, logging period (period until logging), thinning frequency, thinning rate, machines used, etc.), economic conditions (Timber market price, wages, shipping costs, interest rates, etc.), system conditions (subsidies, tax rates, planning systems, etc.), forest space sensing data (outer shape of animals and plants, ridge sounds, wind sounds, animal calls, human sounds) Voice, tree shape, smell, taste, etc.), and is stored in association with each area ID. In addition, it is used to quantitatively evaluate the value of forests, that is, there is some correlation with the value of forests in view of common technical knowledge, and parameters etc. are used when quantitatively evaluating the value of forests. The data available as may be associated with each area ID and stored in a column of the table. It is thought that specifications such as resolution of space satellites (observation satellites) will be updated, and include observation data from far outside the forest, such as irradiating laser light from high altitude to high density. The above-mentioned various types of data are examples, and some types of data among all types may be stored. In addition, the meaning of each data of forest condition, meteorological condition, geological condition, man-made condition, economic condition, institutional condition and forest space sensing data in this embodiment is defined as a concept including the above examples.
The initial data stored in the database may be data already acquired throughout Japan (generally private). So far, the foundation of the database has already been built, so these data can be utilized.
The additional data added to the database includes actual data observed by the user (number, thickness, height of trees) and information collection device 200 (drone, space satellite, forestry machine with sensor, etc.) as big data. There is actual data collected (number of trees, thickness, height).
Further, as the additional data added to the database, there is actual data (profit amount, etc.) as a result of the actual execution by the user based on the estimated data (profit amount, etc.) by AI.
Furthermore, as additional data to be added to the database, results and information based on the results of big data analysis by AI will be added.
The basis is to link these information with the area ID as a key and store them in the database. It should be noted that the area ID may be used as a key to cooperate with various DBs.
In addition, the database may be cloud-based to improve usability.

[UI / output example]
Based on the above database, a UI (user interface) based on a map of the whole of Japan as shown in FIG. 4 is used. The screen may be displayed on a display mechanism such as a display as the output unit 130 of the user terminal 100.
At the microsection, basin, municipality, prefecture, national level, and in some cases at the individual tree level, it is possible to accommodate various scales, and the wider layer is composed of a stack of lower layers on a large scale. And the continuity of information between scales is guaranteed by using known techniques.
In other words, if you stack from a small scale to a large scale and focus on a certain point, you can grasp the point finely, while if you draw a viewpoint, you can understand a wide area well.
Here, the area ID is used as a key, and it is configured by using a known technique so that there is no contradiction in the total value of each layer, and there is also a contradiction between the current situation grasping, prediction, optimization, etc. between layers. It is configured so that it does not exist.
Figure 4 shows the small lots (future) that can be profitable by logging according to the fluctuations in the profit amount (profit-cost = profit) and the timber price (timber price) for each small lot based on the map of the whole of Japan. Profit depends on the timber price at the present time, but it is an example of visualizing and outputting (output) for each small section that is still profitable even if there is a predetermined fluctuation based on the timber price at the present time.
FIG. 4 shows raster data or a vector as an aggregate of micropartitions, as in a digital elevation model (DEM), by extracting micropartitions that are economically viable (sustainable) in the long term in response to fluctuations in material prices. This is an example shown in a map covering the entire area of Japan with one or both of the data. For example, the quantitative output of tree height, thickness, and number density is based on the type of tree, etc., for the set micro-compartment, the current state and future growth amount (trees are cut down at some point). It is necessary to include the viewpoint of sustainability instead of maximizing the profit at one point at any timing.) Is calculated and generated based on the growth rate prediction of the tree, and is newly generated from the present to the future. The height, thickness, number density, etc. of trees are generated and stored in the database. At that time, for example, a self-thinning rate (natural death rate) may be set, and a method for estimating the number density in consideration of future natural death may be used.
Then, the timber profit amount from the present to the future is newly generated for the set minute section and stored in the database. For example, an interest rate of 1% may be set as a discount rate, and a valuation method based on the present value in consideration of future interest rates may be used. Based on the current subsidy rate (subsidy), based on the mode of annual market conditions (there is a market for material prices), the slope of the terrain (mountain disaster risk, sunlight), position (land productivity / fertility) This is an example that considers only the amount of water and nutrition (there are many when it is close to a river). Even if the output takes into consideration other items (deer distribution / cost to drive away), it will output information for each minute section and expand it over a wide area while preserving detailed information from stands and basins to the entire area of Japan. It is also possible.

[Software cloud computing]
The server 300 provides at least a part of the functions of the server 300 shown in the above functional block by software, and the user uses the service (software) provided by the server 300 by the cloud (SaaS / ASP) from the user terminal 100. Then, since the input / output information is exchanged by accessing the database of the server 300, it contributes to the improvement of the usability (improvement of convenience) of the software such as the speeding up and the high accuracy of the output.
With the above software, the database of the server 300 can be accessed from all over Japan, and it can be linked with the input / output data via the software of the user terminal 100 of the user in all areas of the forest that occupies 70% of the country, and it is a big data. It leads to collection.

[AI analysis]
In the AI analysis in the server 300, for example, a learning model is generated from the obtained big data and analyzed as to how much profit the decision to cut in the forest is made.
The model generation unit 305 records and inputs, for example, the calculated estimated profit amount (= timber price, type, thickness, height, etc. of the profit item-topography around the cost item, wages, shipping cost, etc.). It may be to generate a learning model that learns the relationship with the actually measured (observed) profit amount obtained by doing so.
The output as a quantitative evaluation result by AI analysis is the profit amount (corrected estimated value / predicted value) itself as the evaluation result data of the forest value, but it can be further expanded to, for example, the expected profit. It may be an estimated value or a predicted value such as the range of the forest to be cut, the amount of timber production, the required labor amount, etc., which is calculated based on the amount. The output to be generated may be a real number such as wood production or a ratio such as the utilization rate of labor.
For example, the output of wood production is calculated based on the accumulation of the current situation and the future growth amount based on the tree growth rate prediction based on the type of tree etc. for each of multiple micro-sections corresponding to the points on the map. The amount of wood production from the present to the future is generated and stored in the database. For example, a certain discount rate (interest rate), profit amount or profit rate is set, and profitability is taken in consideration of future interest rate and profit amount. Only the timber production that can be obtained may be recorded and output, or the logic for calculating the age of felled trees that maximizes profits may be used.

Further, the model generation unit 305 autonomously generates training data (wood production amount of the above estimated data (corrected) and wood production amount actually observed and sensed as actual data input from a user terminal or the like). If there is an error above a predetermined threshold between the estimated value and the observed value, a new training model is generated by machine learning so as to minimize the error and improve the accuracy, and the training model is used. It may be corrected to provide output values such as tree growth rate and wood production. As a result, new knowledge and feedback from big data can be used to efficiently automate improvements in databases and software, and simulation, optimization, planning, etc. can be monitored by subsequent monitoring, taking into account the linkage with wide-area communication network systems. It is automatically corrected by the observed values of the obtained real-time input information, and a system that does not become obsolete over time is realized. The learning model (predictable technology) may be expressed by a mathematical formula or the like and incorporated into software by adding it to the source code of the program.
Also, in AI analysis, we will recognize the rules without deciding the category by ourselves, or let us judge what data is missing and what data is needed for better performance. By automating a system that acquires big data by AI itself, it is possible to establish a general-purpose AI that enables autonomous improvement in a wide area.
As a specific algorithm, for example, the accuracy of the result may be verified. Conventionally, since only high-frequency data is collected in a large amount, the accuracy of the part with a small amount of data becomes poor and the error becomes large. Considering the reliability of the data used for prediction, if the required estimation (simulation) reliability is not satisfied, AI collects the data by itself via the network.
As shown in Fig. 5, there are many 40-year-old forests throughout Japan on the horizontal axis (forest age) of the graph, and therefore a large amount of data is collected on the vertical axis (number of data), and the observed values (square marks). The deviation between the estimated value (triangle mark) is small, that is, the accuracy is high and the error is small. On the other hand, the number of 100-year-old forests is small, the amount of data is small, the difference between the observed value and the estimated value is large, and the reliability of the required estimation (simulation) may not be satisfied.
Therefore, if it is determined that the reliability of the estimation (simulation) required by AI analysis is not satisfied (the deviation between the observed value and the estimated value exceeds a predetermined threshold value), for example, flight by a drone as an information collecting device 200 Make a plan and output a plan such as what altitude and range to fly in order to collect the data according to the data required for prediction (estimation). Then, the drone is skipped so as to collect the data of the 100-year-old forest intensively, the data is collected, the estimated value is improved, and the difference between the observed value and the estimated value is within the allowable range (within the threshold value). It can be made smaller.

[Spatial axis / time axis]
"Spatial axis": At the level of the whole of Japan, trees grow due to various factors such as temperature, precipitation, sunshine duration, and status, and Japan has four seasons and individual regional characteristics.
"Time axis": Consider not only the current time but also the time axis from the viewpoint of tree growth. For example, estimated data such as yearly, monthly, daily, etc. are calculated and stored in the database. The multifaceted functions of forests, described below, depend on changes in forest conditions. Therefore, by estimating the change in the state of the forest as long-term continuous data by predicting the growth rate of trees, it is possible to perform a simulation extended in the time direction exceeding 100 years. Through forecasting tree growth, long-term simulation, optimization, logging projects, etc. can be planned even in the time direction.
Specific examples of simulation methods in the time direction are as follows.
As a premise, the type of the tree is determined based on the existing image data and the image data collected from the user terminal 100 and the information collecting device 200 (drone or the like). Specifically, the determination is made based on the feature amount of the tree such as the tree shape and the shape / color of the leaves. For this determination, a high-quality trained model generated by the model generator 305 by deep learning using a multi-layered neural network may be used for the relationship between the feature amount and the tree type. .. The appearance of trees changes depending on the season. In addition, the appearance of trees changes depending on the altitude and latitude / longitude. Therefore, the type of the tree may be determined by including these elements in the feature amount.
Next, we predict the future of forests. There is growth according to the type of tree and the environment.
Environmental parameters for tree growth include, for example, temperature in each month, precipitation in each month, sunshine hours in each month, and number of trees per unit area.
The model generation unit 305 is machine learning based on learning data (for example, the growth rate calculated by the calculation unit 303 and the actually measured growth rate acquired by the acquisition unit 302 obtained by recording and inputting by the user). The learning model may be generated by the above, and the evaluation result providing unit 306 may provide the growth rate of the tree as evaluation result data (corrected estimated value / predicted value) after being corrected by the learning model. As a result of predicting the growth rate of trees, data on the size (thickness, height) of one tree is stored in the database for each of a plurality of minute sections (area IDs) corresponding to points on the map. The quantitative output generated as a result of predicting the growth rate of trees is the yield ratio (1) even if it is a real number such as stand accumulation (volume of forests planted all at once ㎥). The number density when it is set to 100%, preferably 0.6 to 0.7, and when it is high, self-thinning starts) or the like may be used. Further, for example, the future climate change may be predicted from the past climate change and reflected in the prediction result of the growth rate of the tree.

[Acquisition of physicality of wide-ranging AI]
Diverse forests throughout Japan with four seasons are vastly distributed. Until now, AI with physicality applicable to diverse and vast forests throughout Japan has not been realized.
As shown in Fig. 1, sensing by a large amount of information gathering device 200 by a national level communication network via a system targeting forests throughout Japan, for example, automatic forest photography by a drone or space satellite camera mechanism or laser irradiation. (Tree position, thickness, height, outline of animals and plants, etc.), automatic recording by voice recorder (ridges in the forest, wind, animal calls, human voices, etc.), sensing by sensor-attached forestry machines, etc. (The shape of the tree to be cut (straight is suitable for wood), thickness, length, etc.) is automatically acquired.
AI will acquire "physicality (five senses)" consisting of "eyes (visual = camera mechanism)", "ears (hearing = voice recorder)", "hands (tactile = sensors of forestry machines, etc.)" .. It may be provided with a "nose (smell)" odor sensor, a "tongue (taste)" taste sensor, and the like.
It should be noted that the harmonious control of a large number of drones may be performed. Harmonic control of a large number of drones refers to self-sacrificing overall optimization control that is difficult for humans (for example, in field observations in an emergency such as a forest fire, a safe route to the observation target site with the preceding drone group. It is a control such as exploring on the premise of self-damage, following the route of the returned drone, and conducting a full-scale exploration by the subsequent drone.)
As a utilization of the above sensing results,
(1) It can be used to calculate wood production, biodiversity, forest CO2 absorption, and risk of mountain disasters by taking pictures in the forest.
For example, based on the sensing results of tree height, thickness, and number density obtained from observations taken in the forest for each of a plurality of minute sections corresponding to points on the map, the above-mentioned tree growth is based on the type of tree and the like. The rate can be calculated.
(2) It can be used for evaluation of biodiversity by voice recording or the like.
For example, the output of biodiversity in forests is based on data such as the types of birds obtained by discriminating from observations by voice recording in forests for each of multiple microsections corresponding to points on the map. Deep learning (deep learning) using a multi-layered neural network from the generation of sex and the cutting of forests (the diversity decreases due to cutting, and even if some cutting is performed, the diversity increases in the entire forest). ) Using the high-quality trained model generated by the model generation unit 305, the newly calculated biodiversity from the present to the future is generated and stored in the database. For example, the concept of Shannon's diversity index or indicator species may be set, and an evaluation method based on observation of a specific rare species may be used. The output to be generated may be a real number such as the number of species or a relativized index such as biodiversity.
(3) It can be used for evaluation such as evaluation of efficiency and profit amount of wood production based on the sensing result by the forestry machine attached to the sensor.
The model generation unit has a multi-layered structure based on learning data (movement time calculated by difference based on observations of forestry machines with sensors between two points on the map (cutting costs decrease in a short time)). It may be possible to provide output such as efficiency of wood production by using a high-quality trained model generated by the model generation unit 305 by deep learning using a neural network.

[Elucidation of synergistic effects and trade-offs of multifaceted functions of forests]
By AI analysis etc. in this embodiment, tree growth amount, wood production amount, CO2 absorption amount, biodiversity, mountain disaster risk, disaster occurrence frequency, biomass fuel yield, cherry blossom flowering time, autumn leaf time, wood It is possible to elucidate the synergistic effects and trade-offs of various multifaceted functions of forests such as profitability on the above-mentioned various scales.
Forests have various aspects such as landscape, money-making, wood production, employment opportunities, water conservation, biological ecology, etc., and simultaneous evaluation of synergistic effects and trade-offs for each. By doing so, it can be useful for various decision-making.
Elucidation of the synergistic effects and trade-offs of the multi-functionality of forests will be clarified as follows.
The evaluation result providing unit 306 simultaneously evaluates each function of the forest and quantifies the height of each function in the time direction and the spatial direction. As a result, the functions that rise at the same time can be regarded as having a synergistic effect. On the contrary, functions in which one is rising and the other is falling can be regarded as having a trade-off relationship. By repeatedly examining the above relationships, synergistic effects and trade-offs may be elucidated, and data on the multifaceted functions of forests that are mutually synergistic and trade-offs may be provided. Simultaneous evaluation also allows real-time visualization of how other functions change when a particular function is enhanced. Similar to the AI learning model described above, a method of learning by comparing the estimated value and the measured value for each function and accelerating the calculation by patterning may be incorporated into the internal calculation process. Further, a high-quality trained model generated by the model generation unit 305 such as clustering by machine learning by "unsupervised learning" may be used.

〔Example〕
The system in this embodiment can be used as the basic material for accountability and consensus building of the government that has invested a large amount of tax by AI analysis and the like, and can perform the following analysis and output. It may be possible to use a high-quality trained model generated by the model generation unit 305.
(1) If you want to minimize the risk of disaster prevention based on quantitative output, you can also determine the logging field and the time and space of logging timing when logging and replanting. For example, logging is linked to the installation of road networks (forest roads, work roads, etc.). Since the installation of the road network increases the risk of disaster prevention due to the weakening of the ground and the outflow of earth and sand, the area where the risk of disaster prevention is potentially high controls the collapse of the mountains by refraining from installing the road network and cutting down. It is possible to perform optimization calculations such as As the tree grows, the depth of the rhizome (the function of deepening the rhizome as it grows and rooting in the cracks of the basement layer to prevent erosion and collapse of the ground), which affects disaster prevention risk, also changes, so it becomes huge over time. A time axis may be added to the estimation by forecasting the forest.
(2) Predict the risk of mountain disasters. Predict the risk of mountain disasters based on the type of trees, age of trees, and slope of terrain. The risk of mountain disasters may be represented by contour density per unit area and mapped. For example, the quantitative output of mountain disaster risk is to calculate the current risk and future risk based on the type of tree, for example, set random numbers and scenarios, and rain due to future guerrilla rainstorms. An evaluation method based on uncertainty or probability theory that considers events may be used. The output to be generated may be a real number such as the amount of timber damage expected due to a mountain disaster, or may be a ratio such as the probability of occurrence of a disaster (steep slope, based on the complexity of the terrain). Since the depth of the rhizome changes as the tree grows, a time axis may be added to the estimation.
(3) For negotiations with individual forest owners, city planning for each municipality, consulting, etc., such as setting areas where logging is not profitable and the frequency of disasters is low as target areas for solar power generation. It will also be possible to utilize it. For example, the quantitative output may be a real number such as a profit amount or a ratio such as a disaster occurrence probability. Furthermore, it is possible to evaluate more advantageous land use options by comparing with the profitability of solar power generation.
(4) Predict the amount of CO2 absorbed (the amount of carbon dioxide absorbed). For example, the amount of CO2 absorbed per year is predicted based on the type, age, location, topography, etc. of the tree. The parameters of CO2 absorption are the frequency of thinning (frequency of thinning, 20% of area x 2 times, 40% of area x 1 time, balance with cost, balance with surrounding forests, at the timing when it becomes cheaper in total). And so on, the CO2 absorption amount is calculated by executing the calculation of the number density, the age of the tree, the position and the like. For example, the quantitative output of calculating the CO2 absorption of trees is the current and future growth rate of trees for the set micro-compartment based on the type of trees obtained from observations by flux towers in the forest. Is calculated, and the amount of CO2 absorbed is newly generated based on the growth rate of the tree from the present to the future and the CO2 conversion rate set for each tree species. At that time, for example, a biomass expansion coefficient may be set, and an evaluation method based on the amount of CO2 absorbed in consideration of branches and leaves and rhizomes from the stem volume may be used.

According to the above-mentioned embodiment, the value such as the economic value of the forest including the viewpoint of tree growth is quantitatively quantified by utilizing the AI applicable to various and vast forests throughout Japan. It is possible to realize a technology for using forest data that can be evaluated. And, as long as the above effect is obtained, it is needless to say that some of the processing steps may be changed in order, may be processed in parallel, or may be omitted, which is not essential.

It should be noted that the above-described embodiment is a preferred embodiment of the present invention, and various modifications can be made without departing from the gist of the present invention. For example, a process for realizing the function of each device / system may be performed by reading and executing a program for realizing the function of each device / system in each device / system. Further, the program is transmitted to another computer system by a transmission wave via a computer-readable recording medium such as a CD-ROM or a magneto-optical disk, or via a transmission medium such as the Internet or a telephone line. May be good. Further, some systems may be realized through human movement.
In addition, specific examples of the above-mentioned processes may be carried out in combination, and the known data and the like used at that time may be acquired from the network in cooperation with a known database or the like, or acquired thereof. The data may be stored and used in the database of this system. That is, the description of details that can be understood and realized by those skilled in the art from common general knowledge may be omitted as appropriate.

10 Forest data utilization system 100 User terminal 200 Information collection device 300 Server

Claims (12)

A storage unit that stores forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition, and forest space sensing data in the area ID that identifies each area on the map in association with each other. ,
A calculation unit that calculates a predicted value based on the data stored in the storage unit for the quantitative value of the forest in the area on the map.
An acquisition unit that acquires the measured value of the quantitative value of the forest in the region, and
The storage unit that stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit as learning data, and the storage unit.
A model generation unit that generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
A server including an evaluation result providing unit that calculates and provides the quantitative value of the forest in the designated area based on the generated learning model. Relationship between the measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions, and forest space sensing data data in each area on the map and the measured values of the quantitative value of forests in the area. A model generator that generates a learning model that has been trained in
The quantitative value of the forest in the designated area of the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition and the forest space sensing data to be evaluated. A server including an evaluation result providing unit that calculates and provides based on the generated learning model from the measured values of the data. The server according to claim 1 or 2, which calculates the quantitative value of the forest in consideration of the time axis of the growth rate of trees. When there is an error between the predicted value of the quantitative value of the forest calculated based on the learning model and the measured value thereof, the model generation unit obtains data necessary for improving the predicted value. The server according to any one of claims 1 to 3, which determines and causes an information collecting device connected via a network to collect the necessary data to generate a new learning model. The server according to any one of claims 1 to 4, wherein the physicality of AI is acquired by using the sensing data related to the five senses collected by the information collecting device connected via the network. The server according to any one of claims 1 to 5, wherein the evaluation result providing unit analyzes and provides data having either a synergistic effect or a trade-off relationship with each other among the multifaceted functions of the forest. .. A forest data utilization system in which one or more user terminals, one or more information collection devices, and a server are connected via a predetermined network.
The server
A storage unit that stores forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition, and forest space sensing data in the area ID that identifies each area on the map in association with each other. ,
A calculation unit that calculates a predicted value based on the data stored in the storage unit for the quantitative value of the forest in the area on the map.
An acquisition unit that acquires an actually measured value of the quantitative value of the forest in the region from at least one of the user terminal and the information collecting device.
The storage unit that stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit as learning data, and the storage unit.
A model generation unit that generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
A communication unit that receives an inquiry about the quantitative value of the forest in the designated area on the map from the user terminal.
A forest data utilization system including an evaluation result providing unit that calculates the quantitative value of the forest in the designated area on the map based on the generated learning model and provides it to the user terminal. .. A forest data utilization system in which one or more user terminals, one or more information collection devices, and a server are connected via a predetermined network.
The server
Measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions, and forest space sensing data in each area on the map from at least one of the user terminal and the information collecting device, and the above. An acquisition unit that acquires the measured value of the quantitative value of the forest in the area, and
A model generator that generates a learning model that learns the relationship between the measured value of the data and the measured value of the quantitative value of the forest.
A communication unit that receives an inquiry about the quantitative value of the forest in the designated area on the map from the user terminal.
The quantitative value of the forest in the designated area on the map is evaluated by the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition and the forest to be evaluated. A forest data utilization system including an evaluation result providing unit that calculates and provides based on the generated learning model from the measured values of the spatial sensing data. It is a method of using forest data using one or more user terminals connected via a predetermined network, and one or more information collecting devices and servers.
Regarding the quantitative value of the forest in each area on the map, the calculation unit of the server uses the area ID for identifying each area as the forest condition, meteorological condition, geological condition, artificial condition, and economic condition in each area. , The process of calculating the predicted value based on the data stored in the storage unit that stores the data of the system conditions and the forest space sensing data in association with each other.
A step in which the acquisition unit of the server acquires an actually measured value of the quantitative value of the forest in the region from at least one of the user terminal and the information collecting device.
A step in which the server stores the predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit in the storage unit as learning data.
A process in which the model generation unit of the server generates a learning model that learns the relationship between the predicted value and the measured value based on the learning data.
A step of the user terminal designating the specific area on the map and transmitting the quantitative value inquiry of the forest to the server.
The process in which the communication unit of the server receives the inquiry from the user terminal, and
A step in which the evaluation result providing unit of the server calculates the quantitative value of the forest in the designated area on the map based on the generated learning model and provides it to the user terminal. ,
A method of using forest data, wherein the user terminal has a step of displaying the quantitative value of the forest in the designated area on the map provided by the server. It is a method of using forest data using one or more user terminals connected via a predetermined network, and one or more information collecting devices and servers.
From at least one of the user terminal and the information gathering device, the acquisition unit of the server can obtain forest conditions, meteorological conditions, geological conditions, artificial conditions, economic conditions, institutional conditions, and forest space sensing data in each area on the map. The step of acquiring the measured value of the data and the measured value of the quantitative value of the forest in the region, and
A process in which the model generation unit of the server generates a learning model that learns the relationship between the measured value of the data and the measured value of the quantitative value of the forest.
A step of the user terminal designating the specific area on the map and transmitting the quantitative value inquiry of the forest to the server.
The process in which the communication unit of the server receives the inquiry from the user terminal, and
The evaluation result providing unit of the server evaluates the quantitative value of the forest in the designated area on the map by the forest condition, the meteorological condition, the geological condition, the artificial condition, and the above. A process of calculating and providing based on the learning model generated from the measured values of the data of the economic conditions, the institutional conditions, and the forest space sensing data.
A method of using forest data, wherein the user terminal has a step of displaying the quantitative value of the forest in the designated area on the map provided by the server. For information processing equipment
According to the calculation unit, regarding the quantitative value of the forest in each area on the map, the area ID that identifies the area is the forest condition, meteorological condition, geological condition, artificial condition, economic condition, institutional condition and The process of calculating the predicted value based on the data stored in the storage unit that stores the forest space sensing data in association with each other.
The process of acquiring the measured value of the quantitative value of the forest in the area by the acquisition unit, and
The predicted value calculated by the calculation unit and the measured value acquired by the acquisition unit stored in the storage unit by the model generation unit are used as learning data, and the relationship between the predicted value and the measured value is obtained. The process of generating the learned learning model and
A program that causes the evaluation result providing unit to execute a process of calculating and providing the quantitative value of the forest in the designated area based on the generated learning model. For information processing equipment
By the model generation unit, the measured values of forest conditions, meteorological conditions, geological conditions, man-made conditions, economic conditions, institutional conditions and forest space sensing data in each area on the map, and the quantitative value of the forest in the area. Processing to generate a learning model that learned the relationship with the measured value,
The evaluation result providing unit determines the quantitative value of the forest in the designated area as the forest condition, the meteorological condition, the geological condition, the artificial condition, the economic condition, the institutional condition, and the evaluation target. A program for executing a process of calculating and providing based on the generated learning model from the measured value of the data of the forest space sensing data.

Images