Information processing device, control method for information processing device, and program

Abstract

To enable the acquisition of drainage construction methods that correspond to the drainage targets in the field. [Solution] The information processing device includes an acquisition means for acquiring target drainage volume, target drainage time, rainfall per unit time, and field images per unit time in a field where drainage work is to be carried out, and an estimation means for inputting the target drainage volume, target drainage time, rainfall per unit time, and field images per unit time acquired by the acquisition means as first input data to a trained model which has been trained to output drainage work information that can be carried out with the drainage target based on input data including the drainage target of the field, and performing estimation, and outputting drainage work information corresponding to the obtained first input data.

Description

Technical Field

[0001] The present invention relates to an information processing device for supporting drainage construction, a control method for the information processing device, and a program.

Background Art

[0002] In a field converted from a paddy field (converted field), since a paddy field with originally poor drainage is used as a field for upland farming, there is a high probability of a problem that the drainage of the field is poor. Generally, in order to improve drainage, open channels or closed conduits are constructed. However, in some cases, the drainage may vary depending on the location of the field without appropriate construction being carried out. Variation in drainage can be a factor leading to uneven growth (poor crop growth in certain areas), so it is desirable to drain the field without uneven drainage. However, it is difficult for even an experienced person to perform appropriate drainage construction so that the field can be drained without uneven drainage. Patent Document 1 discloses a technique for drainage construction in a field, in which the content of vertical hole drainage construction is displayed on a map based on a distribution map of soil hardness and a distribution map of soil moisture content in the field.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] With the technique described in Patent Document ၁, it is possible to improve the drainage in areas with poor drainage, but it is not possible to display the construction content in consideration of the drainage target, such as how much precipitation is to be drained in how many hours. An object of the present invention is to obtain drainage construction content corresponding to the drainage target in a field.

Means for Solving the Problems

[0005] The information processing device according to the present invention is characterized by comprising: an acquisition means for acquiring a target drainage volume, target drainage time, rainfall per unit time, and an image of the field per unit time in a field where drainage work is to be carried out; and an estimation means for inputting the target drainage volume, target drainage time, rainfall per unit time, and image of the field per unit time acquired by the acquisition means as first input data to a trained model which has been trained to output drainage work information that can be carried out with a drainage target based on input data including the drainage target of the field, and performing estimation, and outputting the drainage work information corresponding to the obtained first input data. [Effects of the Invention]

[0006] According to the present invention, it is possible to obtain drainage construction methods that correspond to drainage targets in a field. [Brief explanation of the drawing]

[0007] [Figure 1] This is a diagram illustrating an example of a drainage construction system configuration. [Figure 2] This diagram illustrates an example of the estimated server and tractor hardware configuration. [Figure 3] This diagram illustrates an example of the functional configuration of the estimated server and tractor. [Figure 4] This diagram illustrates the estimation of drainage construction information using a pre-trained model. [Figure 5] This figure shows examples of input and output data for estimating drainage construction information using a pre-trained model. [Figure 6] This is a flowchart explaining the operation of the drainage construction system. [Figure 7] This is a flowchart explaining the operation of the drainage construction system. [Figure 8] This figure shows examples of input data and training data used to train a learning model. [Modes for carrying out the invention]

[0008] Embodiments of the present invention will be described below with reference to the drawings.

[0009] Figures 1(a) and 1(b) show examples of the configuration of a drainage construction system to which the information processing device in this embodiment is applied. Figure 1(a) shows an example of the system configuration during use, in which drainage construction information is acquired by estimation using a trained model and drainage construction is performed based on the acquired drainage construction information. Figure 1(b) shows an example of the system configuration during training of the trained model that estimates drainage construction information. Here, drainage construction information refers to information indicating the details of drainage construction, such as the type of drainage construction (open channel / underground channel), the start position and end position of the drainage construction (open channel / underground channel).

[0010] During use, as shown in Figure 1(a), the drainage construction system is realized by a drainage target value database 101, a weather database 102, a field image database 103, an estimation server 104, and a tractor 106, all connected via the network 100. During learning, as shown in Figure 1(b), the drainage construction system is realized by a weather database 102, an estimation server 104, a learning field drainage performance value database 107, and a learning field image database 108, all connected via the network 100. Note that the configurations shown in Figures 1(a) and 1(b) are examples and are not limited to these; other components not shown may also be included.

[0011] The drainage target value database 101 stores information indicating drainage targets for a field. For example, the drainage target database 101 stores information such as the target drainage volume (the amount of water to be drained) and the target drainage time (how long it should take to drain the target amount) as drainage targets for a field. Information indicating drainage targets for a field is stored in the drainage target value database 101 when the user inputs values ​​to be used as drainage targets, for example, via the control unit. For example, if a field is converted from paddy field to dry field, the target drainage volume is set to the 10-year probability 4-hour precipitation and the target drainage time is set to 4 hours, based on the general drainage standards for converted dry fields.

[0012] Weather database 102 stores historical precipitation per unit time for each region. For example, information such as "20XX / Y / Z / AAA region 10:00 20mm / h" is stored in weather database 102.

[0013] The field image database 103 stores field images per unit time for fields where drainage work is being carried out. Field images are taken and acquired using, for example, drones, artificial satellites, or fixed-point field cameras. Field images are also accompanied by date and time information and field location information. In the following explanation, field location information will be described as GPS location information as an example, but field location information is not limited to GPS location information as long as it is possible to identify the location of the target field. By using the date and time information and GPS location information attached to the field images, it is possible to correlate the past precipitation per unit time for each region in the weather database 102 with the precipitation per unit time for the field during the time period in which the field images were taken. Therefore, it is possible to obtain the precipitation per unit time at the time when the field images per unit time were taken.

[0014] The estimation server 104 can generate a trained model for outputting drainage construction information by training a learning model using information stored in the learning field drainage performance value database 107 and the learning field image database 108. Furthermore, using the generated trained model, the estimation server 104 can output drainage construction information that enables the drainage of the target amount in the target drainage time by performing estimations using information stored in the drainage target value database 101 and the field image database 103. In addition, the estimation server 104 can transmit the outputted drainage construction information to external devices such as the tractor 106 via the network 100.

[0015] The working machine 105 is a working machine that can be detached from the tractor 106. The content of the drainage construction that can be carried out varies depending on the type of working machine attached to the tractor 106. For example, when constructing a bullet culvert, a subsoiler equipped with a mole drain is attached as the working machine, and when constructing a drilled culvert, a cut drain is attached as the working machine.

[0016] The tractor 106 can acquire the drainage construction information output from the estimation server 104 via the network 100. Also, the tractor 106 can acquire the GPS position information of the tractor 106, and can also acquire the working machine information indicating what kind of working machine is attached from the working machine 105. Further, the tractor 106 can create a driving route for each type of working machine regarding the implementation of the drainage construction based on the drainage construction information output from the estimation server 104, and can determine an operable driving route based on the working machine information. Furthermore, the tractor 106 can automatically move to the driving start point of the operable driving route based on, for example, the GPS position information, and can automatically move from the driving start point to the end point of the operable driving route.

[0017] In the learning field drainage performance value database 107, information indicating the drainage performance of the field used for learning the learning model is stored. For example, information indicating the actual drainage volume (the amount of water that could be drained) and the actual drainage time (the time required to drain the actual drainage volume) of the field used for learning the learning model is stored in the learning field drainage performance value database 107. For example, if a certain field in a certain area has a record of being able to drain the 10-year probability 4-hour precipitation in 4 hours, the actual drainage volume is the 10-year probability 4-hour precipitation, and the actual drainage time is 4 hours.

[0018] The learning field image database 108 stores images of the field per unit time for the field used for learning the learning model. The images of the field are added with shooting date and time information and field position information (in this example, GPS position information). By using the shooting date and time information and GPS position information added to the images of the field, it is possible to obtain the precipitation per unit time at the timing when the images of the field per unit time are taken from the weather database 102.

[0019] During the learning of the learning model, for example, the estimation server 104 acquires data on the drainage performance and drainage time performance of the field used for learning the learning model from the learning field drainage performance value database 107. In addition, the estimation server 104 acquires images of the field per unit time for the field used for learning the learning model from the learning field image database 108. The estimation server 104 acquires the precipitation per unit time corresponding to the images of the field from the weather database 102 based on the images of the field acquired from the learning field image database 108. Then, the estimation server 104 uses the acquired drainage performance, drainage time performance, precipitation per unit time, images of the field per unit time, etc. as input data. In addition, the estimation server 104 performs learning of the learning model with the drainage construction information regarding the drainage construction of the field where the drainage performance can be drained by the drainage time performance and there is little drainage unevenness as teacher data, and generates a learned model for outputting the drainage construction information.

[0020] Furthermore, when estimating drainage construction information using a trained model, for example, the estimation server 104 obtains data on the target drainage volume and target drainage time for the field where drainage construction will be carried out from the drainage target value database 101. The estimation server 104 also obtains field images per unit time for the field where drainage construction will be carried out from the field image database 103. Based on the field images obtained from the field image database 103, the estimation server 104 obtains the precipitation per unit time corresponding to the field image from the meteorological database 102. Then, the estimation server 104 uses the obtained target drainage volume, target drainage time, precipitation per unit time, and field images per unit time as input data to perform estimation using the trained model and obtain drainage construction information.

[0021] In the example described above, the estimation server 104 is configured to train the learning model and estimate drainage construction information using the trained model, but this is not the only configuration. For example, another information processing device different from the estimation server 104 may train the learning model and generate a trained model for outputting drainage construction information. The trained model generated by the other information processing device may then be introduced into the estimation server 104, and the estimation server 104 may perform the estimation of drainage construction information.

[0022] Figure 2 illustrates an example of the hardware configuration of the estimation server 104 and the tractor 106. In Figure 2, components identical to those shown in Figure 1 are denoted by the same reference numerals. The estimation server 104 includes a CPU 201, ROM 202, RAM 203, GPU 204, NIC 205, estimation unit 206, HDD 207, and system bus 208. The CPU 201, ROM 202, RAM 203, GPU 204, NIC 205, estimation unit 206, and HDD 207 are communicated together via the system bus 208.

[0023] The CPU (Central Processing Unit) 201 executes computer programs and performs various processes. The CPU 201 also controls various parts of the estimation server 104 via the system bus 208. For example, the CPU 201 performs data transfer between various parts via the system bus 208. The system bus 208 is a bus for sending data such as target drainage volume, target drainage time, rainfall per unit time, and field image per unit time to various parts of the estimation server 104. The ROM (Read Only Memory) 202 is an electrically erasable and recordable non-volatile memory, such as flash memory. The ROM 202 stores various control programs for controlling the estimation server 104, for example. The RAM (Random Access Memory) 203 is a memory that temporarily stores constants, variables, programs, etc. for the operation of the CPU 201. The GPU (Graphics Processing Unit) 204 is an arithmetic unit that can perform efficient calculations by processing more data in parallel.

[0024] NIC205 is a network interface card for connecting to network 100. Estimation server 104 can, for example, send and receive data to network 100 via NIC205, such as target drainage volume, target drainage time, rainfall per unit time, field image per unit time, and drainage construction information. Estimation unit 206 estimates drainage construction information based on target drainage volume, target drainage time, rainfall per unit time, and field image per unit time, either through the coordinated calculations of CPU 201 and GPU 204, or by one of them. If high-speed processing is not required as a system requirement, GPU 204 may not be installed. HDD (Hard Disk Drive) 207 is a non-volatile storage device. HDD 207 can store, for example, trained models, target drainage volume, target drainage time, rainfall per unit time, field image per unit time received via NIC205, and drainage construction information output from estimation unit 206. Alternatively, other storage devices such as SSDs (Solid State Drives) using flash memory may be used instead of HDD207, or in conjunction with HDD207.

[0025] Tractor 106 includes a CPU 211, ROM 212, RAM 213, steering 214, NIC 215, GPS receiver 216, accelerator / brake 217, display device 218, and system bus 219. The CPU 211, ROM 212, RAM 213, steering 214, NIC 215, GPS receiver 216, accelerator / brake 217, and display device 218 are communicated together via the system bus 219.

[0026] The CPU 211 executes computer programs and performs various processes. The CPU 211 also controls various parts of the tractor 106 via the system bus 219. For example, the CPU 211 performs data transfer between various parts via the system bus 219. The system bus 219 is a bus for sending information such as drainage construction information, GPS location information, and implement information output from the estimated server 104 to various parts of the tractor 106. The ROM 212 is an electrically erasable and recordable non-volatile memory, such as flash memory. The ROM 212 stores various control programs for controlling the tractor 106, for example. The RAM 213 is a memory that temporarily stores constants, variables, programs, etc., for the operation of the CPU 211.

[0027] NIC215 is a network interface card for connecting to network 100. The tractor 106 can receive drainage work information, etc., transmitted from the estimated server 104 to network 100 via NIC205, for example. The steering 214 is a device that can control the direction of travel of the tractor 106 and is controlled by the CPU 211. The GPS (Global Positioning System) receiver 216 can receive GPS location information of the location where the tractor 106 is located. The accelerator / brake 217 controls acceleration and deceleration when the tractor 106 is operating automatically. The display device 218 displays the completion of drainage work and the type of implement to be attached, and notifies the user.

[0028] Figure 3 illustrates an example of the functional configuration of the estimated server 104 and the tractor 106. The estimated server 104's CPU 201 executes programs loaded into RAM 203, thereby realizing the various functions of the estimated server 104, which will be described later. Similarly, the tractor 106's CPU 211 executes programs loaded into RAM 213, thereby realizing the various functions of the tractor 106, which will be described later.

[0029] The data receiving unit 301 uses the NIC 205 to receive target drainage volume and target drainage time stored in the drainage target value database 101, and field images per unit time stored in the field image database 103, via the network 100. The received target drainage volume, target drainage time, and field images per unit time are temporarily stored in the RAM 203. When training the learning model, the data receiving unit 301 also receives actual drainage volume and drainage time stored in the learning field drainage performance value database 107, and field images per unit time stored in the learning field image database 108. The received actual drainage volume, drainage time, and field images per unit time are temporarily stored in the RAM 203. In addition, the data receiving unit 301 receives past precipitation per unit time for each region stored in the weather database 102. The data receiving unit 301 then determines the amount of rainfall per unit time at the time the field image was taken, based on the shooting date and time information and GPS location information attached to the field image per unit time, and temporarily stores it in the RAM 203. The data receiving unit 301 is an example of an acquisition means.

[0030] The data storage unit 302 stores data temporarily stored in the RAM 203, such as data received by the data receiving unit 301, drainage construction information output from the estimation unit 303, and trained models learned by the learning unit 304, into the HDD 207.

[0031] The estimation unit 303 loads the trained model generated by the learning unit 304 into the RAM 203 to estimate drainage construction information and temporarily stores the obtained drainage construction information in the RAM 203. Based on the target drainage volume, target drainage time, rainfall per unit time, and field image per unit time temporarily stored in the RAM 203 by the data receiving unit 301, the estimation unit 303 uses the CPU 201 and GPU 204 to estimate the drainage construction information. The estimation unit 303 is an example of an estimation means.

[0032] The learning unit 304 uses the CPU 201 and GPU 204 to train a learning model based on input data temporarily stored in the RAM 203 by the data receiving unit 301, and generates a trained model for outputting drainage construction information. The input data includes, for example, received drainage volume and drainage time data, field images per unit time, and precipitation per unit time at the time the field images were taken. The learning unit 304 trains the learning model using drainage construction information related to drainage constructions that were carried out in fields where drainage volume data could be drained in drainage time data and drainage unevenness was minimal as training data. The generated trained model is temporarily stored in the RAM 203. The learning unit 304 is an example of a learning method.

[0033] One example of a machine learning algorithm used to train a learning model is deep learning, which uses a neural network to generate its own features and connection weighting coefficients for learning. The learning unit 304 may include an error detection unit and an update unit. The error detection unit obtains the error between the output data output from the output layer of the neural network and the training data, according to the input data input to the input layer. The error detection unit may use a loss function to calculate the error between the output data from the neural network and the training data. The update unit updates the connection weighting coefficients between the nodes of the neural network, etc., based on the error obtained by the error detection unit, so as to reduce the error. This update unit updates the connection weighting coefficients, etc., for example, using backpropagation. Backpropagation is a method of adjusting the connection weighting coefficients between the nodes of each neural network so as to reduce the error between the output data from the neural network and the training data.

[0034] The data transmission unit 305 expands the drainage construction information estimated by the estimation unit 303 and stored in the HDD 207 into the RAM 203 and transmits it to the network 100 using the NIC 205.

[0035] The data receiving unit 311 receives drainage construction information via the network 100 using the NIC 215, GPS location information for the tractor 106 using the GPS receiving unit 216, and implement information from the implement 105. The received drainage construction information, GPS location information, and implement information are temporarily stored in the RAM 213.

[0036] The route calculation unit 312 acquires the operating route, start position, and end position for each drainage construction based on the drainage construction information temporarily stored in the RAM 213, and temporarily stores them in the RAM 213. The route selection unit 313 determines the currently available drainage construction content based on the work machine information from the work machine 105, and selects an operating route that can be operated from among the operating routes for each drainage construction.

[0037] The control unit 314 uses GPS position information and traversable driving routes related to the tractor 106 to control the steering 214 and accelerator / brake 217, automatically moving the tractor 106 from the start position to the end position of the traversable driving route. Automatic operation is considered complete when the tractor 106 has reached the end position of the driving route.

[0038] After the automatic operation is completed, the display unit 315 selects one traversable route from among the unused routes, determines the necessary work equipment information for the drainage work to be performed on that route, and displays it on the display device 218. The user attaches the work equipment corresponding to the work equipment information displayed on the display device 218 to the tractor 106. When the work equipment corresponding to the displayed work equipment information is attached to the tractor 106, the data receiving unit 311 acquires new work equipment information from the newly attached work equipment. Subsequently, the route selection unit 313 newly determines the currently possible drainage work content based on the new work equipment information and newly selects an traversable route from among the unused routes. Then, the control unit 314 automatically moves along the newly selected traversable route from the start position to the end position. The display unit 315 repeats the operation of newly displaying the necessary work equipment information for the drainage work to be performed on the route on the display device 218 until there are no more unused routes. When there are no longer any unused operating routes, the display unit 315 displays on the display device 218 that drainage work has been completed.

[0039] Figure 4 illustrates the estimation of drainage construction information using a trained model in this embodiment. The trained model 403 is a trained model for outputting drainage construction information, generated by training the trained model using information stored in the training field drainage performance database 107 and the training field image database 108. The trained model 403 receives the target drainage volume, target drainage time, rainfall per unit time, and field image per unit time as input data 401 for the field where drainage construction will be carried out. The trained model 403 estimates the drainage construction information based on the input data 401 and outputs the obtained drainage construction information as output data 402.

[0040] Figure 5 shows an example of input data (target drainage volume, target drainage time, precipitation per unit time, field image per unit time) and output data (drainage construction information) for estimating drainage construction information using a trained model. In the example input data shown in Figure 5, field images taken every hour are input as field images per unit time. In addition, the precipitation amount derived based on the date and time information and GPS location information attached to the field images per unit time, and information from the weather database 102 is input as precipitation per unit time. In the example output data shown in Figure 5, the details of each drainage construction (type of open or subsurface drain, start position, end position) are output as drainage construction information.

[0041] Figure 6 is a flowchart illustrating the operation of the drainage construction system. Figure 6 shows an example in which the estimation server 104 performs estimation using a trained model for outputting drainage construction information and outputs drainage construction information for the field where drainage construction will be carried out. The processing shown in the flowchart in Figure 6 is realized, for example, by the CPU 201 of the estimation server 104 executing a program loaded in RAM 203.

[0042] Furthermore, the drainage target for the field where drainage work is to be carried out is set by the user and stored in the drainage target value database 101. In addition, field images per unit time for the field where drainage work is to be carried out are taken based on instructions from the user and stored in the field image database 103.

[0043] In step S601, the CPU 201 obtains the target drainage volume and target drainage time set as drainage targets for the field where drainage work will be carried out, from the drainage target value database 101. In step S602, the CPU 201 obtains field images per unit time for the field where drainage work will be carried out from the field image database 103.

[0044] In step S603, the CPU 201 obtains the amount of rainfall per unit time at the time the field images were taken, based on the field images per unit time obtained in step S602. First, the CPU 201 obtains the past amount of rainfall per unit time for each region from the weather database 102. Then, the CPU 201 searches for and obtains the amount of rainfall per unit time at the time the field images were taken, using the date and time information and GPS location information attached to the field images per unit time, and the obtained past amount of rainfall per unit time for each region. Alternatively, the user may set a rainfall amount for each field image per unit time, and the set rainfall amount may be obtained as the amount of rainfall per unit time.

[0045] Furthermore, the execution order of steps S601 to S603 described above can be changed as needed, as long as the target drainage volume, target drainage time, field image per unit time, and precipitation per unit time are obtained before executing the next step, S604.

[0046] In step S604, the CPU 201 uses a trained model for outputting drainage construction information to estimate drainage construction information for the field where drainage construction will be carried out. The CPU 201 inputs the target drainage volume, target drainage time, field image per unit time, and precipitation per unit time acquired in steps S601 to S603 into the trained model as input data and performs estimation of drainage construction information using the trained model. The calculations related to the estimation of drainage construction information may be performed by the CPU 201 alone, by the GPU 204 alone in accordance with the control of the CPU 201, or by the CPU 201 and GPU 204 in cooperation.

[0047] In step S605, the CPU 201 outputs the drainage construction information obtained from the estimation of drainage construction information performed in step S604. In this way, drainage construction information is output as exemplified in Figure 5 as output data.

[0048] Figure 7 is a flowchart illustrating the operation of the drainage construction system. Figure 7 shows an example in which the tractor 106 performs drainage construction based on drainage construction information created by the estimated server 104. The processing shown in the flowchart in Figure 7 is realized, for example, by the CPU 211 of the tractor 106 executing a program or the like that is loaded into the RAM 213.

[0049] In step S701, the CPU 211 acquires the drainage construction information output by the estimated server 104.

[0050] In step S702, the CPU 211 creates an operating route for each drainage construction based on the drainage construction information acquired in step S701. For example, if the drainage construction information was output as text data as shown in Figure 5, the CPU 211 will: "Bullet-type culvert construction route: Starting position / latitude: 39°44'23"78"N Longitude: 104°59'22"60"W Ending point / latitude: 39°44'23" 80"N Longitude: 104°59'22" 60"W" "Drilled underground drainage construction route: Starting position / latitude: 39°44'20"N, longitude: 104°59'22"50"W Ending point / latitude: 39°44'19" 80"N Longitude: 104°59'22" 40"W" Create a driving route like this.

[0051] In step S703, the CPU 211 obtains implement information from the implement 105 attached to the tractor 106. In step S704, the CPU 211 determines the currently available drainage construction work based on the work equipment information acquired in step S703, and selects an operational route from among the operational routes for each drainage construction work. For example, if the CPU 211 determines, based on the acquired work equipment information, that the work equipment attached to the tractor 106 is a cut drain, it determines that the currently available drainage construction work is a perforated drain. In this case, the CPU 211 selects the "perforated drain construction route" as the operational route from among the operational routes created in step S702.

[0052] In step S705, the CPU 211 controls various parts of the tractor 106 to perform drainage work according to the driving route selected in step S704. At this time, the CPU 211 controls the steering 214 and accelerator / brake 217 based on GPS position information acquired by the GPS receiver 216, and automatically drives the tractor 106 from the start position to the end position of the drivable driving route. When the tractor reaches the end position of the driving route, the automatic driving is considered complete. After the automatic driving is completed, the process proceeds to step S706.

[0053] In step S706, the CPU 211 determines whether there are any unused operating routes among the operating routes created in step S702. If the CPU 211 determines that there are unused operating routes (YES in step S706), the process proceeds to step S707. On the other hand, if the CPU 211 determines that there are no unused operating routes (NO in step S706), the drainage work is considered complete, and the process shown in Figure 7 is terminated. When the drainage work is considered complete, the CPU 211 displays an indication that the drainage work has been completed on the display device 218.

[0054] In step S707, the CPU 211 selects one of the unused operating routes and determines the information of the implement to be used for the drainage work to be carried out on that route. The CPU 211 then displays the determined implement information on the display device 218 and prompts the user to change the implement 105. After the CPU 211 determines that the implement 105 to be attached to the tractor 106 has been changed, the process returns to step S703. In this way, the processes described in steps S703 to S707 are repeated until there are no unused operating routes left in the operating routes created in step S702.

[0055] The following describes the training method for the learning model used to estimate drainage construction information. The learning model is trained using a drainage construction system configured as shown in Figure 1(b), for example. Input data for training includes actual drainage volume, actual drainage time, precipitation per unit time, and field images per unit time. Training data consists of drainage construction information from fields where the actual drainage volume was achieved within the actual drainage time. By training the learning model using this input and training data, a trained model capable of outputting drainage construction information is generated.

[0056] As mentioned above, the learning field drainage data database 107 stores drainage data for fields used to train the learning model, such as drainage volume and drainage time. The learning field image database 108 stores field images per unit time for fields used to train the learning model. By using the date and time information and GPS location information attached to the field images, it is possible to correlate the past precipitation per unit time for each region in the weather database 102 with the precipitation per unit time for the field during the time period in which the field images were taken. Therefore, it is possible to obtain the precipitation per unit time at the time the field images were taken.

[0057] Input data and training data may be created based on actual field data. For example, the drainage volume record should be entered as the actual drainage volume that a field was able to drain, and the drainage time record should be entered as the time it took for a field to drain the actual drainage volume (actual drainage time). Additionally, the precipitation per unit time and field image per unit time should be entered as the precipitation per unit time for a field on a particular day and as the field image per unit time for a field on a particular day. The training data should include drainage construction information that was in place when a field was able to drain the actual drainage volume within the actual drainage time.

[0058] Figure 8 shows an example of input data (actual drainage volume, actual drainage time, precipitation per unit time, field image per unit time) and training data (drainage construction information) used for training a learning model. In the example input data shown in Figure 5, field images taken every hour are input as field images per unit time. In addition, precipitation derived from the date and time information and GPS location information attached to the field images per unit time, and information from the weather database 102, is input as precipitation per unit time. In the example training data shown in Figure 5, drainage construction information indicating the details of the drainage construction (type of open or subsurface drain, start position, end position) that was carried out when the actual drainage volume in the corresponding input data was able to be drained in the actual drainage time is input as training data.

[0059] According to this embodiment, the learning model is trained using the drainage volume data, drainage time data, rainfall per unit time, and field images per unit time as input data, and the drainage construction information that was successfully drained as training data, thereby generating a trained model for outputting drainage construction information. The estimation server 104 estimates the drainage construction information by inputting the target drainage volume, target drainage time, rainfall per unit time, and field images per unit time for the field where drainage construction is to be carried out as input data into the trained model for outputting drainage construction information. The estimation server 104 then outputs the drainage construction information obtained by estimating the drainage construction information using the trained model. This makes it possible to output drainage construction information corresponding to the input data, including the drainage target for the field where drainage construction is to be carried out, and to obtain drainage construction content that corresponds to the drainage target for the field. Furthermore, by carrying out drainage construction on the field based on the output drainage construction information, it becomes possible to carry out drainage construction that has good drainage performance and reduces drainage unevenness.

[0060] Incidentally, one factor related to field drainage is the environment surrounding the field. Therefore, when training a learning model, it is also possible to use images of the surrounding environment as input for field images per unit time. The trained model generated in this way can improve the accuracy of outputting drainage construction information so that the target drainage amount can be drained in the target drainage time, provided that the input field images per unit time also include images of the surrounding environment.

[0061] Another factor related to field drainage is drainage construction. Therefore, when training a learning model, the field images per unit time input may include information that shows the content (type) and location (position) of drainage construction already implemented in the field. For example, images with red markers indicating the locations of edge drainage channels and blue markers indicating bullet drainage channels could be used as part of the input data during training. In such a trained model, if the field images per unit time input include information on the content and location of drainage construction in the same format as during training, the accuracy of the output of drainage construction information can be improved so that the target drainage volume can be drained within the target drainage time.

[0062] Another factor related to field drainage is the field's slope. Therefore, it is advisable to add information that indicates the field's slope as input data. When training the learning model, additional information that indicates the field's slope is input and used for training. Examples of information that indicates the field's slope include gradient values ​​or aerial images of multiple fields taken at different angles. By using the trained model thus generated, it is possible to obtain drainage construction information that takes the field's slope into account, and the output accuracy of the drainage construction information can be improved so that the target drainage volume is drained within the target drainage time.

[0063] Another factor related to field drainage is the soil condition of the field. Soil conditions include the type of soil, soil hardness, and soil particle size. Therefore, it is advisable to input additional information that reveals the soil conditions of the field as input data. When training the learning model, additional information that reveals the soil conditions of the field is input and used for training. Examples of information that reveals the soil conditions of the field include images of the field surface and text such as "gley soil, soil hardness, particle size 0.3 mm". By using the trained model thus generated, it is possible to obtain drainage construction information that takes the soil conditions of the field into consideration, and the output accuracy of the drainage construction information can be improved so that the target drainage volume is drained in the target drainage time.

[0064] Furthermore, depending on the user utilizing the pre-trained model, the types of drainage construction work that can be performed may be limited. For example, a user who cannot provide a half-soiler cannot perform subsoil fracturing using a half-soiler. Therefore, it may be possible to add the types of drainage construction work that can be performed as input data. When training the model, the types of drainage construction work that can be performed should be added as input data. By using the resulting trained model, it is possible to limit the drainage construction information output from the pre-trained model.

[0065] Alternatively, instead of field images per unit time as input data, information indicating the water retention capacity of the field per unit time may be used. For example, field images with marks or colors indicating areas with high water retention capacity. When training the learning model, the water retention capacity per unit time is input instead of field images per unit time for training.

[0066] Furthermore, the drainage construction information output by the trained model may be output as text or images. If the output of the trained model is text, the training data used during training will be the drainage construction content written in text. If the output of the trained model is images, the training data used during training will be images of fields with markers added for each type of drainage construction.

[0067] (Other embodiments of the present invention) The present invention can also be realized by supplying a program that implements one or more of the functions of the above-described embodiments to a system or device via a network or storage medium, and by a process in which one or more processors in the computer of that system or device read and execute the program. It can also be realized by a circuit (e.g., an ASIC) that implements one or more functions.

[0068] It should be noted that the embodiments described above are merely examples of how the present invention can be implemented, and the technical scope of the present invention should not be interpreted as being limited by them. In other words, the present invention can be implemented in various forms without departing from its technical concept or its main features.

[0069] The disclosure of this embodiment includes the following configurations and methods, etc. (Composition 1) An acquisition means for acquiring the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in a field where drainage work is to be carried out, An information processing apparatus comprising: an estimation means that inputs the target drainage amount, target drainage time, precipitation per unit time, and field image per unit time acquired by the acquisition means as first input data to a trained model which has been trained to output drainage construction information that can be used to perform drainage based on input data including drainage targets for a field, and outputs drainage construction information corresponding to the obtained first input data. (Configuration 2) The information processing device according to Configuration 1, characterized in that it takes the actual drainage volume, actual drainage time, precipitation per unit time, and images of the field per unit time as input data, and has a learning means to generate the trained model by learning the actual drainage volume and drainage construction information of fields that were able to be drained in the actual drainage time as training data. (Composition 3) The information processing device according to configuration 1 or 2, characterized in that the drainage work information includes the type of drainage work and the start and end positions of the drainage work. (Composition 4) The information processing device according to any one of configurations 1 to 3, characterized in that the field images per unit time used to generate the trained model include field images per unit time that also capture the area around the field. (Composition 5) The information processing device according to any one of configurations 1 to 4, characterized in that the field images per unit time used to generate the trained model include field images per unit time to which information regarding the type and location of the drainage work being carried out is added. (Composition 6) The information processing device according to any one of configurations 1 to 5, characterized in that the input data includes information regarding the slope of the field. (Composition 7) The information processing device according to any one of configurations 1 to 6, characterized in that the input data includes information about the soil of the field. (Composition 8) The aforementioned input data includes the types of drainage work that can be performed. The estimation means is characterized in that it outputs drainage construction information only for the types of drainage construction that can be carried out, as described in any one of configurations 1 to 7. (Composition 9) The information processing device according to any one of configurations 1 to 8, characterized in that, instead of the image of the field per unit time, information regarding the water retention capacity of the field per unit time is used as the input data. (Method 1) The process involves acquiring the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in the field where drainage work is to be carried out. A control method for an information processing device, comprising: an estimation step, in which a trained model, which has been trained to output drainage construction information that enables drainage at a drainage target based on input data including a field drainage target, inputs the target drainage amount, target drainage time, precipitation per unit time, and field image per unit time acquired in the acquisition step as first input data to perform estimation and outputs the drainage construction information corresponding to the obtained first input data. (Program 1) The acquisition step involves obtaining the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in the field where drainage work is to be carried out. A program for causing a computer to perform an estimation step, in which a trained model, which has been trained to output drainage construction information that can perform drainage at a drainage target based on input data including a drainage target for a field, inputs the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time acquired in the acquisition step as first input data to perform estimation and outputs the drainage construction information corresponding to the obtained first input data. [Explanation of symbols]

[0070] 100: Network 101: Drainage target value database 102: Weather database 103: Field image database 104: Estimated server 105: Implement 106: Tractor 107: Field drainage performance value database for learning 108: Field image database for learning

Claims

1. An acquisition means for acquiring the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in a field where drainage work is to be carried out, An information processing apparatus comprising: an estimation means that inputs the target drainage amount, target drainage time, precipitation per unit time, and field image per unit time acquired by the acquisition means as first input data to a trained model which has been trained to output drainage construction information that can be used to perform drainage based on input data including drainage targets for a field, and outputs drainage construction information corresponding to the obtained first input data.

2. The information processing device according to claim 1, characterized in that it has a learning means that takes the actual drainage volume, actual drainage time, precipitation per unit time, and images of the field per unit time as input data, and uses the actual drainage volume and drainage construction information of fields that were able to be drained in the actual drainage time as training data to generate the trained model.

3. The information processing apparatus according to claim 1 or 2, characterized in that the drainage construction information includes the type of drainage construction and the start and end positions of the drainage construction.

4. The information processing apparatus according to claim 1 or 2, characterized in that the field images per unit time used to generate the trained model include field images per unit time that also capture the area around the field.

5. The information processing apparatus according to claim 1 or 2, characterized in that the field images per unit time used to generate the trained model include field images per unit time to which information regarding the type and location of the drainage work being carried out is added.

6. The information processing apparatus according to claim 1 or 2, characterized in that the input data includes information regarding the slope of the field.

7. The information processing device according to claim 1 or 2, characterized in that the input data includes information about the soil of the field.

8. The aforementioned input data includes the types of drainage work that can be performed. The information processing apparatus according to claim 1 or 2, characterized in that the estimation means outputs the drainage construction information only for the types of drainage construction that can be carried out.

9. The information processing device according to claim 1 or 2, characterized in that, instead of the image of the field per unit time, information regarding the water retention capacity of the field per unit time is used as the input data.

10. The process involves acquiring data on the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in the field where drainage work is to be carried out. Control method for an information processing device, comprising: an estimation step, in which a trained model, which has been trained to output drainage construction information that can perform drainage at a drainage target based on input data including a drainage target for a field, inputs the target drainage amount, target drainage time, precipitation per unit time, and field images per unit time acquired in the acquisition step as first input data to perform estimation and outputs the drainage construction information corresponding to the obtained first input data.

11. The acquisition step involves obtaining the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time in the field where drainage work is to be carried out. A program for causing a computer to perform an estimation step, in which a trained model, which has been trained to output drainage construction information that can perform drainage at a drainage target based on input data including a drainage target for a field, inputs the target drainage volume, target drainage time, precipitation per unit time, and field images per unit time acquired in the acquisition step as first input data to perform estimation and outputs the drainage construction information corresponding to the obtained first input data.

Images