Soil property estimation system, soil property estimation method, and soil property estimation program

Abstract

To provide a soil property estimation system, a soil property estimation method, and a soil property estimation program that can easily grasp soil property of the land to be evaluated without spending time and money.SOLUTION: A learning model 8 is used, which machine-learns a correlation between input data including soil analysis information (soil color, soil moisture content, the amount of rocks, etc.) of a column erecting site obtained by analyzing images of a column erecting site taken during a column erecting work and a grounding resistance value measured at the column erecting site, and output data including soil quality information of the column erecting site. Soil quality of the land to be evaluated is estimated from estimation input data including soil analysis information of an evaluation site (soil color, soil moisture content, the amount of rocks, etc.) obtained by analyzing images of the land to be evaluated, and the grounding resistance value measured at the land to be evaluated.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present invention relates to a soil quality estimation system, a soil quality estimation method, and a soil quality estimation program that can easily estimate the soil quality of land to be evaluated without spending a lot of time or money. [Background technology]

[0002] In order to increase the agricultural workforce, it is necessary to increase the number of people aspiring to become farmers. However, for those who are just starting out in farming, it is essential to conduct soil analyses and geological surveys of the desired location in order to select a place to work in agriculture. However, conducting soil analysis and geological surveys requires a considerable amount of time and money, which is a heavy burden for those aspiring to become farmers.

[0003] Meanwhile, when an electric power company erects a pole, it digs a hole in the ground, inserts the pole into it, and then backfills the hole. During this pole erection process, the company takes and saves photographs of the pole site (for example, photographs of the inside of the hole dug for the pole site, photographs of the base of the pole as it is being inserted into the hole, photographs of the base of the pole after it has been inserted into the hole and backfilled, etc.). Since utility poles are scattered over a wide area, if it were possible to easily determine the soil quality using photographs of these poles when they were erected, it would be possible to easily determine the soil quality of the land to be evaluated without spending a lot of time and money.

[0004] The present invention has been made in consideration of the above circumstances, and its main objective is to provide a soil quality estimation system, a soil quality estimation method, and a soil quality estimation program that can easily grasp the soil quality of land to be evaluated without spending a lot of time and money. Summary of the Invention [Means for solving the problem]

[0005] In order to achieve the above object, a soil quality estimation system for determining soil quality according to the present invention comprises: A pole installation site image database is provided that stores images of the pole installation site taken during the pole installation work (photos of the inside of the dug hole, photos of the base of the pole while the pole is being inserted into the hole, photos of the base of the pole after the pole has been inserted into the hole and backfilled, etc.), A learning model storage unit that stores a learning model that has been machine-learned to learn the correlation between input data including soil analysis information (such as soil color, soil moisture content, and rock volume) of the pole site obtained by analyzing the image of the pole site stored in the pole site image database and the ground resistance value measured at the pole site, and output data including soil information of the pole site; An input data acquisition unit that acquires estimation input data including evaluation site soil analysis information (soil color, soil moisture content, rock content, etc.) obtained by analyzing an image of the land to be evaluated and a ground resistance value measured on the land to be evaluated; a soil quality estimation unit that inputs the estimation input data acquired by the input data acquisition unit into the learning model to estimate the soil quality of the land to be evaluated; The present invention is characterized in that it has the following features.

[0006] Therefore, when one wishes to know the soil quality of a piece of land to be evaluated, an image of the land to be evaluated is taken and the ground resistance value of the land is measured. The input data acquisition unit acquires estimation input data including the soil analysis information of the evaluation site (soil color, moisture content, amount of stones and rocks) obtained by analyzing the image of the land to be evaluated and the ground resistance value of the evaluation site, and the soil quality estimation unit estimates the soil quality of the land to be evaluated from the estimation input data obtained by the input data acquisition unit using the learning model stored in the learning model storage unit.

[0007] In this way, when investigating the soil quality of the land to be evaluated, by using a learning model created based on a large amount of learning data consisting of input data including images of the pole construction site taken during the pole construction process and the ground resistance value at that location, and output data including soil information for each pole construction location, it is possible to easily estimate the soil quality of the land to be evaluated from the images and ground resistance value of the land to be evaluated.

[0008] Here, as input data, information on the grounding method used when erecting the pole at the pole site is further added, Information on the grounding method used when erecting the pole on the land to be evaluated may be added as input data for estimation. The grounding method used when erecting the pole should be appropriate depending on the soil quality, so by taking the grounding method into account, it becomes possible to more accurately estimate the soil quality of the land to be evaluated.

[0009] The system further includes a soil utilization database in which information on soil utilization methods suitable for each soil type or information on crops suitable for each soil type is stored, Based on the soil quality estimated by the soil quality estimation unit, information on how to use the soil suitable for that soil quality or information on crops suitable for that soil quality may be presented. Such a system not only provides information about the soil quality, but also about how the soil of the land to be evaluated and the crops that are suitable for that land, making it possible to select land more accurately. Effect of the Invention

[0010] As described above, according to the present invention, it is possible to easily grasp the soil quality of the land to be evaluated without spending a lot of time and money. [Brief description of the drawings]

[0011] [Figure 1] 1 is a schematic diagram showing a soil quality estimation system according to the present invention; [Diagram 2] FIG. 1 is a block diagram showing an example of a machine learning device of a soil quality estimation system according to the present invention. [Diagram 3] (a) is a correlation diagram showing the correlation between soil quality, earth resistivity, ground hardness and softness, and grounding method, and (b) is a table showing the soil quality, grounding method, and resistivity guidelines. [Figure 4] 1 is a diagram showing an example of an image of a pole construction site and learning data used in machine learning. FIG. [Diagram 5] 1 is a flowchart showing an example of a machine learning method performed by the machine learning device according to the present invention. [Figure 6] FIG. 1 is a block diagram showing an example of a soil property estimation device. [Figure 7] 4 is a flowchart showing an example of a soil property estimation method performed by the soil property estimation device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a schematic configuration of a soil property estimation system 1 according to the present invention. As shown in the figure, this soil quality estimation system 1 comprises an imaging device 2 (an imaging device 2a that captures images used in the learning phase, and an imaging device 2b that captures images used in the estimation phase), a pillar site image database 3, an image analysis device 4 (an image analysis device 4a that analyzes images captured by the imaging device 2a, and an image analysis device 4b that analyzes images captured by the imaging device 2b), a machine learning device 5, a soil quality estimation device 6, and a display 7.

[0013] The photographing device 2 may be a portable digital camera, a video camera, or the like, and the photographing device 2a is used to photograph the land where the pole is to be erected (hereinafter referred to as the pole erection site) at each stage of the pole erection work. That is, in the pole erection work, in order to keep a record of the pole erection process, it is necessary to take a photograph of the inside of the dug hole, a photograph of the base part of the pole while the pole is being inserted into the hole, a photograph of the base part of the pole in a state where the lower end of the pole is inserted into the hole and backfilled, etc., and these photographs are taken to keep a record of the work process and are used to collect images for forming input data used in forming a learning model. In addition, the photographing device 2b is used to photograph the land to be evaluated (hereinafter referred to as the evaluation site). That is, it is used to collect images for forming input data for a learned learning model. Here, the image capturing device 2a used in the learning phase and the image capturing device 2b used in the estimation phase may be the same or different.

[0014] The pole site image database 3 stores image data captured by the photographing device 2a in association with utility pole information. The image data stored in the pole site image database 3 is used as needed to form a learning model in the learning phase. In the learning phase, the image data captured by the photographing device 2a is temporarily stored in the pole erection site image database 3 to keep a record of the pole erection process, but may be sent directly to the image analysis device 4a. In the estimation phase, the image captured by the photographing device 2b is sent directly to the image analysis device 4b, but may be temporarily stored in an image database (not shown) to perform pre-processing such as removing unnecessary image parts.

[0015] The image analysis device 4 is used to analyze the image of the land photographed by the photographing device 2 to obtain soil analysis information (soil color, soil moisture content, rock amount, etc.), and the image analysis device 4a used in the learning phase is used to analyze the image of the pole building site to obtain the pole building site soil analysis information and output this pole building site soil analysis information to the machine learning device 5. The image analysis device 4b used in the estimation phase is used to analyze the image of the evaluation site to obtain the evaluation site soil analysis information and output this evaluation site soil analysis information to the soil estimation device 6.

[0016] The machine learning device 5 operates as the main body of the learning phase, and generates a learning model 8 through machine learning to be used for estimating the soil quality of any piece of land from information obtained by analyzing an image of the pole construction site, as well as the ground resistance value and soil quality information at the pole construction site. The soil quality estimation device 6 operates as the main part of the estimation phase, and uses a trained learning model 8 generated by the machine learning device 5 to estimate the soil quality of the location from information obtained by analyzing images of the evaluation site and the ground resistance value of the location. The display unit 7 can be replaced by the display screen of various terminals, and displays information on the soil quality estimated by the soil quality estimation device 6, information on the utilization of this soil, and the like.

[0017] The machine learning device 5 is configured with a general-purpose or dedicated computer. This computer may be a stationary computer or a portable computer, and may also be a client computer, a server computer, or a cloud computer. As shown in FIG. 2, the machine learning device 5 includes a learning data acquisition unit 51, a learning data storage unit 52, a machine learning unit 53, and a trained model storage unit 54.

[0018] The learning data acquisition unit 51 functions as an interface unit that acquires learning data including input data and output data via the communication network 11. This learning data acquisition unit 51 is connected via the communication network 11 to the image analysis device 4a, the worker terminal 10 used by the worker, and the like. The learning data storage unit 52 is a database that stores a plurality of sets of learning data acquired by the learning data acquisition unit 51 .

[0019] The machine learning unit 53 performs machine learning using the learning data stored in the learning data storage unit 52. This machine learning unit 553 inputs multiple sets of learning data to the learning model 8, and causes the learning model 8 to learn the correlation between the input data included in the learning data and the output data corresponding to the input data, thereby generating a learned model. There are various methods for generating a trained model, but it is preferable to use a neural network, for example, which performs supervised learning by the machine learning unit 53.

[0020] The trained model storage unit 54 is a database that stores the trained learning model 8 generated by the machine learning unit 53. The trained model 8 stored in the trained model storage unit 54 is provided to the soil estimation device 6 via any communication network or storage medium. The learning data storage unit 52 and the trained model storage unit 54 may be configured as a single storage device or as separate storage devices.

[0021] The learning data includes, as input data, at least image analysis information (such as soil color, soil moisture content, and amount of rocks) obtained by analyzing an image of the land, and the ground resistance value of the land from which the image was taken. Here, soil color refers to the color pattern of soil when viewed from the outside. Soil color reflects the environment in which the soil was formed and is closely related to the chemical, physical, and biological properties of the soil, so it is an effective and important factor for classifying soil types and identifying soils. In general, soil color is determined by the amount of humus, the form of iron compounds, the amount of aluminum, etc. Soil color is made up of various colors and changes depending on the state of dryness and moisture. Soil moisture (water content) represents the absolute amount of water in the soil and is expressed as a weight or volume content. It is closely related to soil texture, and is therefore a useful and important factor for classifying soil types and identifying soils. The amount of rock affects the soil's water absorption rate, apparent porosity, water retention rate, etc., and has a significant impact on soil quality, so it is a useful and important factor for classifying soil types and identifying soil.

[0022] The earth resistance value also indicates how easily electricity passes through it. For example, if the earth resistance value is high, it is difficult for electricity to pass through, and so it contains a lot of sand, stones, rocks, etc. On the other hand, if the earth resistance value is low, it is easy for electricity to pass through, and so it indicates that the soil contains a lot of moisture, mud, or is near water. For this reason, the ground resistance value differs depending on the soil type, and when the soil type is evaluated based on the earth resistivity, which has a predetermined relationship with this ground resistance value, a correlation diagram such as that shown in Figure 3(a) is obtained. Using this correlation diagram, as shown in Figure 3(b), for example, if the earth resistivity is low at 83 (Ω·m), the soil type is determined to be "clay", if the earth resistivity is 100 (Ω·m), the soil type is "sand", if the earth resistivity is 200 (Ω·m), the soil type is "gravel", and if the earth resistivity is 500 (Ω·m), the soil type is "rock". The earth resistance value is calculated by measuring the ground resistance and using a back calculation method from the measured value. In Fig. 3(a), the horizontal axis indicates the hardness of the ground, for example, the N value, which quantitatively indicates the hardness of the ground, is used. Since the soil type may differ even if the N value is the same, it is possible to identify the soil type by combining it with the earth resistivity.

[0023] (Machine learning methods) The machine learning device 5 forms a learning model using, for example, a neural network model. This neural network model is a publicly known model, and learns the correlation between the input data and the output data by inputting input data included in the learning data to an input layer and comparing output data output from an output layer as an estimation result with output data (teacher data) included in the learning data.

[0024] First, as a preliminary preparation for having the machine learning device 5 perform machine learning, the learning data acquisition unit 51 acquires a desired number of learning data and stores the acquired multiple sets of learning data in the learning data storage unit 52. The number of learning data prepared here is appropriately set in consideration of the estimation accuracy required for the learning model 8.

[0025] There are various ways to prepare learning data, but when an electric power company erects a pole, it takes photographs of the construction process (photos of the inside of the dug hole, photos of the base of the pole as it is being inserted into the hole, photos of the base of the pole after it has been inserted into the hole and backfilled, etc.), and these photographs are linked to the pole information and stored in pole site image database 3. Therefore, by obtaining images of the pole construction process stored as records of past pole construction work from the pole construction site image database 3 and analyzing the images of the pole construction process, soil analysis information (such as soil color, soil moisture content, and rock volume) of the pole construction site can be obtained. In addition, the current earth resistance value of the pole construction site corresponding to the image is measured, and the soil type of the corresponding pole construction site is determined. Here, the soil type can be selected using the classification shown in Figure 3 (mud, clay, ordinary soil, sand, Masa, gravel, boulders, boulders, Masa, bedrock, etc.) which is classified according to the relationship between earth resistivity and softness of the soil. Then, the measured ground resistance value and the determined soil information are input via the worker terminal 10, and the image analysis information and ground resistance value obtained by the image analyzer 4a are used as input data, and the determined soil data is used as output data (teaching data), which are linked to the utility pole information to create a set of learning data. After that, a large number of such learning data are prepared based on data from the time of erecting a large number of poles, as shown in Fig. 4, and are stored in advance in the learning data storage unit 52 as learning data for generating a learning model.

[0026] After the above-mentioned preparations, the machine learning unit 53 performs, for example, a series of processes in the flowchart shown in FIG. 5 to generate a trained model. In step S100, the machine learning unit 53 acquires one set of learning data from the multiple sets of learning data stored in the learning data storage unit 52. The learning data may be acquired in a predetermined order or randomly.

[0027] Next, in step S110, the machine learning unit 53 performs supervised machine learning. That is, the machine learning unit 53 inputs input data included in the acquired set of learning data to the prepared learning model 8, outputs an estimated result, compares this estimated result with output data (teacher data) included in the learning data acquired in step S100, and performs machine learning by a known neural network method. In this way, the machine learning unit 53 causes the learning model 8 to learn the correlation between the input data and the output data (soil information).

[0028] Thereafter, using this learning model under learning, the above-mentioned operations are performed on the remaining prepared learning data, and learning is continued in the machine learning unit 53. At this time, the machine learning unit 53 determines whether or not to continue machine learning based on the error between the output data and the teacher data, the number of learnings, the remaining amount of unlearned learning data stored in the learning data storage unit 52, etc. (step S120).

[0029] That is, if the machine learning unit 53 determines in step S120 to continue machine learning (YES), it performs steps S100 to S110 on the learning model 8 being trained using untrained learning data. On the other hand, if the machine learning unit 53 determines in step S120 not to continue machine learning (NO), it stores the generated trained learning model 8 in the trained model storage unit 54 in step S130 and ends the machine learning.

[0030] (Soil quality estimation device) Next, a soil quality estimation device 6 that estimates the soil quality of the evaluated land using the trained model generated using the above-mentioned technique will be described. The soil quality estimation device 6 is composed of a general-purpose or dedicated computer. This computer may be a stationary computer or a portable computer, and may also be a client computer, a server computer, or a cloud computer. As shown in Fig. 6, the soil quality estimation device 6 includes an input data acquisition unit 61, an estimation unit 62, a trained model storage unit 63, an output processing unit 64, and a soil utilization database 65.

[0031] The input data acquisition unit 61 acquires information obtained by analyzing the image of the evaluation site captured by the photographing device 2b using the image analysis device 4b, that is, evaluation site soil analysis information such as the soil color, soil moisture content, and rock content of the evaluation site. In addition, the earth resistance value measured at the evaluation site is input via the operator terminal 10, and a set of data consisting of the evaluation site soil analysis information and the evaluation site ground resistance value is acquired as input data. When a set of input data is input by the input data acquisition unit 61, the estimation unit 62 inputs the input data acquired by the input data acquisition unit 61 to the learning model 8 stored in the trained model storage unit 63, and performs estimation processing (estimating the soil quality of the land to be evaluated). For this estimation processing, the trained learning model 8 trained by the machine learning device 5 is used.

[0032] The estimation unit 62 may not only have the function of performing estimation processing using the learning model 8, but may also include a preprocessing function of adjusting the input data acquired by the input data acquisition unit 61 to a desired format, etc., and inputting it to the learning model 8 as preprocessing of the estimation processing.

[0033] The trained model storage unit 63 is a database that stores trained learning models 8 used in the estimation process of the estimation unit 62. The number of learning models 8 stored in the trained model storage unit 63 is not limited to one. For example, multiple learning models 8 with different numbers of input data or different machine learning techniques may be stored and used by selecting them appropriately.

[0034] An example of a case in which the number of input data is different is when the grounding method used at the time of pole erection is specified based on records made at the time of pole erection and added to the input data. Since the earthing method depends on the soil quality of the area where the method is used, there are appropriate earthing methods (indicated in [ ]) according to the soil quality as shown in Figure 3 above. For example, if the soil quality is "sand", the "top point" method is suitable as the earthing method, and if the soil quality is "rock", the "multi-earth B" or "deep buried electrode" method is suitable as the earthing method. In this way, since the earthing method is closely related to the soil quality, if the earthing method at the time of erection is recorded, adding that information to the input data makes it possible to form a learning model that can accurately estimate the soil quality.

[0035] The soil utilization database 65 is a database that stores information on the methods of using soil suitable for each type of soil and the crops suitable for each type of soil. The output processing unit 64 acquires soil information of the evaluation site in order to display the results estimated by the estimation unit 62 on the display 7, and also acquires information on soil utilization methods suitable for the estimated soil quality of the evaluation site and information on crops suitable for that soil quality by referring to the soil utilization database 65, and displays these acquired information on the display 7.

[0036] FIG. 7 is a flow chart showing an example of a soil quality estimation method using the soil quality estimation device 6, and the soil quality of the land to be evaluated is estimated as follows. First, in step S200, a photograph of the land to be evaluated (evaluation site) is taken, and the photographed image is analyzed by the image analyzer 4b to obtain soil analysis information (soil color, soil moisture content, rock content, etc.) of the evaluation site. At the same time, the earth resistance value of the evaluation site is measured and obtained.

[0037] Then, in step S210, the acquired image analysis information and ground resistance value of the evaluation site are input as input data to the trained learning model 8, and the soil quality of the evaluation site is estimated. Then, in step S220, information about the soil quality estimated by the learning model 8 is displayed on the display. The information relating to soil type here includes the soil type classification (mud, clay, ordinary soil, sand, Masa, gravel, boulders, boulders, granular Masa, or bedrock), and may also include the numerical value of earth resistivity and information indicating the softness or hardness of the soil. Furthermore, in step S230, based on the estimated soil quality, the soil utilization database 65 is referenced to extract information on a soil utilization method suitable for the estimated soil quality or information on crops suitable for the estimated soil quality, and this is displayed on the display.

[0038] Therefore, according to the above-mentioned soil quality estimation system 1, when obtaining soil information of the land to be evaluated, it is possible to easily and accurately grasp the soil quality of the evaluation site by obtaining images of the evaluation site and ground resistance value data without requesting soil analysis or geological surveys. Furthermore, in the above-described embodiment, not only the soil quality of the evaluated site is presented, but also the method of using the soil of the evaluated site and crops suitable for the land, making it possible to more accurately select a location for engaging in agriculture, for example.

[0039] In the above configuration, in order to obtain soil information of the evaluation site, the soil analysis information of the evaluation site (soil color, soil moisture content, amount of rocks, etc.) obtained by analyzing images of the evaluation site and the ground resistance value of the evaluation site are used as input data. However, if the grounding method used at the time of pole construction is added to the input data in the learning phase, it is desirable to also add the grounding method to the input data in the estimation phase. However, if there are no utility poles at the evaluation site, the grounding method used during the construction of other structures at the evaluation site may be added as input data.

[0040] In the above embodiment, a case has been described in which a neural network is adopted as a specific method of machine learning by the machine learning unit 53, but the machine learning unit 53 may adopt any other machine learning method. For example, tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural net types (including deep learning) such as recurrent neural networks and convolutional neural networks, clustering types such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors and k-means, multivariate analyses such as principal component analysis, factor analysis, and logistic regression, support vector machines, and the like may be adopted.

[0041] In addition, the above-mentioned machine learning device 5 and soil quality estimation device 6 can also be provided in the form of a program (machine learning program or soil quality estimation program) for executing each step of the above-mentioned machine learning method or soil quality estimation method. [Explanation of symbols]

[0042] 1. Soil quality estimation system 3. Pillar site image database 61 Input data acquisition section 62 Estimation part 63 Trained model memory 65 Soil Utilization Database

Claims

1. A soil quality estimation system for estimating soil quality, comprising: A pole erection site image database is provided for storing images of the pole erection site taken during the pole erection work, A learning model storage unit that stores a learning model that has been machine-learned to learn the correlation between input data including soil analysis information of the pole site obtained by analyzing the image of the pole site stored in the pole site image database and the ground resistance value measured at the pole site, and output data including soil information of the pole site; An input data acquisition unit that acquires estimation input data including evaluation site soil analysis information obtained by analyzing an image of the land to be evaluated and a ground resistance value measured on the land to be evaluated; a soil quality estimation unit that inputs the input data for estimation acquired by the input data acquisition unit into the learning model to estimate the soil quality of the land to be evaluated; A soil quality estimation system comprising:

2. As the input data, information on the grounding method used at the time of erecting the pole at the pole site is further added, 2. The soil quality estimation system according to claim 1, further comprising information on a grounding method used when erecting a pole on the land to be evaluated, as said input data for estimation.

3. Further comprising a soil utilization database storing information on a method of using soil suitable for each soil type or information on crops suitable for each soil type; 3. The soil quality estimation system according to claim 1, further comprising: a soil quality estimation unit that estimates the soil quality based on the soil quality, and that presents information on a method of using the soil suitable for the soil quality or information on crops suitable for the soil quality.

4. A soil quality estimation method for estimating soil quality, comprising: An input data acquisition step for acquiring an image and a ground resistance value of the land to be evaluated; A soil quality estimation step of estimating the soil quality of the land to be evaluated from the estimation input data including the evaluation site soil analysis information obtained by analyzing the image of the land to be evaluated acquired in the input data acquisition step and the ground resistance value of the land to be evaluated acquired in the input data acquisition step, using a learning model that has been machine-learned to learn the correlation between input data including the soil analysis information of the pole building site obtained by analyzing the image of the pole building site taken during the pole building work and the ground resistance value of the pole building site measured at the pole building site, and output data including soil quality information of the pole building site; A soil quality estimation method comprising the steps of:

5. The input data acquisition step includes acquiring information on a grounding method used when erecting a pole on the land to be evaluated, The input data of the learning model further includes information on the grounding method used when erecting the pole at the pole site.

5. The method for estimating soil quality according to claim 4.

6. A soil quality estimation program for causing a computer to execute each step of the soil quality estimation method according to claim 4 or 5.

Images