JP7424066B2 - Geological evaluation system, geological evaluation method and geological evaluation program - Google Patents

Description

The present invention relates to a geological evaluation system, a geological evaluation method, and a geological evaluation program for evaluating geology in order to create geological maps and the like.

In dam construction work, etc., we clarify the characteristics of the rocks that make up the bedrock at the site where construction or material collection is planned, as well as the geological conditions of the underlying bedrock, and conduct geotechnical studies necessary for safe and economical dam design and construction. In addition to understanding the properties as accurately as possible, we are determining the optimal construction method and material collection method. Such investigations into the properties of rocks are generally carried out by geological specialists at the planned locations for construction or material collection. In this case, a geological engineer removes a rock from an outcrop using a tool such as a rock hammer or a clinometer, observes the rock's characteristics by sight or touch, and determines the name of the rock and the engineering properties of the rock. For this reason, the skills and knowledge of geological specialists were required. In addition, when creating a geological map, it is necessary to clearly indicate the reference position (reference line) on the field with paint, etc. to determine the position on the geological map, or to measure the position of cracks by offset surveying. , there was a lot of outdoor work.

Therefore, a rock determination system that can be easily implemented at the sampling site is being considered (see, for example, Patent Document 1). The rock classification system described in this document is compatible with a rock classification frame that describes the questions and answers indicating the characteristics of each rock and the answer confidence, a fuzzy rule that describes the ambiguous characteristics of each rock, and the fuzzy rule. The system includes a knowledge database device consisting of membership functions. This rock determination system performs a rule-based inference mechanism and a fuzzy inference mechanism in correspondence with the knowledge database device to determine the type of rock.

Japanese Patent Application Publication No. 6-220833

However, the judgment of the engineering properties of rocks and bedrock by geological engineers is difficult because it is based on images of the rock surface and information on impact response at the site. Furthermore, when the evaluation target is wide-ranging, it takes time to acquire the impact response information for the evaluation target. For this reason, it was difficult to efficiently evaluate the geology.

A geological evaluation system that solves the above problems includes an image information storage section that stores photographed images of the ground surface in association with the ground surface position, and a geological information storage section that stores geological information in association with the ground surface position. , a geological evaluation system for performing geological evaluation, comprising: a control unit for predicting geology; the geological information of the unit area is obtained from the geological information storage section in correspondence with the ground surface position, and the captured image associated with the unit area containing a single piece of geological information is A prediction model for predicting geological information based on the photographed image is generated using the geological information as training information, and a prediction model for predicting the geological information is generated based on the photographed image, and a prediction model for predicting the geological information is generated using the photographed image of the ground surface position of the evaluation target area and the prediction model. Predict geological information in the target area.

According to the present invention, it is possible to efficiently and appropriately evaluate geology.

FIG. 1 is a system configuration diagram of a geological evaluation device in an embodiment. FIG. 2 is an explanatory diagram of the hardware configuration of the embodiment. FIG. 4 is an explanatory diagram of the data structure stored in the data storage unit in the embodiment, in which (a) is the image information storage unit, (b) is the geological map storage unit, (c) is the teacher information storage unit, and (d) is the Prediction result storage unit. 1 is a flowchart showing the overall processing procedure of a geological evaluation process in an embodiment. 5 is a flowchart illustrating the processing procedure of image data creation processing in the embodiment. FIG. 3 is an explanatory diagram illustrating a photographing process in the embodiment. 5 is a flowchart illustrating the processing procedure of teacher data generation processing in the embodiment. FIG. 3 is a diagram for explaining the training data generation process in the embodiment, in which (a) is a photographed image, (b) is a lithology classification diagram, and (c) is a diagram for explaining filtering. 2 is a flowchart illustrating the processing procedure of prediction processing in the embodiment. FIG. 7 is an explanatory diagram illustrating a repulsion degree recording process of a reexamination area in an embodiment. 5 is a flowchart illustrating the processing procedure of correction processing in the embodiment.

An embodiment of a geological evaluation system, a geological evaluation method, and a geological evaluation program will be described below with reference to FIGS. 1 to 11. This embodiment will be described as a geological evaluation system that evaluates the geology of a dam construction site (evaluation target area) and creates a geological map according to the evaluation. In this case, at the same construction site, machine learning is performed using teacher information of the location (learning target area) where a geological specialist has created a geological map, and other locations where a geological specialist has not created a geological map. Evaluate the geology using the learning results.

As shown in FIG. 1, in this embodiment, a geological evaluation device 20 connectable to an imaging device 10 is used.
The imaging device 10 is, for example, a camera, and generates an image of a subject. This imaging device 10 is mounted on an unmanned aircraft 11 such as a so-called drone. This unmanned aircraft 11 is made to fly while scanning at a position at a predetermined distance from the learning target area or the evaluation target area. Then, using the imaging device 10, the entire target area is divided and photographed.

(Hardware configuration)
The hardware configuration of the information processing device H10 that constitutes the geological evaluation device 20 will be explained using FIG. 2. The information processing device H10 includes a communication device H11, an input device H12, a display device H13, a storage section H14, and a processor H15. Note that this hardware configuration is just an example, and it can also be implemented using other hardware.

The communication device H11 is an interface that establishes a communication path with other devices and executes data transmission and reception, and is, for example, a network interface card, a wireless interface, or the like.

The input device H12 is a device that allows an operator to input instructions necessary for geological evaluation, information on a geological map created by a geological specialist, geological information to be corrected, etc., and is, for example, a mouse or a keyboard. The display device H13 is a display or the like that displays various information, and outputs a geological map or the like that maps the predicted (evaluated) geology in the evaluation target area.

The storage unit H14 is a storage device that stores data and various programs for executing various functions of the geological evaluation device 20. Examples of the storage unit H14 include ROM, RAM, hard disk, and the like. Further, data may be stored using a cloud as the storage unit H14.

The processor H15 controls each process in the geological evaluation device 20 using programs and data stored in the storage unit H14. Examples of the processor H15 include a CPU, an MPU, and the like. This processor H15 develops a program stored in a ROM or the like into a RAM and executes various processes for geological evaluation.

The processor H15 is not limited to performing software processing for all processes that it executes. For example, the processor H15 may include a dedicated hardware circuit (for example, an application-specific integrated circuit: ASIC) that performs hardware processing for at least part of the processing that it executes. That is, the processor H15 includes (1) one or more processors that operate according to a computer program (software), (2) one or more dedicated hardware circuits that execute at least some of various processes, or ( 3) It can be configured as a circuit including a combination thereof. A processor includes a CPU and memory, such as RAM and ROM, where the memory stores program codes or instructions configured to cause the CPU to perform processing. Memory or computer-readable media includes any available media that can be accessed by a general purpose or special purpose computer.

(Function of geological evaluation device 20)
The functions of the geological evaluation device 20 will be explained using FIG. 1. This geological evaluation device 20 is a computer system that evaluates the geology of an evaluation target area. This geological evaluation device 20 includes a control section 21, an image information storage section 22, a geological map storage section 23 as a geological information storage section, a teacher information storage section 24, a learning result storage section 25, and a prediction result storage section 26.

The control unit 21 includes a control means (CPU, RAM, ROM, etc.) and performs processing (including an image management stage, a learning stage, a prediction stage, a geological map creation stage, a correction stage, etc.) to be described later. For this purpose, by executing the geological evaluation program stored in the memory, the control unit 21 functions as an image management unit 211, a learning unit 212, a prediction unit 213, a geological map creation unit 214, a correction unit 215, and the like.

The image management unit 211 executes processing for managing images taken by the imaging device 10.
The learning unit 212 executes learning processing for evaluating geology by machine learning using teacher information. Here, the learning unit 212 uses the photographed image and the geological map information to learn to generate a prediction model that receives the photographed image as input and outputs the geological map information. In this embodiment, for example, CNN (Convolutional Neural Network) is used as machine learning, but it is not limited to this.

The prediction unit 213 uses the learning results obtained by the learning unit 212 to execute a process of evaluating geology. In a CNN, information inputted in an input layer is used to make a final determination based on features in a fully connected layer via a multi-layered convolution layer and a pooling layer. The convolutional layer captures the characteristics of local regions of rock quality and rock grade through various filters. In the pooling layer, the feature points obtained in the convolution layer are replaced with representative numerical values. In this embodiment, the prediction unit 213 holds learning results for determining rock quality and rock class, respectively. In addition, in this embodiment, in the fully connected layer, a classifier (for example, SVM: Support Vector Machine) is used to compare analysis results with lithology or lithology according to the feature values output from the convolution layer and the pooling layer. and determine the threshold for analysis.

The geological map creation unit 214 creates a geological map based on the geological information predicted by the prediction unit 213, and executes processing to output this geological map. In this embodiment, the geological map creation unit 214 creates a predicted rock quality classification map and a predicted rock class classification map, which will be described later.
The modification unit 215 executes a process of modifying the predicted rock class information.

As shown in FIG. 3A, the image information storage unit 22 stores image management data used for learning and prediction for geological evaluation. This image management data is recorded when image data is generated by photographing the exposed ground (ground surface) at a dam construction site in image data creation processing to be described later. Each image management data includes photographed image data 220 and whole image data 225.

The photographed image data 220 is image data of a photographed image (full color image) of a subject. This photographed image data 220 records data regarding the position (coordinates) at which the photographed image was photographed. Based on this photographing position, an entire image can be created by arranging the photographed images.
The entire image data 225 is image data of the entire learning target range or the entire evaluation target range created using the captured image data 220.

As shown in FIG. 3(b), the geological map storage unit 23 stores data related to a geological map (geological information) created by a geological expert. This geological map is recorded when a geological expert creates and inputs the geological map. The geological map storage unit 23 stores rock type classification map data 231 and rock class classification map data 232. Each section map data is associated with data regarding the spatial arrangement (latitude, longitude, altitude) as a geological location.

The lithology classification map data is a map in which lithology types are classified and displayed for the learning target area. Rock classifications include volcanic rocks, sedimentary rocks, and metamorphic rocks. In this case, the lithology classification map includes the position corresponding to the marking in the photographed object. Here, the markings are made with paint, rivets, etc. provided by reference line surveying, and are placed at predetermined equal intervals of several meters.

The rock class classification map data is a map in which rock classes are classified and displayed for the learning target area. Rock class classifications include A class rock, B class rock, CH class rock, CM class rock, CL class rock, D class rock, etc. In this case, the rock class classification map includes the position corresponding to the marking in the photographed object.

As shown in FIG. 3(c), the teacher information storage unit 24 stores teacher information for performing machine learning. This teacher information is recorded before learning processing is performed. The teacher information stores rock quality teacher information 241 and rock grade teacher information 242. These teacher information (241, 242), for example, are divided into grids of a predetermined learning unit size (for example, 0.5 m x 0.5 m) and created in grid units.

The rock quality teacher information 241 includes a grid identifier, a grid position, a photographed image, and rock quality information.
The rock grade teacher information 242 includes a grid identifier, a grid position, a photographed image, and rock grade information.

The grid identifier is an identifier for specifying each grid as a unit area.
The grid position is a coordinate for specifying the position of this grid.
The photographed image is an image taken of the ground surface at the position of this grid.
The rock quality information and rock class information are the rock quality classification and rock class classification at the position of this grid.

The learning result storage unit 25 stores a predictive model (learning result) for creating a geological map. This learning result is recorded when the learning process described later is executed. In this embodiment, prediction models (a rock quality prediction model and a rock class prediction model) for respectively predicting rock quality and rock class are stored using image information for each grid.

As shown in FIG. 3(d), the prediction result storage unit 26 stores prediction result information 260 and reconsideration information 265 predicted based on the photographed image. This prediction result information 260 is recorded when prediction processing is performed for each area of a specific size in the evaluation target range. Here, each region to be predicted uses a range (grid) of the same size as the grid used for the teacher information. Reconsideration information 265 is recorded when there is a possibility of modifying prediction result information 260. The prediction result information 260 includes data regarding a grid identifier, grid position, divided image to be evaluated, predicted rock quality, and predicted rock class, and the reconsideration information 265 includes data regarding repulsion degree and corrected rock class.

The grid identifier is an identifier for specifying each grid indicating each prediction result.
The grid position is a coordinate for specifying the position of this grid. This grid position is recorded as a different position from the grid position of the teacher information.
The evaluation target divided image is a captured image (evaluation target) of this grid.
The predicted lithology is the lithology predicted for this grid.
The predicted rock grade is the predicted rock grade for this grid.
The degree of repulsion is the degree of repulsion determined by the Schmidt hammer test at this grid position. This resilience is obtained in a hammer test to confirm the predicted rock grade.
The revised rock class is a rock class that is a modification of the predicted rock class.

Next, the processing procedure when geology is evaluated by the geological evaluation device 20 will be explained using FIGS. 4 to 11. In this embodiment, a learning process, a prediction process, a correction process, and a relearning process are performed in order. Here, the learning process is a process for learning geological evaluation from captured images, and the prediction process is a process for performing geological evaluation in the evaluation target area. The correction process is a process of correcting the predicted geological evaluation, and the relearning process is a process of learning again using the corrected information.

<Learning process>
As shown in FIG. 4, in the learning process, image data generation processing (step SA1), teacher data generation processing (step SA2), and learning processing (step SA3) are performed.

(Image data creation process)
As shown in FIG. 5, in the image data creation process (step SA1), photographing preparation work is performed (step S1-1). Specifically, the ground surface to be studied is wetted, cleaned, and dried, and the wet/dry state and cleaning state of the surface are unified in a predetermined state. Furthermore, in order to correspond to the geological map, markings are provided in the photographic target range by reference line surveying at lengths that are integral multiples of the divided images.

Next, a photographing operation is performed (step S1-2). Here, the unmanned aircraft 11 is mounted with the imaging device 10 and is flown at a predetermined time in the daytime when the solar radiation conditions are the same (for example, from 12:00 to 1:00 on a sunny day). In this case, the unmanned aircraft 11 is operated to fly at a predetermined height so that the imaging device 10 scans the ground surface to be studied.

As shown in FIG. 6, after photographing range A1, the imaging device 10 of the unmanned aerial vehicle 11 photographs the next range A2 so as to overlap this range A1. Then, the imaging device 10 stores the photographed image in the memory in association with the photographing position.

After that, the control unit 21 of the geological evaluation device 20 executes a captured image acquisition process (step S1-3). Specifically, the image management section 211 of the control section 21 acquires a photographed image from the imaging device 10 and records it in the image information storage section 22 as photographed image data 220.

Next, the control unit 21 of the geological evaluation device 20 executes an entire image generation process (step S1-4). Specifically, the image management unit 211 of the control unit 21 arranges the plurality of acquired captured images using the position information of the captured images so that the peripheral portions of the captured images overlap. In this case, about 80% of the periphery of one image is overlapped with the adjacent image. Then, the image management unit 211 connects each image at the overlapping portion, generates whole image data 225 with the learning target area as one image, and stores it in the image information storage unit 22.

(Teacher data generation process)
Next, the teacher data generation process (step SA2) will be explained using FIG.
First, the control unit 21 of the geological evaluation device 20 executes a process for identifying divided images of each grid (step S2-1). Specifically, the learning unit 212 of the control unit 21 analyzes the entire image according to the size of one grid based on the markings reflected in the entire image data 225 stored in the image information storage unit 22. Generate divided images divided into each grid. Then, the learning unit 212 assigns a grid identifier to each grid, and stores it in the memory in association with the position of the grid and each divided image.

Next, the control unit 21 of the geological evaluation device 20 executes a process for identifying the geological map of each grid (step S2-2). Specifically, the learning unit 212 of the control unit 21 identifies the area of the geological map (lithology classification map and rock class classification map) corresponding to each grid based on the markings.

Next, the control unit 21 of the geological evaluation device 20 executes a process of associating the divided images with the geological map (step S2-3). Specifically, the learning unit 212 of the control unit 21 associates the divided images and the geological map that are associated with the same grid and stores them in the memory.

For example, a divided image obtained by dividing the captured image shown in FIG. 8(a) into each grid is associated with the rock mass grade of each grid shown in FIG. 8(b) based on the grid position. Note that FIG. 8(b) is a rock class classification diagram divided into grids, with the shaded portions being CH class and the other portions being CM class.

Next, the control unit 21 of the geological evaluation device 20 executes grid filtering processing (step S2-4). Specifically, the learning unit 212 of the control unit 21 extracts only divided images to which a single division is assigned, excluding divided images that include multiple divisions, in the geological map corresponding to each divided image. do. Here, grids that include multiple rock classifications and grids that include multiple rock class classifications are deleted. For example, in Figure 8(c), the shaded grid includes multiple rock class classifications, so the divided images and geological map (rock class information) associated with the shaded grid are excluded. Specify the divided images and geological map associated with the grid as teacher information.

Then, the control unit 21 of the geological evaluation device 20 executes recording processing as teacher information (step S2-5). Specifically, the learning unit 212 of the control unit 21 stores the specified teacher information in the teacher information storage unit 24.

(Learning process)
Next, a learning process (step SA3) for generating a predictive model will be explained.
First, the learning unit 212 of the control unit 21 performs deep learning using divided images of teacher information as an input layer and geological information as an output layer. Here, the learning unit 212 generates a lithology prediction model for predicting lithology classification (lithology information) in an area of the grid size by learning using the divided images and lithology classification map of the teacher information. do. Further, the learning unit 212 generates a rock class prediction model for predicting rock class classification (rock class information) in an area of the grid size by learning using the divided images and the rock class classification map of the teacher information. , are recorded in the learning result storage section 25.

<Prediction process>
As shown in FIG. 4, in the prediction process, image data creation processing (step SB1), prediction processing (step SB2), and geological map display processing (step SB3) are executed in order.

In the image data creation process (step SB1), similarly to the image data creation process (step SA1) in the learning process, photographing preparation work, photographing work, photographed image acquisition processing, and overall image generation processing are executed. As a result, the entire image to be evaluated is generated.

(prediction processing)
The prediction process (step SB2) will be explained using FIG. 9.
First, the control unit 21 of the geological evaluation device 20 executes division processing of the evaluation target image (step S3-1). Specifically, the prediction unit 213 of the control unit 21 generates an evaluation target divided image in which the entire image to be evaluated is divided into grids based on the intervals of markings, and stores the divided images in memory in association with the positions of the grids. .

Next, the control unit 21 of the geological evaluation device 20 executes geological information prediction processing for each grid (step S3-2). Specifically, the prediction unit 213 of the control unit 21 acquires divided images for each grid, and predicts geological information using the geological prediction model.

In this case, the prediction unit 213 acquires the rock quality prediction model from the learning result storage unit 25, and predicts the rock quality in each grid using the rock quality prediction model using the divided images as an input layer. Then, the predicted rock quality is stored in the prediction result storage unit 26 in association with the grid identifier of the grid, the grid position, and the evaluation target divided image. Furthermore, the prediction unit 213 acquires the rock class prediction model from the learning result storage unit 25, predicts the rock class in each grid using the rock class prediction model using the divided images as an input layer, and calculates the grid identifier of the grid, The prediction result is stored in the prediction result storage unit 26 in association with the grid position and the divided image to be evaluated.

(Geological map display processing)
As shown in FIG. 4, when prediction for all divided images included in the evaluation target area is completed, the control unit 21 of the geological evaluation device 20 executes a geological map display process (step SB3). Specifically, the geological mapping unit 214 of the control unit 21 maps the geological information (lithology and rock class) of each grid stored in the prediction result storage unit 26 using the coordinates of the grid, and Generate a predicted lithology classification map and a predicted rock class classification map for the entire area. Here, regarding rock quality and rock class, the same type is specified according to the identifier of each type, and the range (classification) of each type is specified. Then, the geological map creation section 214 of the control section 21 displays each created section map on the display device H13. In this case, each grid is displayed in a state that can be specified arbitrarily on the rock class classification map.

<Correction process>
Next, as shown in FIG. 4, a correction process is performed. Here, processing for obtaining a reexamination area in the geological map (step SC1), correction processing for correction (step SC2), and redisplay processing (step SC3) are executed in order.

In the reexamination area acquisition process (step SC1), the worker first checks the lithology classification map and rock class classification map displayed on the display device H13, and makes predictions based on the lithology and continuity state. Specify the grid of locations where the rock grade is in doubt. In this case, the control unit 21 of the geological evaluation device 20 generates reconsideration information 265 in which no data is recorded in association with the prediction result information 260 of the designated grid.

Then, as shown in FIG. 10, an instruction to hit the rock in the reexamination area with the rock inspection hammer 40 is displayed on the display device H13. The operator performs a hammer test in the indicated review area. This rock inspection hammer 40 is connected to a measuring device that measures the repulsive force (degree of repulsion) from the rock surface. When the rock surface is struck by the rock inspection hammer 40, the rock inspection hammer 40 measures the impact response and calculates the degree of repulsion. In this case, the degree of repulsion is associated with the impact position (grid position) of the rock at which the impact was made.

Next, as shown in FIG. 11, in the correction process (step SC2), the control unit 21 of the geological evaluation device 20 executes the recording process of the degree of repulsion in association with the impact position (step S4-1). Specifically, when the worker inputs the degree of repulsion at a position corresponding to the position of a grid in the reexamination area, the correction unit 215 of the control unit 21 converts the repulsion degree of the grid identifier of this grid into reexamination information. 265.

Next, the control unit 21 of the geological evaluation device 20 executes a bedrock grade correction process (step S4-2). Specifically, the correction unit 215 of the control unit 21 displays the predicted rock classification map and the repulsion degree of the reexamination area. Then, the user checks the displayed rock classification map and repulsion degree, and inputs the rock mass grade to be corrected into the grid of the reexamination area. The modification unit 215 records the input rock class as a modified rock class in the reconsideration information 265 of the grid identifier that has been modified and input.

Next, the control unit 21 of the geological evaluation device 20 executes a redisplay process (step SC3). Specifically, the geological map creation unit 214 of the control unit 21 executes the same process as step SB3 and displays each created divisional map on the display device H13.

<Relearning process>
Next, as shown in FIG. 4, a relearning step is performed. This step is executed when the corrected rock class of the reexamination information 265 is recorded in the correction process (step SC2) of the correction process.

In this relearning process, the control unit 21 of the geological evaluation device 20 executes a teacher data update process (step SD1). Specifically, the learning unit 212 of the control unit 21 extracts the reconsideration information 265 in which the degree of repulsion is recorded and the prediction result information 260 associated therewith from the prediction result storage unit 26. Then, the learning unit 212 generates new teacher information that associates the grid identifier, grid position, and evaluation target divided image of the extracted prediction result information 260 with the revised rock class of the reexamination information 265, and includes this grid identifier. It is recorded in the teacher information storage section 24 as rock class teacher information 242.

Next, the control unit 21 of the geological evaluation device 20 executes a learning process similarly to step SA3 (step SD2). In this case, the learning unit 212 of the control unit 21 re-creates the rock class prediction model for predicting the rock class classification in the area of the grid size by learning using the divided images and the rock class classification map of the teacher information. The rock class prediction model is generated and recorded in the learning result storage unit 25.

(effect)
The control unit 21 of the geological evaluation device 20 executes a learning process using teacher information that associates a geological map containing a single piece of geological information with a divided image. This makes it possible to accurately predict the geological map of the grid.

According to this embodiment, the following effects can be obtained.
(1) In the present embodiment, the control unit 21 of the geological evaluation device 20 executes a learning process for each grid using teacher information that associates divided images with a geological map containing single geological information, and Generate a predictive model. Then, the control unit 21 predicts geological information using a geological prediction model and an evaluation target divided image obtained by dividing an image of the ground surface to be evaluated into grids of the same size as the grid of the teacher data. This allows accurate geological information to be predicted for the grid. In addition, the burden on geological specialists can be reduced and geological evaluations can be carried out efficiently and appropriately.

(2) In the present embodiment, the control unit 21 of the geological evaluation device 20 associates divided images and geological information based on markings. This makes it possible to efficiently associate image information and geological map information and generate teacher data.

(3) In this embodiment, the control unit 21 of the geological evaluation device 20 predicts rock quality and rock class for each grid, and creates a geological map. Thereby, the photographed image can be output as a geological map.

(4) In this embodiment, the wet/dry conditions and cleaning conditions of the ground surface are unified in the photographing preparation work (step S1-1), and photographs are taken at the same time in the photographing work. Accurate machine learning can be performed by eliminating unevenness in photographed images due to dry/wet conditions, sunlight irradiation conditions, and differences due to cleaning conditions.

(5) In the present embodiment, the control unit 21 of the geological evaluation device 20 generates whole image data 225 by overlapping and connecting the peripheral parts of the photographed images, and generates divided images in which the whole image is divided into grids. do. Thereby, the entire image can be generated using the central region in the photographed image where less distortion occurs.

(6) In the present embodiment, the control unit 21 of the geological evaluation device 20 executes a learning process using the updated predicted rock mass grade and the evaluation target divided images as teacher data (step SD2). Thereby, it is possible to perform re-learning using the corrected teaching data and generate an accurate predictive model.

(7) In this embodiment, the degree of repulsion in a grid where the predicted rock mass grade is questionable is acquired. As a result, it is only necessary to measure the degree of repulsion of the grid in question, and it is possible to efficiently identify the correct rock mass grade.

This embodiment can be modified and implemented as follows. This embodiment and the following modified examples can be implemented in combination with each other within a technically consistent range.
- In the above embodiment, the control unit 21 of the geological evaluation device 20 unifies the wet/dry state and cleaning state of the surface in the photographing preparation process (step S1-1) of the image data creation process (step SA1), and In (step S1-2), images were taken at a predetermined time so that the solar radiation conditions were the same. Instead of adjusting the conditions at the time of photographing, the white balance of the imaging device 10 may be adjusted so that the color correction samples have substantially the same color.

- In the above embodiment, the divided images of each grid are associated with the geological map based on the markings. The correspondence between the grid divided image and the geological map is not limited to the method using markings, as long as the mutual positions can be adjusted. For example, longitude, latitude, etc. may be used.

- In the above embodiment, the learning results are used to evaluate the geology of places where geological specialists have not created geological maps. Here, it has been found that the correct answer rate is better if the grids serving as teacher data are scattered near the grid to be evaluated. Therefore, it is preferable that the photographed images serving as the teacher data are provided discretely within the evaluation target area.

- In the above embodiment, the control unit 21 of the geological evaluation device 20 executed the learning process using a geological map created by a geological expert. The geological information associated with the geological location is not limited to a geological map, and may be geological information at a plurality of points determined by a geological specialist in the evaluation target area.

- In the above embodiment, the control unit 21 of the geological evaluation device 20 calculates a geological prediction model by deep learning using CNN. The machine learning used by the control unit 21 is not limited to this. For example, a geological prediction model may be calculated that configures a network in which various pieces of information included in captured images and geological information are interconnected while being weighted. In this case, a template prediction model (eg, template network) for predicting geological information from captured images is prepared. For example, as this template prediction model, a network in which nodes are connected using statistical weighting values is used so that it can be used for general purposes. Then, images taken at multiple locations in the learning target area are input into the template prediction model, and the output values from the template prediction model are compared with the values of the geological map created by a geological specialist engineer, and the two are found to be close. Repeat the feedback to adjust the weighting. Then, when the differences between the two at a plurality of positions are included in an allowable range, this adjusted prediction model (network) is used as a geological prediction model.

A1, A2... range, 10... imaging device, 11... unmanned aircraft, 20... geological evaluation device, 21... control unit, 22... image information storage unit, 24... teacher information storage unit, 25... learning result storage unit, 26... Prediction result storage section, 40...Rock inspection hammer, 211...Image management section, 212...Learning section, 213...Prediction section, 214...Geological map creation section, 215...Correction section, 220...Photographed image data, 225...Whole image data , 241...Rock quality teacher information, 242...Rock grade teacher information, 260...Prediction result information, 265...Reconsideration information.

Claims (7)

an image information storage unit that stores a captured image of the ground surface in association with a ground surface position;
a geological information storage unit storing at least one of a rock quality classification and a rock class classification in association with a ground surface position;
A geological evaluation system that executes geological evaluation, comprising a control unit that predicts the classification ,
The control section,
dividing the captured image stored in the image information storage unit into unit areas in association with the ground surface position;
obtaining a section associated with the unit area from the geological information storage unit in correspondence with the ground surface position;
generating a prediction model for predicting the division based on the photographed image, using the captured image and the division associated with the unit area to which only a single division is associated as training information;
A geological evaluation system characterized by predicting divisions in the evaluation target area using captured images of ground surface positions of the evaluation target area and the prediction model. The control unit includes:
Connecting the captured images stored in the image information storage unit to generate a whole image with the learning target area as one image,
2. The geological evaluation system according to claim 1, wherein when dividing into the unit areas, the generated entire image is divided in association with the ground surface position. The image information storage unit stores a captured image of the ground surface in association with a ground surface position based on markings on the ground surface provided by reference line surveying,
The geological information storage unit stores the classification in association with the ground surface position based on the marking on the ground surface,
The control unit includes:
The geological evaluation system according to claim 1 or 2, wherein the division into the unit areas is performed in association with the position of the marking. The method is characterized in that, as the photographed image of the teacher information and the photographed image of the evaluation target area , an image of the marked ground surface is photographed under predetermined conditions such as a cleaning state, a wet/dry state, and a sunlight irradiation state of the ground surface. The geological evaluation system according to claim 1. When the control unit obtains a section different from the predicted section among the predicted sections , the control section generates new teacher information that associates the newly obtained section with the photographed image of the unit area; The geological evaluation system according to claim 1 or 4 , wherein the prediction model is regenerated using the new teacher information. an image information storage unit that stores a captured image of the ground surface in association with a ground surface position;
a geological information storage unit storing at least one of a rock quality classification and a rock class classification in association with a ground surface position;
A method of performing geological evaluation using a geological evaluation system comprising a control unit that predicts the classification ,
The control section,
dividing the captured image stored in the image information storage unit into unit areas in association with the ground surface position;
obtaining a section associated with the unit area from the geological information storage unit in correspondence with the ground surface position;
generating a prediction model for predicting the division based on the photographed image, using the captured image and the division associated with the unit area to which only a single division is associated as training information;
A geological evaluation method characterized by predicting divisions in the evaluation target area using a photographed image of the ground surface position of the evaluation target area and the prediction model. an image information storage unit that stores a captured image of the ground surface in association with a ground surface position;
a geological information storage unit storing at least one of a rock quality classification and a rock class classification in association with a ground surface position;
A program that executes geological evaluation using a geological evaluation system comprising a control unit that predicts the classification ,
The control section,
dividing the captured image stored in the image information storage unit into unit areas in association with the ground surface position;
obtaining a section associated with the unit area from the geological information storage unit in correspondence with the ground surface position;
generating a prediction model for predicting the division based on the photographed image, using the captured image and the division associated with the unit area to which only a single division is associated as training information;
A geological evaluation program characterized in that the geological evaluation program functions as a means for predicting divisions in the evaluation target area using a captured image of the ground surface position of the evaluation target area and the prediction model.