Method for estimating state of hull

Abstract

According to the present invention, on the basis of an image, obtained by imaging a mesh to be evaluated on a surface of a ship to be evaluated, and a three-dimensional model of a hull, which is information for estimating a water flow vector generated in the mesh while the ship is sailing, the contribution of the mesh to a frictional resistance value correlated with the required horsepower of the ship is specified.

Description

Methods for evaluating the condition of a ship's hull 【0001】 This invention relates to a technique for evaluating the condition of a ship's hull. 【0002】 Patent Document 1 describes setting a 3D model showing the shape of an object, calculating a simulated image based on the 3D model and radar parameters, outputting training data that associates the simulated image with the type of object, and performing deep learning using the training data. Patent Document 2 describes using an imaging means suspended into the sea from a work vessel along with a work member to image a target mark attached to a work object laid on the seabed or in the sea, processing the obtained image to generate a virtual 3D model including the work member and the work object, generating an image of the virtual 3D model viewed from a desired direction, and using it to support the work member of the work vessel in processing the work object. 【0003】 Japanese Patent Publication No. 2020-3379 Japanese Patent Publication No. 2001-180579 【0004】 Methods for evaluating the condition of a ship's hull include, for example, having divers photograph parts of the hull with cameras and determining the degree of fouling based on the captured images, or analyzing the ship's operational data to determine the degree of fouling from changes in propulsion performance. However, the former method makes it difficult to objectively and quantitatively evaluate the overall fouling condition of the hull. The latter method only provides a rough indication of the overall fouling condition of the ship, including the hull and propellers. Therefore, these methods make it difficult to accurately evaluate the condition of the hull. 【0005】 The present invention aims to improve the accuracy of evaluating the condition of a ship's hull. 【0006】 One aspect of the present invention provides a method for identifying the contribution of a certain area to the required horsepower of a vessel or a required horsepower index, which is a value correlated with the required horsepower, based on an image taken of a certain area of ​​the surface of a vessel to be evaluated and water flow information, which is the vector of the water flow generated in that area while the vessel is sailing or information for estimating that vector. 【0007】The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated and the position of the partial region in the three-dimensional shape. The process for identifying the contribution may include the process of obtaining the required horsepower index output by a trained model, which has been trained using training data that includes an image of an arbitrary partial region of the surface of an arbitrary vessel, the three-dimensional shape of the entire surface of the submerged portion of the vessel, and the position of the partial region in the three-dimensional shape as explanatory variables, and the required horsepower index for the partial region as the objective variable. The trained model receives an image of an arbitrary partial region of the vessel to be evaluated, the three-dimensional shape of the entire surface of the submerged portion of the vessel, and the position of the partial region in the three-dimensional shape as input to the trained model. 【0008】 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated, the position of the partial region in the three-dimensional shape, and water behavior information which is information about the behavior of water around the vessel to be evaluated. The process of identifying the contribution may also include the process of obtaining the required horsepower index output by a trained model that has been trained using training data in which the explanatory variables include an image taken of an arbitrary partial region of the surface of an arbitrary vessel, the three-dimensional shape of the entire surface of the submerged portion of the vessel, the position of the partial region in the three-dimensional shape, and water behavior information regarding the behavior of water around the vessel, and the required horsepower index for the partial region is input to a trained model that has been trained using machine learning, and in which the image taken of the partial region of the vessel to be evaluated, the three-dimensional shape of the entire surface of the submerged portion of the vessel, the position of the partial region in the three-dimensional shape, and water behavior information regarding the behavior of water around the vessel are input. 【0009】 The process for identifying the contribution may include a process for identifying at least one of the roughness and frictional resistance coefficient of a portion of the vessel being evaluated based on an image of that portion of the vessel being evaluated, and a process for identifying the contribution based on at least one of the roughness and frictional resistance coefficient and water flow information relating to that portion of the vessel being evaluated. 【0010】The process of identifying at least one of the roughness and the frictional resistance coefficient may include a process of inputting an image of a portion of the area to be evaluated of the ship to be evaluated into a trained model that has been trained using training data in which an image of an arbitrary object contaminated by immersion in water is taken as an explanatory variable, and at least one of the roughness and frictional resistance coefficient of the object is taken as an objective variable, in order to obtain at least one of the roughness and frictional resistance coefficient output by the trained model. 【0011】 The method described above may include a process of identifying at least one of the roughness and frictional resistance coefficient at a time different from the time when the partial area of ​​the vessel being evaluated was photographed, based on at least one of the roughness and frictional resistance coefficient identified with respect to a partial area of ​​the vessel being evaluated with respect to each of several different time periods, and a process of identifying the contribution of the partial area to the required horsepower index of the vessel being evaluated at a time different from the time when the partial area of ​​the vessel being evaluated was photographed, based on at least one of the roughness and frictional resistance coefficient identified with respect to a time different from the time when the partial area of ​​the vessel being evaluated was photographed, and water flow information relating to the partial area. 【0012】 The process of identifying at least one of the roughness and the frictional resistance coefficient may include a process of identifying contamination information, which is information relating to one or more types of contamination in a part of the ship being evaluated and the area occupied by each of those one or more types of contamination in the part, based on an image taken of that part of the ship being evaluated, and a process of identifying at least one of the roughness and the frictional resistance coefficient for that part based on the contamination information. 【0013】 The process for identifying the contamination information may include a process for obtaining contamination information output by a trained model that has been trained using training data in which images of an arbitrary object contaminated by immersion in water are included as explanatory variables and contamination information of the object is included as an objective variable, by inputting images of a part of the ship to be evaluated. 【0014】The method described above may include: a process of identifying fouling information at a time different from the time when the partial area of ​​the vessel under evaluation was photographed, based on fouling information identified with respect to a partial area of ​​the vessel under evaluation with respect to each of several different time periods; a process of identifying at least one of roughness and frictional resistance coefficient with respect to a partial area of ​​the vessel under evaluation with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed, based on fouling information identified with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed; and a process of identifying the contribution of the partial area to the required horsepower index of the vessel under evaluation at a time different from the time when the partial area was photographed, based on at least one of roughness and frictional resistance coefficient identified with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed and water flow information with respect to the partial area. 【0015】 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated and the position of the partial region in the three-dimensional shape, and the process for identifying the contribution may include the process of identifying the vector of the water flow generated in the partial region while the vessel is sailing, based on the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated and the position of the partial region of the vessel in the three-dimensional shape, and the process of identifying the contribution based on at least one of the roughness and frictional resistance coefficient identified based on an image taken of the partial region and the vector of the water flow. 【0016】 The process of identifying the water flow vector may include a process of obtaining the water flow vector output by a trained model that has been trained using training data in which the three-dimensional shape of the entire surface of the submerged portion of any ship and the position of any part of the ship in the three-dimensional shape are included as explanatory variables, and the water flow vector generated in the part of the ship while the ship is sailing is included as the objective variable, by inputting the three-dimensional shape of the entire surface of the submerged portion of the ship to be evaluated and the position of the part of the ship to be evaluated in the three-dimensional shape. 【0017】The process for identifying the water flow vector may include a process for identifying the water flow vector generated in a part of the area to be evaluated while the vessel is underway, using CFD with the three-dimensional shape of the vessel being evaluated. 【0018】 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated, the position of the partial region in the three-dimensional shape, and the location and shape of fouling identified from images of the partial region or its surroundings. The process for identifying the contribution may include a process for identifying the vector of water flow generated in the partial region while the vessel is sailing, based on the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated, the position of the partial region of the vessel in the three-dimensional shape, and the location and shape of fouling identified from images of the partial region or its surroundings, and a process for identifying the contribution based on at least one of roughness and frictional resistance coefficient identified from images of the partial region and the vector of water flow. 【0019】 The process of identifying the water flow vector may include a process of obtaining the water flow vector output by a trained model that has been trained using training data in which the three-dimensional shape of the entire surface of the submerged portion of any ship, the position of any part of the ship in the three-dimensional shape, and the position and shape of fouling identified from images of the part or the area surrounding the part are included as explanatory variables, and the water flow vector generated in the part while the ship is sailing is included as the input of the three-dimensional shape of the entire surface of the submerged portion of the ship to be evaluated, the position of the part of the ship to be evaluated in the three-dimensional shape, and the position and shape of fouling identified from images of the part or the area surrounding the part, and which has been trained using training data in which the three-dimensional shape of the entire surface of the submerged portion of the ship to be evaluated, the position of the part of the ship to be evaluated in the three-dimensional shape, and the position and shape of fouling identified from images of the part or the area surrounding the part are input. 【0020】 The method described above may include a process of determining the contribution of the entire submerged portion of the required horsepower index of the vessel by summing the contributions of each of the multiple partial regions of the vessel to be evaluated obtained by dividing the surface of the vessel to be evaluated. 【0021】The method includes a process of identifying the contribution of the entire submerged portion of the hull to the required horsepower index of the ship by integrating the contributions of the partial regions of the plurality of evaluation targets obtained by dividing the surface of the ship to be evaluated, and a process of identifying at least one of the roughness and the friction resistance coefficient for the partial region so that the difference between the contribution of the entire submerged portion of the hull to the required horsepower index of the ship measured during each of a plurality of voyages by the same ship to be evaluated or a plurality of ships of the same ship type, and the value obtained by integrating the contributions of the partial regions of the plurality of evaluation targets constituting the entire submerged portion of the hull of the ship for each voyage decreases. The method may also include a process of adjusting the relationship information indicating the relationship between the fouling information used to identify at least one of the roughness and the friction resistance coefficient from the fouling information and at least one of the roughness and the friction resistance coefficient. 【0022】 The method includes a process of identifying the contribution of the entire submerged portion of the hull to the required horsepower index of the ship by integrating the contributions of the partial regions of the plurality of evaluation targets obtained by dividing the surface of the ship to be evaluated, and a process of adjusting the parameters of the CFD so that the difference between the contribution of the entire submerged portion of the hull to the required horsepower index of the ship measured during each of a plurality of voyages by the same ship to be evaluated or a plurality of ships of the same ship type, and the value obtained by integrating the contributions of the partial regions of the plurality of evaluation targets constituting the entire submerged portion of the hull of the ship for each voyage decreases. 【0023】 Another aspect of the present invention provides a method including a process of performing machine learning using teacher data including, as explanatory variables, an image of an arbitrary partial region of the surface of an arbitrary ship, the three-dimensional shape of the surface of the entire submerged portion of the ship, and the position of the partial region in the three-dimensional shape, and including, as the objective variable, the contribution of the partial region to the required horsepower of the ship or the required horsepower index having a correlation with the required horsepower, to generate or update a learned model. 【0024】Yet another aspect of the present invention provides a method comprising a process of performing machine learning using teacher data that includes, as explanatory variables, an image of an arbitrary partial area on the surface of an arbitrary ship, the three-dimensional shape of the entire surface of the submerged portion of the ship, the position of the partial area in the three-dimensional shape, and water behavior information regarding the behavior of water around the ship, and includes, as an objective variable, the contribution of the partial area to the required horsepower of the ship or a required horsepower index having a correlation with the required horsepower, to generate or update a learned model. 【0025】 Yet another aspect of the present invention provides a method comprising a process of performing machine learning using teacher data that includes, as an explanatory variable, an image of an arbitrary object fouled by immersion in water, and includes, as an objective variable, at least one of the roughness and the friction coefficient of the object, to generate or update a learned model. 【0026】 Yet another aspect of the present invention provides a method comprising a process of performing machine learning using teacher data that includes, as an explanatory variable, an image of an arbitrary object fouled by immersion in water, and includes, as an objective variable, fouling information that is information regarding one or more types of fouling of the object and the occupied area of each of the one or more types of fouling on the object, to generate or update a learned model. 【0027】 Yet another aspect of the present invention provides a method comprising a process of performing machine learning using teacher data that includes, as explanatory variables, the three-dimensional shape of the entire surface of the submerged portion of an arbitrary ship and the position of an arbitrary partial area of the ship in the three-dimensional shape, and includes, as an objective variable, the vector of the water flow that occurs in the partial area while the ship is in navigation, to generate or update a learned model. 【0028】 Yet another aspect of the present invention provides a method comprising a process of performing machine learning using teacher data that includes, as explanatory variables, the three-dimensional shape of the entire surface of the submerged portion of an arbitrary ship, the position of an arbitrary partial area of the ship in the three-dimensional shape, and the position and shape of fouling specified from an image of the partial area or the area around the partial area, and includes, as an objective variable, the vector of the water flow that occurs in the partial area while the ship is in navigation, to generate or update a learned model. 【0029】A further aspect of the present invention provides a method comprising the process of obtaining a plurality of images by photographing different areas of a vessel in different directions using a camera moving underwater, and the process of generating mapping image data from the plurality of images that represent images corresponding to any direction and any area of ​​the surface of the submerged portion of the vessel. 【0030】 A further aspect of the present invention provides a method comprising the process of obtaining a plurality of images by photographing different regions of a ship in different directions using a camera moving underwater, and the process of generating three-dimensional shape data representing the three-dimensional shape of the surface of the submerged portion of the ship from the plurality of images. 【0031】 A further aspect of the present invention provides a method comprising the process of obtaining a plurality of images by photographing different areas of a ship in different directions using a camera moving underwater, and the process of identifying the shooting position of each of the plurality of images from the plurality of images. 【0032】 A further aspect of the present invention provides a method comprising the process of obtaining multiple images by photographing different regions of a vessel in different directions using a camera moving underwater, and the process of measuring the vector of the water flow generated on the surface of the submerged portion of the vessel using a measuring device that moves underwater along with the camera. 【0033】 A further aspect of the present invention provides a method comprising the process of obtaining a plurality of images by photographing different regions of a vessel in different directions using a camera moving underwater, and the process of identifying the vector of the water flow occurring on the surface of the submerged portion of the vessel from the plurality of images. 【0034】 According to the present invention, the accuracy of evaluating the condition of the ship's hull is improved. 【0035】 A diagram showing an example of an evaluation system according to the first embodiment. A diagram showing an example of the configuration of a server device. A diagram illustrating a three-dimensional model of a ship's hull. A diagram explaining a method for identifying mesh fouling information using a machine learning model. A flowchart illustrating a process for quantitatively evaluating the fouling effect. A diagram illustrating a process for correcting roughness and water flow vectors. A diagram showing an example of the configuration of a server device according to the second embodiment. A flowchart illustrating a process for quantitatively evaluating the fouling effect according to the second embodiment. 【0036】 I. First Embodiment 1. Diagram 1 shows an example of the evaluation system 1 according to the first embodiment. The evaluation system 1 takes an image of the entire hull and uses the captured image to evaluate the effect of hull fouling. The evaluation of the effect of hull fouling assesses the impact of hull fouling on the navigation performance of the vessel 10. The evaluation of the effect of hull fouling is performed by quantitatively evaluating the values ​​that change due to hull fouling. 【0037】 The evaluation system 1 comprises an underwater drone 20, a control device 30, and a server device 40. The underwater drone 20 and the control device 30 are connected via a communication cable 2. The control device 30 and the server device 40 are connected via a network 3, such as the Internet. The server device 40 is also connected via the network 3 and a communication satellite 4 to a terminal device (not shown) mounted on the ship 10. This allows the server device 40 to acquire the outputs of various sensors mounted on the ship 10. 【0038】 The underwater drone 20 submerges while the vessel 10 is at anchor and moves underwater, capturing images of all areas of the vessel's hull. Here, the vessel's hull refers to the submerged portion of the vessel's surface, excluding the propellers. The underwater drone 20 is equipped with a camera 22 and a communication interface 23. 【0039】 Camera 22 captures images of various areas of the hull from different directions. The images captured by camera 22 are, for example, video. However, the images captured by camera 22 are not limited to video; for example, they may be multiple still images taken at predetermined time intervals. Camera 22 is an example of a camera that moves underwater according to the present invention. Video and multiple still images are examples of multiple images according to the present invention. Communication IF 23 transmits the images captured by camera 22 to the control device 30 via the communication cable 2. Note that the term "image" here refers to a digital image. 【0040】The control device 30 is installed on the quay and controls the operation of the underwater drone 20 according to the operator's commands. The control device 30 also transmits images received from the underwater drone 20 to the server device 40 via the network 3. 【0041】 The server device 40 is installed on land and managed and operated by an organization that provides shipping services. The server device 40 performs processing to evaluate the fouling effects on the hull based on images taken by the underwater drone 20. 【0042】 Figure 2 shows an example configuration of a server device 40. The server device 40 comprises a processor 41, memory 42, storage 43, communication interface 44, input unit 45, and display unit 46. The processor 41 is connected to the memory 42, storage 43, communication interface 44, input unit 45, and display unit 46 via a bus. 【0043】 The processor 41 controls various parts of the server device 40 and performs various calculations by executing programs. An example of the processor 41 is one or more CPUs (Central Processing Units). The memory 42 is a computer-readable storage medium used as the work area of ​​the processor 41. Programs and data executed by the processor 41 are stored in the memory 42. An example of the memory 42 is ROM (Read Only Memory) or RAM (Random Access Memory). The storage 43 is a computer-readable storage medium that stores various types of data used by the processor 41. An example of the storage 43 is an HHD (Hard Disk Drive) or SSD (Solid State Drive). The communication IF 44 is connected to the network 3 and communicates data with other devices via the network 3 according to a predetermined communication standard. The input unit 45 inputs signals to the processor 41 according to user operations. An example of the input unit 45 is a keyboard or mouse. The display unit 46 displays various types of information under the control of the processor 41. An example of the display unit 46 is a liquid crystal display. 【0044】The processor 41 functions as a generation means 411, an acquisition means 412, an image processing means 413, an identification means 414, a calculation means 415, an output means 416, and a modification means 417 by executing a program stored in memory 42. These functions are software modules realized through the cooperation of software with hardware resources. 【0045】 The generation means 411 generates a machine learning model 431 for identifying hull fouling information from images of the hull. The machine learning model 431 is generated by performing machine learning using training data. The training data consists of images of any object that has been fouled by immersion in water. The object may be a ship hull or a steel plate. The images are pre-labeled with fouling information. The fouling information includes the type of fouling, the degree of fouling, and the area ratio of the fouling. The type of fouling may also be a concept that includes the degree of fouling. The area ratio of the fouling is information about the area occupied by the fouling. The area ratio of the fouling is obtained by calculating the ratio of the fouled portion to the total area. 【0046】 The generation means 411 generates a machine learning model 431 by machine learning using images of any object soiled by immersion in water as explanatory variables and soiling information as the target variable, with these as training data. Machine learning includes deep learning. The generation means 411 stores the generated machine learning model 431 in the storage 43. The generation means 411 may also update the machine learning model 431 by retraining or the like. The machine learning model 431 is an example of a trained model that has been trained according to the present invention. 【0047】 The acquisition means 412 acquires images of the hull of the vessel 10 to be evaluated, which are taken by the underwater drone 20. The acquisition means 412 may receive images of the hull from the control device 30, or it may read images from a storage medium on which images of the hull are stored. 【0048】The image processing means 413 applies image processing to the image of the hull acquired by the acquisition means 412 to generate a three-dimensional model of the hull. This three-dimensional model shows the three-dimensional shape and positional relationship of the surface of the submerged portion of the hull. The three-dimensional model also includes information on the shooting position of the hull image. For example, the image processing means 413 generates the three-dimensional model of the hull using SfM (Structure From Motion) and MVS (Multi-View Stereo). As described above, the image of the hull is obtained by photographing each region of the hull from different directions. Since the image of the hull is taken underwater, it is expected that the quality will be worse than an image taken on land due to the effects of refraction, scattering of auxiliary light, and suspended particles in the water. Therefore, the image processing means 413 also has functions such as scaling and adjusting the brightness and saturation of the image to improve these quality issues. In SfM, the shooting position and direction are estimated based on the correspondence of feature points that are commonly included in multiple images of the hull, and a low-density point cloud is generated by triangulation. In MVS, the depth of each pixel constituting the image is estimated based on the shooting position and shooting direction estimated by SfM, and a high-density point cloud is generated by integrating these. In addition, the image processing means 413 may adjust the scale using a drawing of the ship's hull or the like. 【0049】 The image processing means 413 synthesizes a three-dimensional model of the hull with textures, or captured images with shooting position information, to generate mesh images for each mesh of the hull. The mesh images are orthographic projection images obtained by projection with parallel light from infinity. For example, the image processing means 413 divides the three-dimensional model into multiple meshes to generate divided images, and then generates an orthographic projection image by orthographically transforming each divided image. A mesh is an example of a part of the area to be evaluated according to the present invention. Orthographic projection images are advantageous for recognizing the state of fouling on the hull because they have less distortion due to perspective and angle from the viewpoint. Each mesh of the three-dimensional model of the hull is associated with a mesh image of that mesh. 【0050】Figure 3 illustrates a three-dimensional model of a ship's hull. As shown in Figure 3, this three-dimensional model represents the three-dimensional shape of the submerged portion of the hull. The three-dimensional model is divided into multiple meshes. In the example shown in Figure 3, the mesh shape is, for example, a rectangle, but the mesh shape is not limited to a rectangle and may be other shapes such as a triangle. 【0051】 The identification means 414 identifies contamination information for each mesh based on the mesh image of each mesh generated by the image processing means 413. For example, the identification means 414 inputs the mesh image as an explanatory variable into a machine learning model 431 stored in storage 43, and obtains contamination information output from the machine learning model 431 as an objective variable. The contamination information includes the type of contamination, the degree of contamination, and the area ratio of the contamination. 【0052】 Figure 4 illustrates a method for identifying mesh contamination information using a machine learning model 431. In the example shown in Figure 4, when a mesh image of mesh Z is input to the machine learning model 431, the machine learning model 431 outputs contamination information for mesh Z. The contamination information for mesh Z includes a set of contamination types (rust), severity (severity of this contamination), and area percentage (30%), and a set of contamination types (algae), severity (mild contamination), and area percentage (10%). This indicates that 30% of mesh Z's total area is contaminated with severe rust, 10% is contaminated with mild algae, and the remaining 60% is uncontaminated. Each mesh may contain a single type and degree of contamination, or it may contain multiple types and degrees of contamination. 【0053】 The calculation means 415 calculates the frictional resistance value of each mesh based on the fouling information identified by the identification means 414. The frictional resistance value of a mesh is the contribution of the mesh to a value that correlates with the required horsepower of the ship 10, and is therefore an example of the contribution of a certain range of the required horsepower index of the ship according to the present invention. Required horsepower refers to the horsepower required to obtain a predetermined ship speed. When the mesh of the ship's hull becomes fouled, the frictional resistance value of that mesh increases, and the required horsepower of the ship 10 increases. 【0054】Specifically, the calculation means 415 identifies the roughness of each mesh based on the fouling information. Roughness is a coefficient that indicates the increased frictional resistance force due to fouling of the mesh on the hull surface. Roughness refers to the degree of unevenness on the surface of an object and is defined by the maximum height or the arithmetic mean. Originally, the frictional force that an object receives from a fluid is determined by the viscosity of the fluid and the flow velocity on the surface of the object, based on Newton's law of viscosity. Therefore, theoretically, the roughness of the surface of the object itself does not directly affect the frictional force. However, when protrusions are created on the surface of an object due to fouling, the flow velocity gradient increases, and as a result the resistance force increases. Based on this mechanism, in CFD (Computational Fluid Dynamics) calculations, by defining roughness as a parameter, it is possible to predict the flow velocity distribution on the surface of the object and calculate the frictional force. It should be noted that it is not practical to divide the surface of an object into infinitely fine meshes in CFD calculations. Therefore, a method is often used that estimates the velocity gradient using the principles of fluid dynamics and empirical know-how based on roughness and calculates the frictional force. Furthermore, the calculation means 415 identifies the water flow vector for each mesh based on a three-dimensional model of the hull. The three-dimensional model of the hull shows the three-dimensional shape of the surface of the submerged portion of the hull and the relative position of each mesh in the three-dimensional shape of the hull. The three-dimensional shape of the surface of the submerged portion of the hull and the relative position of the mesh in the three-dimensional shape affect the water flow vector of the mesh. A water flow vector is a vector of water flow generated in the mesh while the vessel 10 is in motion. The water flow vector includes the velocity and direction of the water flow. In one example, the calculation means 415 estimates the water flow vector for each mesh by CFD calculation using the three-dimensional model. In another example, the calculation means 415 may estimate the water flow vector for each mesh based on model testing. Since the water flow vector is information about the vector of water flow generated in the mesh while the vessel 10 is in motion, it is an example of water flow information according to the present invention. Then, the calculation means 415 calculates the frictional resistance value of each mesh based on the roughness and the water flow vector. 【0055】The calculation means 415 calculates the total frictional resistance value of the hull by accumulating the frictional resistance values ​​of each mesh. The total frictional resistance value of the hull is the contribution of the entire submerged portion of the hull to the total frictional resistance value of the ship 10, and is therefore an example of the contribution of the entire submerged portion of the hull to the required horsepower index of the ship according to the present invention. 【0056】 The output means 416 outputs the calculation results calculated by the calculation means 415. These calculation results include the frictional resistance value of each mesh of the hull and the frictional resistance value of the entire hull. For example, the output means 416 stores the frictional resistance value of each mesh of the hull and the frictional resistance value of the entire hull in the storage 43. Note that the output by the output means 416 is not limited to storage in the storage 43, but may also be displayed on the display unit 46, printed, or transmitted to an external device. 【0057】 The correction means 417 corrects the roughness and water flow vector used to calculate the frictional resistance value of the mesh by data assimilation. Data assimilation is a method of correcting predicted data with actual observed data so that the estimation model yields results closer to reality. Data assimilation methods include, for example, optimal interpolation and Kalman filtering. Since roughness depends on the physical properties and not on the ship 10, the correction means 417 applies the roughness correction to all ships 10, including the target ship 10. On the other hand, since the water flow vector depends on the shape of the ship 10, the correction means 417 applies the water flow vector correction to the target ship 10 and other ships 10 of the same type as the target ship 10. 【0058】 2. Operation (1) Quantitative evaluation of fouling effects Figure 5 is a flowchart illustrating the process for quantitatively evaluating fouling effects. Here, the vessel to be evaluated is assumed to be vessel 10A. In this case, before this process is started, the hull of vessel 10A is photographed by an underwater drone 20. The photography of the hull by the underwater drone 20 is performed, for example, while vessel 10A is anchored at a quay or out at sea. 【0059】The underwater drone 20 moves through the water and uses the camera 22 to photograph all areas of the hull that are submerged in water. At this time, the camera 22 photographs each area of ​​the hull from multiple different directions. The underwater drone 20 transmits the images captured by the camera 22 to the control device 30 via the communication IF 23. The control device 30 transmits the images received from the underwater drone 20 to the server device 40 via the network 3. 【0060】 In step S11, the acquisition means 412 acquires the images captured by the underwater drone 20 by receiving them from the control device 30 and stores them in the storage 43. Note that the images captured by the underwater drone 20 do not necessarily have to be transmitted to the server device 40 via the network 3. For example, the control device 30 may store the images received from the underwater drone 20 in a storage medium, and when this storage medium is connected, the server device 40 may read the images captured by the underwater drone 20 from this storage medium. 【0061】 In step S12, the image processing means 413 generates a three-dimensional model of the ship's hull from the image acquired in step S11. For example, the image processing means 413 generates the three-dimensional model of the ship's hull shown in Figure 3 using SfM and MVS. 【0062】 In step S13, the image processing means 413 generates mesh images of each mesh of the hull from the three-dimensional model of the hull generated in step S12. For example, as shown in Figure 3, the image processing means 413 divides the three-dimensional model into multiple meshes to generate divided images, and then performs an orthorectified transformation on each divided image to generate mesh images of each mesh of the hull. 【0063】In step S14, the identification means 414 uses the machine learning model 431 stored in the storage 43 to identify the fouling information of each mesh of the hull from the mesh image generated in step S13. For example, as shown in Figure 4, the identification means 414 inputs the mesh image of mesh Z as an explanatory variable into the machine learning model 431 and obtains the fouling information of mesh Z output from the machine learning model 431 as an objective variable. The fouling information of mesh Z includes a set of fouling type (rust), degree of fouling (severity), and area percentage of this fouling (30%), and a set of fouling type (algae), degree of fouling (mild), and area percentage of this fouling (10%). The identification means 414 repeats this process for all meshes. 【0064】 In step S15, the calculation means 415 calculates the frictional resistance value of each mesh based on the fouling information of each mesh identified in step S14. Here, an example of calculating the frictional resistance value of mesh Z will be given. As shown in Figure 4, the fouling information of mesh Z includes a set of fouling type (rust), degree of fouling (severity), and area percentage of this fouling (30%), and a set of fouling type (algae), degree of fouling (mild), and area percentage of this fouling (10%). The calculation means 415 determines the roughness of the hull mesh Z using the following formula (1). Select the wall function or turbulence model that takes into account the roughness calculated using the above formula (1), and then apply the wall shear stress τ MeshZ The friction velocity is iteratively calculated using CFD, and the calculation is repeated until the water flow vector converges. The water flow vector indicates the velocity and direction of the water flow in mesh Z. Here, the wall shear stress τ is calculated. MeshZ This represents the frictional resistance value in mesh Z. 【0065】 Here, k s|Mesh Z k is the average roughness of mesh Z. s k is the roughness. sis specified based on the type and degree of fouling. The correspondence between the type and degree of fouling and the roughness is determined in advance by literature and experiments. The roughness ks|rust,hev is the roughness corresponding to severe rust in this correspondence. The roughness ks|alga,light is the roughness corresponding to light algae in this correspondence. The roughness k s|flat is the roughness corresponding to the state without fouling in this correspondence. Also, in the above formula (1), in order to perform weighting according to the fouling area ratio, the roughness of each fouling state of mesh Z is multiplied by the fouling area ratio. Since the fouling area ratio of severe rust is 30%, in the above formula (1), 0.3 is multiplied by the term including the roughness ks|rust,hev. Since the fouling area ratio of light algae is 10%, in the above formula (1), 0.1 is multiplied by the term including the roughness ks|alga,light. Since the area ratio of the state without fouling is the remaining 60%, in the above formula (1), 0.6 is multiplied by the term including the roughness k s|flat . Then, by performing weighted averaging on the roughness of the severe rust part, the light algae part, and the part without fouling of mesh Z, the average roughness k s|Mesh Z of mesh Z is obtained. The calculation means 415 calculates the roughness for other meshes in the same manner. 【0066】 In step S16, the calculation means 415 calculates the frictional resistance value of the entire hull by integrating the frictional resistance values of all the meshes calculated in step S15. The calculation means 415 calculates the frictional resistance value of the entire hull according to the following formula (2). 【0067】 Here, D f|Ship is the frictional resistance value of the entire hull, τ Mesh is the frictional resistance value of each mesh of the hull, and S is the area of each mesh. By integrating the frictional resistance values τ Mesh of all the meshes of the hull, the frictional resistance value D f|Ship of the entire hull is obtained. 【0068】In step S17, the output means 416 outputs the frictional resistance values ​​of each mesh of the hull calculated in step S15 and the frictional resistance value of the entire hull calculated in step S16. For example, the output means 416 stores the frictional resistance values ​​of each mesh of the hull and the frictional resistance value of the entire hull in the ship database stored in the storage 43. 【0069】 (2) Roughness Correction Figure 6 is a diagram illustrating the process of correcting roughness and water flow vectors. The correction means 417 corrects the roughness used to calculate the frictional resistance value of the mesh by data assimilation. As a result, the roughness is adjusted so that the frictional resistance value of the entire hull is closer to the observed value. Here, the target vessel 10 is assumed to be vessel 10A. 【0070】 The correction means 417 uses the frictional resistance value of the entire hull of the vessel 10A, calculated in step S16 described above, as prediction data. The correction means 417 also calculates the frictional resistance value of the entire hull based on the operational profile of the vessel 10A using the following formula (3), and uses this as observation data. 【0071】 Here, D f|Ship_obs D is the observed frictional resistance value of the entire hull, BHP is the horsepower of the main engine, η is the propulsion efficiency, V is the ship speed, and F is the resistance element due to sea weather, etc., excluding the increase in frictional resistance due to fouling. The horsepower of the main engine BHP is measured by a shaft horsepower meter installed on the ship's main engine. The propulsion efficiency η is a value specific to ship 10. The ship speed V is measured by a speedometer installed on ship 10. The resistance element F is calculated using a known method using values ​​obtained from the operational profile and sea weather observed along the ship 10's route. Observation data D f|Ship_obs This is the frictional resistance value of the entire hull measured while the vessel 10A was in motion, and is therefore an example of the contribution of the entire submerged portion of the hull to the required horsepower index of the vessel measured during motion according to the present invention. 【0072】 Furthermore, in order to further improve the accuracy of the correction, the correction means 417 uses the observation data D f|Ship_obsThe effects of propeller fouling and wind-induced thrust may be removed. For example, the correction means 417 estimates the effect of propeller fouling on thrust based on the thrust force of the propeller shaft measured by a sensor installed on the ship 10A, and uses this as observation data D f|Ship_obs It may be removed from. Alternatively, the correction means 417 may compare the predicted friction resistance value of the entire hull, calculated using a water flow vector that reflects the influence of the surrounding water behavior, with the observed friction resistance value of the entire hull that includes the influence of the surrounding water behavior, or it may compare the predicted friction resistance value of the entire hull, calculated using a water flow vector in still water without the influence of the surrounding water behavior, with the observed friction resistance value of the entire hull from which the influence of the surrounding water behavior has been removed. In the former case, the marine weather conditions encountered by the ship 10 are input in the CFD calculation and a water flow vector is calculated. The friction resistance value of each mesh is calculated using this water flow vector, and the friction resistance value of the entire hull obtained by integrating them is used as predicted data and compared with the actual observed data. In the latter case, the increase in friction resistance due to fouling can be quantified by converting the observed data to match the calm water performance using a known method. The friction resistance value of each mesh is calculated using a water flow vector in calm water, and the friction resistance value of the entire hull obtained by integrating them is used as predicted data and compared with the observed data converted to match the calm water performance. 【0073】 The correction means 417 calculates the difference between the predicted data and the observed data using the following formula (4). 【0074】 Here, ΔD is the difference, D f|Ship_obs D is observational data. f|Ship This is the predicted data. Observed data D f|Ship_obs This is calculated using the above formula (3). Predicted data D f|Ship This is read from the ship database in storage 43. 【0075】 The correction means 417 corrects the roughness of each mesh of the ship 10A using the following formula (5). 【0076】 Here, k f|foulis the mesh roughness, K is the correction factor, and ΔD is the difference. Roughness k f|foul The correction factor K is expressed in vector form, which lists the types of contamination. The correction factor K may be calculated considering the area ratio of contamination and data accuracy. Mesh roughness k f|foul In all vessels 10, the difference ΔD is corrected by the correction factor K so that it approaches 0. In other words, the constraint is set so that when the sum of the differences ΔD is used as the evaluation function, this sum becomes as small as possible. For example, if the difference ΔD is less than 0, it means that the frictional resistance value of the predicted data is greater than the actual value. Therefore, in this case, the roughness is corrected by the correction factor K so that it becomes smaller. On the other hand, if the difference ΔD is greater than 0, it means that the frictional resistance value of the predicted data is smaller than the actual value. Therefore, in this case, the roughness is corrected by the correction factor K so that it becomes larger. Note that the roughness itself may be corrected by the correction factor K, or the relationship information showing the relationship between fouling information and roughness may be corrected. 【0077】 The roughness corrected in this way is applied not only to the target vessel 10A but to all vessels 10. Therefore, as shown in Figure 6, the roughness corrected based on the predicted and observed data of vessel 10A is used not only for calculating the frictional resistance value of vessel 10A but also for calculating the frictional resistance value of other vessels, such as vessel 10B. Mesh roughness k f|foul The roughness is corrected by a correction factor K so that the difference ΔD is minimized. For example, if the difference ΔD is less than 0, it means that the predicted frictional resistance value is greater than the actual value. Therefore, in this case, the roughness is corrected by the correction factor K so that it becomes smaller. On the other hand, if the difference ΔD is greater than 0, it means that the predicted frictional resistance value is smaller than the actual value. Therefore, in this case, the roughness is corrected by the correction factor K so that it becomes larger. Note that the roughness itself may be corrected by the correction factor K, or the relationship information showing the relationship between fouling information and roughness may be corrected. By calculating the frictional resistance value of the hull using the corrected roughness, the accuracy of the frictional resistance value is improved. 【0078】(3) Correction of water flow vector The correction means 417 corrects the water flow vector used to calculate the frictional resistance value of the mesh by data assimilation. As a result, the water flow vector is adjusted so that the frictional resistance value of the entire hull is closer to the observed value. Here, the target vessel 10 is assumed to be vessel 10A. 【0079】 The correction means 417 uses the frictional resistance value of the entire hull of the vessel 10A calculated in step S16 above as prediction data, and the frictional resistance value of the entire hull of the vessel 10A calculated by the above formula (3) as observation data. The correction means 417 calculates the difference between the prediction data and the observation data using the above formula (4). 【0080】 The correction means 417 corrects the water flow vector of each mesh of the ship 10A using the following formula (6). 【0081】 Here, V mesh V is the mesh water flow vector, K is the correction factor, and ΔD is the difference. Water flow vector V mesh The difference ΔD is corrected by the correction factor K so as to minimize it. For example, if the difference ΔD is less than 0, it means that the frictional resistance value of the predicted data is greater than the actual value. Therefore, in this case, the water flow vector is corrected by the correction factor K so as to be smaller. On the other hand, if the difference ΔD is greater than 0, it means that the frictional resistance value of the predicted data is smaller than the actual value. Therefore, in this case, the water flow vector is corrected by the correction factor K so as to be larger. When the water flow vector is estimated by CFD calculation, the water flow vector is an example of the CFD parameters according to the present invention. 【0082】 The water flow vectors modified in this way are applied only to the target vessel 10A and vessel 10 of the same type as the target vessel 10A. Therefore, as shown in Figure 6, the water flow vectors modified based on the predicted and observed data of vessel 10A are used to calculate the frictional resistance value of the target vessel 10A, but not to calculate the frictional resistance value of vessel 10B, which is of a different type from the target vessel 10A. By calculating the frictional resistance value of the hull using the modified water flow vectors, the accuracy of the frictional resistance value is improved. 【0083】 According to the first embodiment described above, the detailed fouling state of the entire hull can be objectively and quantitatively evaluated, and a highly accurate frictional resistance value of the hull can be calculated based on this fouling state. By evaluating the fouling effect on the hull based on such a frictional resistance value, the accuracy of the fouling effect evaluation is improved. 【0084】 II. Second Embodiment In the second embodiment, the frictional resistance value of the entire hull is determined from the hull image and three-dimensional model using the machine learning model 432. 【0085】 Figure 7 shows an example of the configuration of the server device 40A according to the second embodiment. The configuration of the evaluation system 1 according to the second embodiment is basically the same as the configuration of the evaluation system 1 according to the first embodiment, but the configuration of the server device 40A is slightly different. Components that are the same as in the first embodiment are denoted by the same reference numerals and their descriptions are omitted. 【0086】 The server device 40A comprises a processor 41, memory 42, storage 43, communication IF 44, input unit 45, and display unit 46. The processor 41 functions as a generation means 411A, an acquisition means 412, an image processing means 413, an identification means 414A, a calculation means 415, and an output means 416 by executing a program stored in memory 42. These functions are software modules realized by software cooperating with hardware resources. Note that the server device 40A does not necessarily have a modification means 417. 【0087】The generation means 411A generates a machine learning model 432 for identifying the frictional resistance value of each mesh of a ship's hull from images and a three-dimensional model of the hull. The machine learning model 432 is generated by performing machine learning using training data. The training data includes images of any ship's hull that has been soiled by immersion in water, and a three-dimensional model of that ship's hull. The images and three-dimensional model of the ship's hull are pre-labeled with the frictional resistance value of each mesh of the hull. The generation means 411A generates the machine learning model 432 by performing machine learning using the images and three-dimensional model of the ship's hull as explanatory variables and the frictional resistance value of each mesh of the hull as the target variable, using these as training data. Deep learning is included in the machine learning. The generation means 411A stores the generated machine learning model 432 in the storage 43. The generation means 411A may also update the machine learning model 432 by retraining or the like. The machine learning model 432 is an example of a trained model according to the present invention. 【0088】 The explanatory variables of the machine learning model 432 include a three-dimensional model of the hull. This is because the three-dimensional model of the hull is indirectly correlated with the frictional resistance value of each mesh in the hull, which is the objective variable. The three-dimensional model of the hull shows the three-dimensional shape of the surface of the submerged portion of the hull and the relative position of each mesh in the three-dimensional shape of the hull. The three-dimensional shape of the surface of the submerged portion of the hull and the relative position of each mesh in the three-dimensional shape affect the water flow vector of each mesh. Since the water flow vector of each mesh is correlated with the frictional resistance value of each mesh in the hull, the three-dimensional model of the hull is indirectly correlated with the frictional resistance value of each mesh in the hull. The three-dimensional model of the hull is an example of water flow information according to the present invention, as it is information for estimating the water flow vectors generated in the mesh while the ship 10 is sailing. 【0089】Furthermore, the explanatory variables of the machine learning model 432 include images of each mesh of the hull. This is because the images of each mesh of the hull are indirectly correlated with the objective variable, which is the frictional resistance value of each mesh of the hull. The images of each mesh of the hull visually represent the fouling state of each mesh. The fouling state of each mesh corresponds to the roughness of each mesh. Since the roughness of each mesh is correlated with the frictional resistance value of each mesh of the hull, the images of each mesh of the hull are indirectly correlated with the frictional resistance value of each mesh of the hull. 【0090】 Furthermore, instead of a three-dimensional model, flow field information or water flow vectors for each mesh of the hull may be used as explanatory variables in the machine learning model 432. This flow field information is an example of water flow information according to the present invention, as it is information for estimating the water flow vectors for each mesh of the hull. This flow field information and the water flow vectors for each mesh of the hull can be obtained, for example, by CFD calculation. 【0091】 Furthermore, the water flow vectors of each mesh of the hull are affected by the wave and current conditions around the ship 10. Therefore, the explanatory variables of the machine learning model 432 may also include wave information and current information. Wave information includes, for example, wave direction, wave height, and wave period. Current information includes, for example, tidal direction and tidal speed. Wave information and current information are information about the behavior of water around the ship 10, and are therefore examples of water behavior information according to the present invention. 【0092】Figure 8 is a flowchart illustrating a process for quantitatively evaluating the fouling effect according to the second embodiment. The processes in steps S21 to S23 are the same as those in steps S11 to S13 described in the first embodiment, so their explanation is omitted. In step S24, the identification means 414A uses the machine learning model 432 stored in the storage 43 to identify the frictional resistance value of each mesh of the hull from the three-dimensional model of the hull generated in step S22 and the mesh image of the hull generated in step S23. The frictional resistance value of each mesh is the contribution of the mesh to the value that correlates with the required horsepower of the ship 10, and is therefore an example of the contribution of a certain area of ​​the required horsepower index according to the present invention. Specifically, the identification means 414 inputs the three-dimensional model of the hull and the mesh image of each mesh of the hull as explanatory variables into the machine learning model 432, and obtains the frictional resistance value of the mesh output from the machine learning model 432 as the objective variable. Furthermore, if the explanatory variables of the machine learning model 432 include wave information and tidal current information, the identification means 414 further inputs the wave information and tidal current information as explanatory variables into the machine learning model 432. The processing in steps S25 to S26 is the same as the processing in steps S16 and S17 described in the first embodiment, so the explanation is omitted. 【0093】 As in the first embodiment described above, when calculating the frictional resistance value of the hull using a formula that utilizes roughness and water flow vectors, if the accuracy of the roughness and water flow vectors is low, the accuracy of the frictional resistance value of the hull will also be low. However, in the second embodiment, the frictional resistance value of the hull can be determined without using a formula that utilizes roughness and water flow vectors, thus increasing the accuracy of the frictional resistance value of the hull. By evaluating the fouling effect of the hull based on such a frictional resistance value, the accuracy of the fouling effect evaluation is improved. 【0094】 III. Modifications The present invention is not limited to the embodiments described above, and may be modified as follows. The following modifications may be used individually or in combination. 【0095】1. Modification 1 In the first and second embodiments described above, the hull of the ship may be photographed in a manner that does not use an underwater drone. In one example, a rail may be installed on the quay, and a drive device may move the camera along the rail, and the camera may photograph the hull of the ship as it moves along the rail. This camera is an example of an underwater moving camera according to the present invention. In another example, a camera may be attached to the tip of a rod-shaped member, and a photographer may move the rod-shaped member underwater, and the camera may photograph the hull of the ship as it moves underwater. This camera is an example of an underwater moving camera according to the present invention. In yet another example, a diver may dive with a camera and photograph the hull of the ship with the camera. This camera is an example of an underwater moving camera according to the present invention. Images of the hull of the ship can also be obtained with the configuration according to this modification. 【0096】 2. Modification 2 In the first and second embodiments described above, the method for generating a three-dimensional model of the hull from images of the hull is not limited to SfM and MVS. The image processing means 413 may generate a three-dimensional model of the hull from images of the hull by methods other than SfM and MVS. For example, the image processing means 413 generates a three-dimensional model by synthesizing images of the hull based on the shooting position and relative position of the hull images. Methods for determining the shooting position of the hull images include methods using acoustic positioning and methods using visual odometry. A three-dimensional model of the hull can also be generated with the configuration according to this modification. 【0097】 3. Modification 3 In the first and second embodiments described above, a three-dimensional model of the hull does not necessarily have to be generated. For example, the image processing means 413 may generate mesh images of each mesh of the hull from an image of the hull taken and the position from which it was taken. The frictional resistance value of the hull can also be determined with this modified configuration. 【0098】4. Modification 4 In the first embodiment described above, the water flow vector of each mesh of the hull is affected by the wave and current conditions around the ship 10. Therefore, the water flow vector of each mesh may be determined by CFD calculation or model testing based on wave information and current information in addition to the three-dimensional model of the hull. The three-dimensional model of the hull shows the three-dimensional shape of the surface of the submerged portion of the hull and the relative position of each mesh in the three-dimensional shape of the hull. Wave information includes, for example, wave direction, wave height, and wave period. Current information includes, for example, tidal direction and tidal speed. Wave information and current information are examples of water behavior information according to the present invention, as they are information about the behavior of water around the ship 10. According to this modification, the accuracy of the water flow vector used to calculate the frictional resistance value of the hull is improved. 【0099】 Furthermore, the water flow vectors of each mesh in the hull may be affected by mesh fouling. Therefore, the water flow vectors of each mesh may be determined by CFD calculation or model testing based on mesh fouling information in addition to the three-dimensional model of the hull. The three-dimensional model of the hull shows the three-dimensional shape of the surface of the submerged portion of the hull and the relative position of each mesh in the three-dimensional shape of the hull. The fouling information includes the location and shape of the fouling. The location and shape of the fouling of a mesh may be determined, for example, by image recognition from the mesh image of that mesh on the hull or the mesh images of surrounding meshes. According to this modification, the accuracy of the water flow vectors used to calculate the frictional resistance value of the hull is improved. 【0100】 5. Modification 5 In the first embodiment described above, the method for identifying the water flow vector of each mesh is not limited to a method based on CFD calculation or model testing. In one example, a flow meter is provided on an underwater drone, and the water flow vector of each mesh on the hull surface is measured by this flow meter. The flow meter is an example of a measuring device according to the present invention. According to this modification, the accuracy of the water flow vector used to calculate the frictional resistance value of the hull is improved. 【0101】In short, another aspect of the present invention may provide a method comprising the process of obtaining multiple images by photographing different regions of a ship in different directions using a camera moving underwater, and the process of measuring the vector of the water flow generated on the surface of the submerged portion of the ship using a measuring device that moves underwater along with the camera. 【0102】 In another example, water movement may be visualized using tracers such as ink, bubbles, or wind socks, and images of the hull may be taken along with the tracers. By analyzing these images, the water flow vectors for each mesh on the hull surface may be identified. According to this modification, the water flow vectors used to calculate the frictional resistance value of the hull can be accurately identified without equipping the underwater drone with dedicated devices such as measuring instruments. 【0103】 In short, another aspect of the present invention may provide a method comprising the process of obtaining multiple images by photographing different regions of a ship in different directions using a camera moving underwater, and the process of identifying the vector of the water flow occurring on the surface of the submerged portion of the ship from the multiple images. 【0104】 6. Modification 6 In the first and second embodiments described above, the value used to evaluate the effect of hull fouling is not limited to the frictional resistance value of the hull. For example, instead of or in addition to the frictional resistance value of the hull, the required horsepower of the vessel 10 or other values ​​correlated with this required horsepower may be used. The required horsepower of the vessel 10 increases when the hull becomes fouled. Therefore, the effect of hull fouling can also be evaluated using the required horsepower of the vessel 10. The required horsepower of the vessel 10 is an example of the required horsepower index according to the present invention. 【0105】Other values ​​that correlate with the required horsepower of the vessel 10 include the fuel consumption of the vessel 10. Fuel consumption increases as the hull becomes fouled. When the fuel consumption of the vessel 10 is used, in the first embodiment, the fuel consumption of the vessel 10 may be calculated using the frictional resistance value of the entire hull. In the second embodiment, a machine learning model may be generated by machine learning with the fuel consumption of the vessel 10 as the target variable, and the fuel consumption of the vessel 10 may be identified using this machine learning model. Since the fuel consumption of the vessel 10 correlates with the required horsepower of the vessel 10, it is an example of the required horsepower index according to the present invention. The effect of hull fouling can also be evaluated with this modified configuration. 【0106】 7. Modification 7 In the first embodiment described above, only the frictional resistance value of the hull at the time the image of the hull was taken can be obtained. However, it is not always possible to take an image of the hull at a desired time. Therefore, the calculation means 415 may estimate the roughness of the hull at a time when no image of the hull has been taken, and use the estimated roughness to calculate the frictional resistance value of the hull. 【0107】 Here, a process is performed to quantitatively evaluate the fouling effect described above based on images of the hull taken in January and July, and the roughness in January and July is identified. January and July are examples of multiple different periods according to the present invention. The calculation means 415 identifies the roughness in March from the roughness in January and July by time series analysis. March is an example of a period different from the period when the images were taken according to the present invention. Then, the calculation means 415 identifies the frictional resistance value of each mesh in March based on the roughness in March identified by time series analysis and the water flow vector of each mesh, using the same method as in the first embodiment described above. According to this modification, the roughness and frictional resistance value of the hull at a period different from the period when the images of the hull were taken can be obtained. 【0108】In short, the method according to the present invention may include a process of identifying the roughness at a time different from the time when the partial area of ​​the vessel under evaluation was photographed, based on the roughness identified with respect to a partial area of ​​the vessel under evaluation with respect to each of several different time periods, and a process of identifying the contribution of the partial area to the required horsepower index of the vessel under evaluation at a time different from the time when the partial area of ​​the vessel under evaluation was photographed, based on the roughness identified with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed and water flow information related to the partial area. 【0109】 8. Modification 8 In the first embodiment described above, only the frictional resistance value of the hull at the time the image of the hull was taken can be obtained. However, it is not always possible to take an image of the hull at a desired time. Therefore, the calculation means 415 may estimate the fouling information of the hull at a time when no image of the hull has been taken, and use the estimated fouling information to calculate the frictional resistance value of the hull. 【0110】 Here, a process is performed to quantitatively evaluate the aforementioned fouling effects based on images of the hull taken in January and July, and fouling information for January and July is identified. January and July are examples of multiple different periods according to the present invention. The calculation means 415 identifies fouling information for March from the fouling information for January and July by time-series analysis. March is an example of a period different from the time when the images were taken according to the present invention. Then, the calculation means 415 identifies the roughness of each mesh based on the fouling information for March identified by time-series analysis in the same manner as in the first embodiment described above, and identifies the frictional resistance value of each mesh in March based on the identified roughness and the water flow vector of each mesh. According to this modification, fouling information and frictional resistance values ​​of the hull at a time different from the time when the images of the hull were taken can be obtained. 【0111】In short, the method according to the present invention may include: a process of identifying fouling information at a time different from the time when the partial area of ​​the vessel under evaluation was photographed, based on fouling information identified with respect to a partial area of ​​the vessel under evaluation for each of several different time periods; a process of identifying the roughness of the partial area based on fouling information identified with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed; and a process of identifying the contribution of the partial area to the required horsepower index of the vessel under evaluation at a time different from the time when the partial area was photographed, based on the roughness identified with respect to a time different from the time when the partial area of ​​the vessel under evaluation was photographed and water flow information related to the partial area. 【0112】 9. Modification 9 In the first embodiment described above, roughness was determined using an analytical method, but roughness may be determined by methods other than analytical methods. Other methods include methods using machine learning models. 【0113】 For example, the roughness of each mesh may be identified from a mesh image of the hull using a machine learning model. The generation means 411 generates a machine learning model by machine learning using images of any object soiled by immersion in water as explanatory variables and the roughness of the hull mesh as the target variable, with these as training data. The arbitrary object may be the hull or a sheet of steel. The roughness of the hull mesh is labeled on this image. The generation means 411 may also update this machine learning model by retraining or the like. The machine learning model is an example of a trained model that has been trained according to the present invention. The identification means 414 inputs the mesh image of each mesh of the hull into this machine learning model to obtain the mesh roughness output from the machine learning model. The calculation means 415 may calculate the frictional resistance value of the mesh using the mesh roughness obtained in this way. According to this modification, the accuracy of the roughness used to calculate the frictional resistance value of the hull is improved. 【0114】In short, the method according to the present invention may include a process to determine roughness, in which an image of a portion of the area to be evaluated of the ship being evaluated is input to a trained model that has been trained using training data in which an image of an arbitrary object that has been soiled by immersion in water is included as an explanatory variable and the roughness of the object is included as an objective variable, and the roughness output by the trained model is obtained. 【0115】 10. Modification 10 In the first embodiment described above, the water flow vector was identified using an analytical method, but the water flow vector may be identified by a method other than an analytical method. Other methods include methods using a machine learning model. 【0116】 For example, a machine learning model may be used to identify the water flow vector for each mesh from a three-dimensional model of the hull. The generation means 411 generates a machine learning model by using the three-dimensional model of the hull as an explanatory variable and the water flow vector for each mesh of the hull as an objective variable, and performing machine learning on these as training data. The three-dimensional model of the hull shows the three-dimensional shape of the surface of the submerged part of the hull and the relative position of each mesh in the three-dimensional shape of the hull. The water flow vector for each mesh of the hull is labeled on the three-dimensional model of the hull. The generation means 411 may also update this machine learning model by retraining or the like. The machine learning model is an example of a trained model that has been trained according to the present invention. The identification means 414 inputs the three-dimensional model of the hull into this machine learning model to obtain the water flow vector for each mesh output from the machine learning model. The calculation means 415 may calculate the frictional resistance value of the mesh using the water flow vector for each mesh identified in this way. According to this modification, the accuracy of the water flow vector used to calculate the frictional resistance value of the hull is improved. 【0117】In short, the method according to the present invention may include a process to obtain the water flow vector output by a trained model that has been trained using training data in which the three-dimensional shape of the entire surface of the submerged portion of any ship and the position of any part of the ship in the three-dimensional shape are included as explanatory variables, and the water flow vector generated in the part of the ship while it is sailing is included as the objective variable, by inputting the three-dimensional shape of the entire surface of the submerged portion of the ship to be evaluated and the position of the part of the ship to be evaluated in the three-dimensional shape. 【0118】 Furthermore, the water flow vectors for each mesh of the hull are affected by the wave and current conditions around the ship 10. Therefore, the explanatory variables of the machine learning model for identifying the water flow vectors for each mesh may also include wave information and current information. Wave information may include, for example, wave direction, wave height, and wave period. Current information may include, for example, tidal direction and tidal speed. Wave information and current information are information about the behavior of water around the ship 10 and are therefore examples of water behavior information according to the present invention. According to this modification, the accuracy of the water flow vectors used to calculate the frictional resistance value of the hull is improved. 【0119】 In short, in the method according to the present invention, the water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel to be evaluated, the position of a part of the said three-dimensional shape, and water behavior information which is information about the behavior of the water around the vessel to be evaluated. The process of identifying the contribution may include the process of obtaining the required horsepower index output by a trained model that has been machine-learned using training data in which an image taken of an arbitrary part of the surface of an arbitrary vessel, the three-dimensional shape of the entire surface of the submerged portion of the vessel, the position of the said part of the said three-dimensional shape, and water behavior information relating to the behavior of the water around the vessel are included as explanatory variables, and the required horsepower index for the said part of the vessel is included as the objective variable. The trained model is then input with an image taken of a part of the vessel to be evaluated, the three-dimensional shape of the entire surface of the submerged portion of the vessel, the position of the said part of the said three-dimensional shape, and water behavior information relating to the behavior of the water around the vessel. 【0120】Furthermore, the water flow vectors of each mesh on the hull may be affected by mesh fouling. Therefore, the explanatory variables of the machine learning model used to identify the water flow vector of each mesh may also include information on mesh fouling. This fouling information includes the location and shape of the fouling. The location and shape of the mesh fouling may be identified, for example, by image recognition from a mesh image of that mesh on the hull or a mesh image of the surrounding meshes. This modification improves the accuracy of the water flow vectors used to calculate the frictional resistance of the hull. 【0121】 In short, the method according to the present invention may include a process to obtain a water flow vector output by a trained model that has been trained using training data in which the three-dimensional shape of the surface of the entire submerged portion of any ship, the position of any part of the ship in the three-dimensional shape, and the position and shape of fouling identified from images of the part or the area surrounding the part are included as explanatory variables, and the water flow vector generated in the part while the ship is sailing is included as the objective variable. The trained model is then input with the three-dimensional shape of the surface of the entire submerged portion of the ship to be evaluated, the position of the part of the ship to be evaluated in the three-dimensional shape, and the position and shape of fouling identified from images of the part or the area surrounding the part, and the trained model is output. 【0122】 11. Modification 11 Another aspect of the present invention may provide a method comprising a process for generating or updating the machine learning model 431 described in the first embodiment, the machine learning model 432 described in the second embodiment, and the various machine learning models described in the modification. 【0123】 12. Modification 12 Another embodiment of the present invention may provide a method for generating a three-dimensional model of a ship's hull. The three-dimensional model of the ship's hull may have a surface texture, or it may be three-dimensional shape data representing only the three-dimensional shape of the ship's hull without a surface texture. This three-dimensional model of the ship's hull may be used for purposes other than determining the frictional resistance value of the ship's hull. According to this modification, the three-dimensional shape of the ship's hull can be easily grasped. 【0124】In short, yet another aspect of the present invention may provide a method comprising the process of obtaining multiple images by photographing different areas of a ship in different directions using a camera moving underwater, and the process of generating three-dimensional shape data representing the three-dimensional shape of the surface of the submerged portion of the ship from the multiple images. 【0125】 13. Modification 13 Another aspect of the present invention may be provided as a method for generating mapping image data representing images corresponding to the mesh of a ship's hull. The mapping image data is data relating each mesh of a three-dimensional model of the ship's hull to a mesh image of that mesh. The mesh image is an image corresponding to any direction and any region of the ship's surface. This mapping image data may be used for purposes other than determining the frictional resistance value of the ship's hull. According to this modification, the condition of the ship's hull can be determined. 【0126】 In short, yet another aspect of the present invention may provide a method comprising the process of obtaining multiple images by photographing different areas of a ship in different directions using a camera moving underwater, and the process of generating mapping image data from the multiple images that represents an image corresponding to an arbitrary area of ​​the surface of the submerged part of the ship. 【0127】 14. Modification 14 Another aspect of the present invention may provide a method for determining the location of a photograph taken of a ship's hull. The location of the photograph may be determined by SfM as described in the first embodiment. The location of the photograph may be used for purposes other than determining the frictional resistance value of the ship's hull. According to this modification, the location of the photograph taken of the ship's hull image can be easily determined. 【0128】 In short, yet another aspect of the present invention may provide a method comprising the process of obtaining a plurality of images by photographing different areas of a ship in different directions with a camera moving underwater, and the process of identifying the shooting position of each of the plurality of images. 【0129】15. Modification 15 In the embodiments described above, the configuration, functions, and operation of the evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40 are illustrative and not limited thereto. The evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40 may be configured to include one or more of the above-described devices or parts, or they may be configured to omit some of the devices or parts. Furthermore, at least some of the functions of the evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40 may be provided by other devices, the functions of one device may be distributed among multiple devices, or the functions of multiple devices may be combined into one device. Moreover, the operating procedures of the evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40 may be rearranged or some operating procedures may be omitted, as long as they are not contradictory. 【0130】 16. Modification 16 Another embodiment of the present invention may provide a method having steps of processing performed in the evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40. Yet another embodiment of the present invention may provide a program to be executed in the evaluation system 1, the vessel 10, the underwater drone 20, the control device 30, and the server device 40. This program may be provided stored on a computer-readable recording medium or provided by download via the Internet or the like. 【0131】 17. Modification 17 In the second embodiment described above, when a fouling index is input to the machine learning model 432, the server device 40A may have a modification means for modifying the fouling index by data assimilation. The fouling index is a value that quantitatively shows the effect of fouling, and is determined, for example, based on a graph where the horizontal axis is the type and degree of fouling and the vertical axis is the fouling index. The modification means uses the frictional resistance value of the entire hull of the ship 10A calculated in step S25 described above as prediction data, and modifies the fouling index so that the difference between the prediction data and the observation data is minimized. According to this modification, the accuracy of the frictional resistance value is improved. 【0132】18. Modification 18 In the first embodiment described above, the roughness is determined from the contamination information and the friction resistance value is calculated from the roughness, but a friction resistance coefficient may be used instead of or in addition to the roughness. Similarly, in the modification described above, a friction resistance coefficient may be used instead of or in addition to the roughness. A friction resistance coefficient is a coefficient that relates the frictional resistance force acting from a fluid on an object to the fluid velocity and the surface area of ​​the object. The friction resistance coefficient is determined by factors such as the fluid velocity and the roughness of the object surface. Furthermore, not only for simple shapes such as flat plates, but also for complex shapes, it is possible to define a shape coefficient (corresponding to the friction coefficient) by organizing the relationship between friction resistance, surface area, and fluid velocity when subjected to fluid velocity from one direction. When a friction resistance coefficient is used, the processing in steps S15 to S17 may be changed as follows, for example. 【0133】 In step S15, the calculation means 415 calculates the frictional resistance value of each mesh based on the fouling information of each mesh identified in step S14. Here, an example of calculating the frictional resistance value of mesh Z will be given. As shown in Figure 4, the fouling information of mesh Z includes a set of fouling type (rust), degree of fouling (severity), and area percentage of this fouling (30%), and a set of fouling type (algae), degree of fouling (mild), and area percentage of this fouling (10%). The calculation means 415 calculates the frictional resistance value of mesh Z of the hull using the following formula (7). 【0134】 Here, D f|Mesh Z ρ is the frictional resistance value of mesh Z [N], and ρ is the fluid density [kg / m³]. 3 ], V is the fluid velocity in mesh Z [m / s], and S is the surface area of ​​mesh Z [m 2 ], C f ρ is the frictional resistance coefficient. Density ρ is a constant. Velocity V represents the velocity of the water flow in mesh Z. Velocity V is estimated, for example, based on CFD calculations based on a three-dimensional model or based on model testing. Frictional resistance coefficient C fIt is determined based on the type and degree of contamination. The correspondence between the type and degree of contamination and the frictional resistance coefficient is predetermined by literature and experiments. Frictional resistance coefficient C f|rust,hev This is the frictional resistance coefficient corresponding to severe rust in this relationship. Frictional resistance coefficient C f|alga,light This is the frictional resistance coefficient corresponding to mild algae in this correspondence. Frictional resistance coefficient C f|flat This is the frictional resistance coefficient corresponding to the state without contamination in this correspondence. Furthermore, in the above formula (7), in order to weight according to the area ratio of contamination, the frictional resistance value of each contamination state in mesh Z is multiplied by the area ratio of that contamination. Since the area ratio of contamination, which is severe rust, is 30%, in the above formula (7), the frictional resistance coefficient C f|rust,hev Terms containing this are multiplied by 0.3. Since the area percentage of light algal contamination is 10%, in the above formula (7), the frictional resistance coefficient C f|alga,light Terms containing this are multiplied by 0.1. Since the area ratio of the unstained state is the remaining 60%, in the above formula (7), the frictional resistance coefficient C f|flat Terms containing this are multiplied by 0.6. Then, by multiplying the frictional resistance values ​​of the heavily rusted parts of mesh Z, the lightly algae-covered parts, and the uncontaminated parts, the frictional resistance value D of mesh Z is obtained. f|Mesh Z The result is obtained. The calculation means 415 calculates the frictional resistance value for other meshes in the same manner. 【0135】 In step S16, the calculation means 415 calculates the total frictional resistance value of the hull by integrating the frictional resistance values ​​of all the meshes calculated in step S15. The calculation means 415 calculates the total frictional resistance value of the hull using the following formula (8). 【0136】 Here, D f|Ship D is the frictional resistance value of the entire hull. f|Mesh This is the frictional resistance value of each mesh in the hull. The total frictional resistance value of all meshes in the hull is D. f|Mesh By accumulating these, the total frictional resistance value D of the hull can be calculated. f|Ship You can obtain this. 【0137】In step S17, the output means 416 outputs the frictional resistance values ​​of each mesh of the hull calculated in step S15 and the frictional resistance value of the entire hull calculated in step S16. For example, the output means 416 stores the frictional resistance values ​​of each mesh of the hull and the frictional resistance value of the entire hull in the ship database stored in the storage 43. 【0138】 1: Evaluation system, 10, 10A, 10B: Ship, 20: Underwater drone, 22: Camera, 30: Control device, 40, 40A: Server device, 41: Processor, 42: Memory, 43: Storage, 44: Communication IF, 45: Input unit, 46: Display unit, 411, 411A: Generation means, 412: Acquisition means, 413: Image processing means, 414, 414A: Identification means, 415: Calculation means, 416: Output means, 417: Correction means, 431, 432: Machine learning model

Claims

A process to identify the contribution of a specific area to the required horsepower of a vessel or a required horsepower index, which is a value correlated with the required horsepower, based on an image taken of a specific area of ​​the vessel's surface and water flow vectors or water flow information used to estimate those vectors, generated in that area while the vessel is sailing. A method for providing this. 　 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel being evaluated, and the position of the said three-dimensional shape of the said portion of the region. The process of identifying the contribution includes a process of obtaining the required horsepower index output by a trained model, which has been trained using training data that includes an image of a portion of the surface of any ship, the three-dimensional shape of the entire submerged portion of the ship, and the position of the portion in the three-dimensional shape as explanatory variables, and the required horsepower index for the portion as the dependent variable, by inputting an image of a portion of the ship to be evaluated, the three-dimensional shape of the entire submerged portion of the ship, and the position of the portion in the three-dimensional shape as the dependent variable. The method according to claim 1. 　 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel under evaluation, the position of the portion of the vessel in the three-dimensional shape, and water behavior information which is information regarding the behavior of the water around the vessel under evaluation. The process of identifying the contribution includes a process of obtaining the required horsepower index output by a trained model that has been machine-learned using training data in which the explanatory variables include an image taken of an arbitrary portion of the surface of any ship, the three-dimensional shape of the entire submerged portion of the ship, the position of the portion in the three-dimensional shape, and water behavior information relating to the behavior of the water around the ship, and the required horsepower index relating to the portion in question as the dependent variable, and inputting an image taken of the portion of the ship to be evaluated, the three-dimensional shape of the entire submerged portion of the ship, the position of the portion in the three-dimensional shape, and water behavior information relating to the behavior of the water around the ship. The method according to claim 1. 　 The process for identifying the contribution includes: a process for identifying at least one of the roughness and frictional resistance coefficient of a portion of the vessel being evaluated based on an image of that portion of the vessel being evaluated; and a process for identifying the contribution based on at least one of the roughness and frictional resistance coefficient and water flow information relating to that portion of the vessel being evaluated. The method according to claim 1. 　 The process of identifying at least one of the roughness and frictional resistance coefficient includes a process of inputting an image of a portion of the vessel to be evaluated into a trained model that has been machine-trained using training data in which an image of an arbitrary object contaminated by immersion in water is included as an explanatory variable, and at least one of the roughness and frictional resistance coefficient of the object is included as an objective variable, in order to obtain at least one of the roughness and frictional resistance coefficient output by the trained model. The method according to claim 4. 　 A process to identify at least one of the roughness and frictional resistance coefficient at a time different from the time when the said area was photographed, based on at least one of the roughness and frictional resistance coefficient identified with respect to a part of the ship being evaluated for each of several different time periods, A process to identify the portion of the required horsepower index of the vessel under evaluation that is contributed to by that portion, based on at least one of the roughness and frictional resistance coefficient identified for a time different from when that portion was photographed, and water flow information relating to that portion; The method according to claim 4, comprising: 　 The process of identifying at least one of the roughness and the frictional resistance coefficient includes: a process of identifying contamination information, which is information relating to one or more types of contamination in a part of the ship being evaluated and the area occupied by each of those one or more types of contamination in that part, based on an image taken of that part of the ship being evaluated; and a process of identifying at least one of the roughness and the frictional resistance coefficient for that part based on the contamination information. The method according to claim 4. 　 The process for identifying the aforementioned contamination information includes a process for obtaining contamination information output by a trained model that has been trained using training data in which images of an arbitrary object contaminated by immersion in water are included as explanatory variables and contamination information of said object is included as the objective variable, by inputting images of a portion of the area to be evaluated of the ship to be evaluated. The method according to claim 7. 　 A process to identify fouling information at a time different from the time when the A process to identify at least one of the roughness and frictional resistance coefficient for a portion of the vessel subject to evaluation, based on fouling information identified at a time different from the time when the portion of the vessel subject to evaluation was photographed. A process to identify the portion of the required horsepower index of the vessel under evaluation that is contributed to by that portion, based on at least one of the roughness and frictional resistance coefficient identified for a time different from when that portion was photographed, and water flow information relating to that portion; The method according to claim 7, comprising: 　 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel being evaluated, and the position of the said three-dimensional shape of the said portion of the region. The process for identifying the contribution includes: a process for identifying the vector of the water flow generated in a certain area of ​​the vessel while it is sailing, based on the three-dimensional shape of the entire surface of the submerged portion of the vessel being evaluated and the position of that area of ​​the vessel being evaluated in that three-dimensional shape; and a process for identifying the contribution based on at least one of the roughness and frictional resistance coefficient identified based on an image of that area and the vector of the water flow. The method according to claim 4. 　 The process for identifying the water flow vector includes a process to obtain the water flow vector output by a trained model that has been machine-learned using training data in which the three-dimensional shape of the entire surface of the submerged portion of any ship and the position of any part of the ship in that three-dimensional shape are included as explanatory variables, and the water flow vector generated in that part of the ship while it is sailing is included as the target variable, by inputting the three-dimensional shape of the entire surface of the submerged portion of the ship being evaluated and the position of the part of the ship being evaluated in that three-dimensional shape. The method according to claim 10. 　 The process of identifying the water flow vector includes a process of identifying the water flow vector that occurs in a part of the area of ​​the vessel under evaluation while the vessel is under navigation, using CFD with respect to the three-dimensional shape of the vessel under evaluation. The method according to claim 10. 　 The water flow information includes the three-dimensional shape of the entire surface of the submerged portion of the vessel being evaluated, the position of the portion of the vessel in the three-dimensional shape, and the location and shape of the contamination identified from the image of the portion of the vessel or the area surrounding it. The process for identifying the contribution includes: a process for identifying the vector of water flow generated in a certain area while the vessel is underway, based on the three-dimensional shape of the entire surface of the submerged portion of the vessel under evaluation, the position of a certain area of ​​the vessel under evaluation in that three-dimensional shape, and the location and shape of fouling identified from images of that area or its surroundings; and a process for identifying the contribution based on at least one of the roughness and frictional resistance coefficient identified from images of that area and the vector of water flow. The method according to claim 4. 　 The process of identifying the water flow vector includes a process of obtaining the water flow vector output by a trained model that has been machine-learned using training data in which the three-dimensional shape of the entire surface of the submerged portion of any ship, the position of any part of the ship in the three-dimensional shape, and the position and shape of fouling identified from images of the part or its surroundings are included as explanatory variables, and the water flow vector generated in the part while the ship is sailing is included as the objective variable, and inputting the three-dimensional shape of the entire surface of the submerged portion of the ship to be evaluated, the position of the part of the ship to be evaluated in the three-dimensional shape, and the position and shape of fouling identified from images of the part or its surroundings. The method according to claim 13. 　 A process to identify the contribution of the entire submerged portion of the hull to the required horsepower index of the ship, by summing the contributions of each of the multiple partial regions of the ship to be evaluated obtained by dividing the surface of the ship to be evaluated. The method according to claim 1, comprising: 　 A process to identify the contribution of the entire submerged portion of the hull to the required horsepower index of the ship, by summing the contributions of the required horsepower index of each of the multiple partial regions of the ship to be evaluated obtained by dividing the surface of the ship to be evaluated, In the process of identifying at least one of roughness and frictional resistance coefficient for a given area, a process is performed to adjust relational information showing the relationship between the fouling information used to identify at least one of roughness and frictional resistance coefficient from fouling information and at least one of roughness and frictional resistance coefficient, in order to reduce the difference between the contribution of the entire submerged portion of the hull to the required horsepower index of the vessel measured during each of multiple voyages by the same vessel or multiple vessels being evaluated, and the contribution of each of the specified areas of the required horsepower index for each of multiple areas of the evaluation that constitute the entire submerged portion of the hull, in order to identify at least one of roughness and frictional resistance coefficient for the said area. The method according to claim 7, comprising: 　 A process to identify the contribution of the entire submerged portion of the hull to the required horsepower index of the ship, by summing the contributions of the required horsepower index of each of the multiple partial regions of the ship to be evaluated obtained by dividing the surface of the ship to be evaluated, A process to adjust the parameters of the CFD such that the difference between the contribution of the entire submerged portion of the hull to the required horsepower index of the vessel measured during each of multiple voyages by the same vessel under evaluation or multiple vessels of the same hull type, and the value obtained by summing the contribution of each of the specified required horsepower indexes for each of the multiple evaluation areas constituting the entire submerged portion of the hull of the vessel during said voyage, is reduced. The method according to claim 12, comprising: 　 This process involves generating or updating a trained model using training data that includes images of any portion of the surface of any vessel, the three-dimensional shape of the entire submerged portion of the vessel's surface, and the position of the portion in question within that three-dimensional shape as explanatory variables, and the vessel's required horsepower or the contribution of the portion in question to the required horsepower index, which is a value correlated with the required horsepower, as the dependent variable. A method for providing this. 　 This process involves generating or updating a trained model using training data that includes, as explanatory variables, an image of any portion of the surface of any vessel, the three-dimensional shape of the entire submerged portion of the vessel, the position of the portion in question within that three-dimensional shape, and water behavior information regarding the behavior of the water around the vessel, while the dependent variable is the vessel's required horsepower or the contribution of the portion in question to the required horsepower index, which is a value correlated with the required horsepower. A method for providing this. 　 This process involves generating or updating a trained model by performing machine learning using training data that includes images of any object soiled by immersion in water as explanatory variables, and at least one of the object's roughness and frictional resistance coefficient as the dependent variable. A method for providing this. 　 This process involves generating or updating a trained model using training data that includes images of any object soiled by immersion in water as explanatory variables, and soiling information, which is information about one or more types of soiling on the object and the area occupied by each of those types of soiling on the object, as the dependent variable. A method for providing this. 　 This process involves generating or updating a trained model by performing machine learning using training data that includes the three-dimensional shape of the entire submerged portion of any given vessel, the position of any part of the vessel within that three-dimensional shape as explanatory variables, and the vector of the water flow generated in that part of the vessel while it is sailing as the dependent variable. A method for providing this. 　 This process involves generating or updating a trained model by performing machine learning using training data that includes, as explanatory variables, the three-dimensional shape of the entire surface of the submerged portion of any vessel, the position of any portion of the vessel in that three-dimensional shape, and the location and shape of fouling identified from images of that portion or its surroundings, and as the dependent variable, the vector of the water flow generated in that portion while the vessel is sailing. A method for providing this. 　 A process that uses a camera moving underwater to photograph different areas of a ship from different directions to obtain multiple images, A process for generating mapping image data from the aforementioned multiple images, representing images corresponding to any direction and any region of the surface of the submerged part of the vessel. A method for providing this. 　 A process that uses a camera moving underwater to photograph different areas of a ship from different directions to obtain multiple images, A process for generating three-dimensional shape data representing the three-dimensional shape of the surface of the submerged portion of the ship from the aforementioned multiple images. A method for providing this. 　 A process that uses a camera moving underwater to photograph different areas of a ship from different directions to obtain multiple images, A process to identify the shooting location of each of the multiple images from the multiple images. A method for providing this. 　 A process that uses a camera moving underwater to photograph different areas of a ship from different directions to obtain multiple images, A process to measure the vector of the water flow generated on the surface of the submerged part of the vessel using a measuring device that moves underwater in conjunction with the camera. A method for providing this. 　 A process that uses a camera moving underwater to photograph different areas of a ship from different directions to obtain multiple images, A process to identify the vector of the water flow occurring on the surface of the submerged portion of the ship from the aforementioned multiple images. A method for providing this.

Images