Image analysis system, image analysis method, and image analysis program

Abstract

This image analysis system is provided with a satellite image acquisition unit, an environmental data acquisition unit, a detection unit, an estimation unit, and an output unit. The satellite image acquisition unit acquires a satellite image of an area to be analyzed. The environmental data acquisition unit acquires environmental data of the area to be analyzed. The detection unit detects an object to be analyzed in the satellite image. The estimation unit estimates the accuracy of detection, on the basis of the environmental data. The output unit outputs a result of the detection and information indicating the accuracy of the detection.

Description

Technical Field

[0001] The present disclosure relates to an image analysis system and the like.

Background Art

[0002] Satellite images captured by artificial satellites are widely used for analyzing objects on the earth's surface. For example, SAR (Synthetic Aperture Radar) can perform observation without being affected by the presence or absence of sunlight and clouds, so satellite images captured by SAR are suitable for analyzing objects on the earth's surface. For this reason, satellite images captured by SAR are sometimes used to identify objects existing in a region to be analyzed. On the other hand, since imaging is performed by observing the reflection of electromagnetic waves on the earth's surface, for example, a mapped image captured by SAR is an image represented by monochrome gradations. In an image represented by monochrome gradations, for example, the contrast ratio between the sea or land and the object to be analyzed is not sufficient, and it may be difficult to distinguish the object appearing in the satellite image. For this reason, for example, analysis of an object appearing in a satellite image may be further performed using data other than the satellite image.

[0003] The data analysis device of Patent Document 1 acquires a satellite image and observation data of the surface environment of the earth. The data analysis device of Patent Document 1 generates analysis data obtained by correcting the analysis data based on the satellite image based on the observation data of the surface environment of the earth.

Prior Art Literature

Patent Literature

[0004]

Patent Literature 1

Summary of Invention

Problem to be Solved by the Invention

[0005] With the technology described in Patent Document 1, it may be difficult to identify an object existing in a region to be analyzed.

[0006] This disclosure aims to provide an image analysis system, etc., that can easily identify objects present in the area being analyzed, in order to solve the above-mentioned problems. [Means for solving the problem]

[0007] To solve the above problems, the image analysis system disclosed herein comprises: an image acquisition means for acquiring satellite images of the area to be analyzed; an environmental data acquisition means for acquiring environmental data of the area to be analyzed; a detection means for detecting objects to be analyzed that are captured in the satellite images; an estimation means for estimating the accuracy of detection based on the environmental data; and an output means for outputting the detection result and information indicating the accuracy of detection.

[0008] The image analysis method disclosed herein acquires satellite images of the area to be analyzed, acquires environmental data of the area to be analyzed, detects the object to be analyzed in the satellite images, estimates the accuracy of detection based on the environmental data, and outputs the detection result and information indicating the accuracy of detection.

[0009] The recording medium of this disclosure non-temporarily records an image analysis program that causes a computer to execute the following processes: acquiring satellite images of the area to be analyzed; acquiring environmental data of the area to be analyzed; detecting objects to be analyzed that are captured in the satellite images; estimating the accuracy of detection based on the environmental data; and outputting the detection results and information indicating the accuracy of detection. [Effects of the Invention]

[0010] According to this disclosure, it is possible to easily identify objects present in the area being analyzed. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows an example of the configuration in an embodiment of the present disclosure. [Figure 2] This diagram schematically illustrates how the Earth's surface is imaged by artificial satellites. [Figure 3] This figure shows an example of satellite imagery in an embodiment of the present disclosure. [Figure 4] This figure shows an example of satellite imagery in an embodiment of the present disclosure. [Figure 5] This figure shows an example of satellite imagery in an embodiment of the present disclosure. [Figure 6] This figure shows an example of the configuration of an image analysis system according to the embodiment of this disclosure. [Figure 7] This figure shows an example of a screen displaying analysis results in an embodiment of the disclosure. [Figure 8] This figure shows an example of the operation flow of an image analysis system in an embodiment of the disclosure. [Figure 9] This figure shows an example of the hardware configuration of an image analysis system in an embodiment of the disclosure. [Modes for carrying out the invention]

[0012] Embodiments of this disclosure will be described in detail with reference to the figures. Figure 1 is a diagram showing an example of the configuration of an analysis system. The analysis system comprises, for example, an image analysis system 10, a terminal device 20, a satellite image management server 30, and an environmental data management server 40. The image analysis system 10 connects to the terminal device 20, for example, via a network. The image analysis system 10 connects to the satellite image management server 30, for example, via a network. The image analysis system 10 connects to the environmental data management server 40, for example, via a network. There may also be multiple terminal devices 20. The number of terminal devices 20 can be set as appropriate.

[0013] The image analysis system 10 is, for example, a system for analyzing satellite images. Satellite images are, for example, images of the Earth's surface taken by an imaging device mounted on an artificial satellite. Satellite images are, for example, taken by SAR (Synthetic Aperture Radar). Satellite images may also be taken by imaging devices other than SAR. Satellite images are used, for example, to detect objects that appear in the satellite image. The detection of objects that appear in the satellite image is performed, for example, by image recognition. However, for example, images taken by SAR are monochrome images based on electromagnetic waves reflected from the Earth's surface, so the accuracy of object detection by image recognition may not be sufficient. For example, when the object to be analyzed is a ship, image recognition may detect an object other than a ship as a ship. For this reason, when performing analysis using images taken by SAR, for example, after object detection by image recognition, the person in charge of analysis may identify the objects that appear in the image. The person in charge of analysis may, for example, visually check the image taken by SAR to identify the objects that appear in the image taken by SAR.

[0014] Figure 2 is a schematic diagram illustrating an example of how the Earth's surface is imaged by a satellite. In the example in Figure 2, a ship is sailing on the sea. In the example in Figure 2, the area in which the ship is sailing is imaged by an imaging device mounted on the satellite. The Earth's surface is imaged, for example, by SAR (Surface-to-Air Resonance). The satellite image captured by the imaging device mounted on the satellite is output to the satellite image management server 30, for example, via a ground station. In the example in Figure 2, environmental data is also observed by observation sensors installed on the sea. The observation sensors installed on the sea are, for example, observation buoys that observe water temperature and wave conditions. Wave conditions include, for example, wave height, wave direction, and wave period. The observation sensors are not limited to observation buoys. Also, the items of environmental data observed by the observation sensors are not limited to those mentioned above. The observed environmental data is output to the environmental data management server 40, for example.

[0015] The image analysis system 10 acquires satellite images captured by artificial satellites from, for example, the satellite image management server 30. The image analysis system 10 also acquires environmental data observed by observation sensors from, for example, the environmental data management server 40. The image analysis system 10 uses, for example, an image analysis model to detect objects of analysis in the acquired satellite images. Then, based on, for example, the acquired environmental data of the area of ​​analysis, the image analysis system 10 estimates the accuracy of detecting the objects of analysis in the detection results by the image recognition model. The accuracy of detecting the objects of analysis is, for example, an indicator of the likelihood of the object detection result. That is, the accuracy of detecting the objects of analysis is an indicator of the likelihood that the object detected by the image recognition model is the object of analysis. For example, if the object of analysis is a ship, and the image analysis system 10 detects that a ship is in the satellite image, the accuracy of detecting the objects of analysis is an indicator of the possibility that the object is a ship. The image recognition model is a learning model that detects objects in satellite images. The image recognition model will be explained later.

[0016] Furthermore, the environmental data is, for example, data relating to an environment that can affect whether an object is present. The environmental data is, for example, observation data obtained by observing an environment that can affect whether an object is present. The environmental data may be data relating to terrain. The environmental data may be water depth data. The water depth data may be data obtained by observing changes in water depth due to tides. For example, when the analysis target is a ship, navigation or anchoring of the ship can be affected by water depth, tidal currents, waves, wind speed, and terrain. A large ship cannot, for example, navigate through shallow-water areas. In addition, for ordinary small ships, it is dangerous to navigate in areas with fast tidal currents and high wave heights, so they navigate avoiding areas with fast tidal currents and high wave heights. In addition, areas near steep cliffs are not suitable for anchoring ships. As described above, the possibility that a ship exists at a target analysis point can vary depending on the situation indicated by the environmental data at the target analysis point. Therefore, when the analysis target is a ship, the probability that a ship exists can be estimated based on the environmental data. For this reason, for example, when the analysis target is a ship, the image analysis system 10 can estimate the accuracy of ship detection in a detection result based on environmental data of a region to be analyzed when the ship is detected by an image recognition model.

[0017] Furthermore, when the object of analysis is a vessel, it may be difficult to distinguish between a vessel and other objects. Other objects include, for example, marine organisms. The presence or absence of marine organisms can be influenced by, for example, the environment. The presence or absence of marine organisms can be influenced by, for example, water depth, water temperature, air temperature, currents, waves, wind speed, and topography. The environmental factors that can influence the presence of marine organisms are not limited to those listed above. For example, when the area under analysis is a sea area where marine organisms may exist, and a vessel is detected by image recognition, the image analysis system 10 estimates the accuracy of the detection by image recognition based on environmental data related to the presence of marine organisms. For example, if the environmental data suggests that the possibility of it being a marine organism is low, the image analysis system 10 can estimate that the accuracy of detecting a vessel in the detection results by the image recognition model is high. Conversely, if the environmental data suggests that the possibility of it being a marine organism is high, the image analysis system 10 estimates that the accuracy of detecting a vessel in the detection results by the image recognition model is low.

[0018] The image analysis system 10 outputs, for example, to a terminal device 20 a detection result of an object to be analyzed captured in a satellite image and a detection accuracy. The terminal device 20 outputs, for example, to a display device (not shown) the detection result of the object to be analyzed captured in the satellite image and the detection accuracy. A person in charge of analysis identifies the object by referring to, for example, the detection result of the object to be analyzed captured in the satellite image displayed on the display device and the detection accuracy. For example, when a ship that is an analysis target is detected, and an estimation result of detection accuracy based on environmental data indicates that there is a high possibility that the object is not a ship, the person in charge of analysis determines whether the object is a ship with reference to the estimation result of detection accuracy. When the object is detected as a ship, and the estimation result based on environmental data indicates that there is a high possibility that the object is not a ship, the person in charge of analysis determines whether the object detected as a ship is a ship by referring to, for example, a satellite image obtained when the object was identified as a non-ship object in past analysis. When the object is detected as a ship, and the estimation result based on environmental data indicates that there is a high possibility that the object is not a ship, the person in charge of analysis identifies the object by considering, for example, the possibility that a suspicious ship is present. This is because if a ship is present in an area where the possibility of a ship existing is low according to environmental data, the ship may be navigating or anchored with an unusual intention. The unusual intention refers to, for example, a purpose different from the normal purpose in the sea area where the ship is navigating or anchored. As described above, identifying the object captured in the satellite image by referring to the detection result of the object to be analyzed and the detection accuracy of the detection result can facilitate the work of identifying the object captured in the satellite image.

[0019] Figure 3 shows an example of a satellite image obtained by imaging the ground surface with SAR. The example of the satellite image in FIG. 3 is an example of a satellite image obtained by imaging the periphery of a bay. The example of the satellite image in FIG. 3 is an example of a satellite image in a case where no ships or other objects exist at all in the imaged area. In the example of the satellite image in FIG. 3, the low-brightness portion is a land area. Further, in the example of the satellite image in FIG. 3, the high-brightness portion is a sea area.

[0020] Figure 4 shows an example of a satellite image captured by SAR in the same region as the example satellite image in Figure 3, where ships and other objects are present in the same region. In the example image in Figure 4, ships and other objects are depicted as elliptical shapes in the ocean region. In the example image in Figure 4, objects are also depicted as elliptical and circular shapes in the land region. The image analysis system 10 detects, for example, the object to be analyzed in the satellite image of the example in Figure 4. If the object to be analyzed is a ship, the image analysis system 10 detects the ship in the satellite image of the example in Figure 4 using, for example, an image recognition model. Then, if a ship is detected from the satellite image of the example in Figure 4, the image analysis system 10 estimates the probability that the detected object is a ship based on environmental data in the region shown in the satellite image of the example in Figure 4.

[0021] Figure 5 shows an example of a display screen showing the detection results and estimated detection accuracy of the object being analyzed. The example screen in Figure 5 shows the detection results and estimated detection accuracy of the object being analyzed when the object being analyzed is detected in the satellite image shown in the example in Figure 4. In the example screen in Figure 5, the area where the object being analyzed is detected is enclosed by a solid or dashed rectangle. In the example screen in Figure 5, the area enclosed by a solid rectangle represents, for example, an area where the detection accuracy is above a certain standard. In the example screen in Figure 5, the area enclosed by a dashed rectangle represents, for example, an area where the detection accuracy is below a certain standard. By outputting an image like the example screen in Figure 5, the analyst can, for example, identify areas where the detection accuracy by the image recognition model is low and requires more detailed analysis. The analyst can, for example, consider that the estimated accuracy based on environmental data suggests a low probability of it being a ship, and then analyze the area enclosed by the dashed rectangle in detail to identify the object in the satellite image.

[0022] Here, we will describe a specific example of the configuration of the image analysis system 10. Figure 6 is a diagram showing an example of the configuration of the image analysis system 10.

[0023] The image analysis system 10, as a basic configuration, includes a satellite image acquisition unit 12, an environmental data acquisition unit 13, a detection unit 14, an estimation unit 15, and an output unit 16. The image analysis system 10 also includes, for example, an acquisition unit 11 and a storage unit 17.

[0024] The acquisition unit 11 acquires, for example, information about the object to be analyzed. The acquisition unit 11 acquires, for example, information about the object to be analyzed from the terminal device 20. The information about the object to be analyzed is, for example, information indicating at least one of the region to be analyzed and the object to be analyzed.

[0025] The region to be analyzed is, for example, information indicating the scope of the analysis. The region to be analyzed may be predetermined. For example, if the region to be analyzed by the image analysis system 10 is limited to a specific region, the acquisition unit 11 does not need to acquire information indicating the region to be analyzed.

[0026] Information about the object to be analyzed is, for example, information indicating what is being detected. The object to be analyzed is, for example, a ship. If the object to be analyzed is a ship, the ship to be analyzed is, for example, a ship that is sailing or anchored. The ship to be analyzed may be both a ship sailing and an anchored ship. The ship to be analyzed may also be a ship located on land. The object to be analyzed may be an aircraft, vehicle, building, or storage item located on land. The object to be analyzed is not limited to the above. The object to be analyzed may also be predetermined. For example, if the image analysis system 10 is used only for analyzing the presence or absence of a ship, the acquisition unit 11 does not need to acquire new information indicating the detection of the presence of a ship.

[0027] Information indicating the object being analyzed may also be the type of object being analyzed. For example, if the object being analyzed is a ship, the type of object being analyzed may be, for example, the type of ship. The type of ship may be, for example, a large ship, a medium ship, or a small ship. The type of ship may be a tanker, a passenger ship, a cargo ship, a ferry, a workboat, or a fishing vessel. The type of ship may be a ship capable of carrying an aircraft squadron, a conventional ship, or a submarine. The type of ship is not limited to the above. Also, the type of object is not limited to the above.

[0028] Information regarding the subject of analysis may include information indicating the timing of the analysis. This information could include, for example, the timing of when the satellite imagery being analyzed was acquired and the timing of when the environmental data was observed. The timing of the analysis can be defined by date and time, days, or a period. The period can be defined, for example, using the first and last days of the period. However, the method of defining the period is not limited to the above. Furthermore, information regarding the subject of analysis may also include information indicating which areas within the region being analyzed. This information is not limited to the above. Additionally, information regarding the subject of analysis may be defined, for example, by the person conducting the analysis.

[0029] The satellite image acquisition unit 12 acquires satellite images of the region to be analyzed. The satellite image acquisition unit 12 acquires satellite images to which information such as the location and date and time of acquisition is attached. The information about the location where the image was taken is, for example, the latitude and longitude of the central point of the image. The information about the location where the image was taken is not limited to the above, but can be any information that identifies the location where the image was taken. The satellite image acquisition unit 12 acquires satellite images of the region to be analyzed, for example, via the satellite image management server 30. Alternatively, the satellite image acquisition unit 12 may acquire satellite images of the region to be analyzed via a storage medium.

[0030] If the acquisition unit 11 has acquired information indicating the region to be analyzed, the satellite image acquisition unit 12 may acquire satellite images of the region indicated by the information indicating the region to be analyzed. If the acquisition unit 11 has acquired information indicating the timing of the analysis target, the satellite image acquisition unit 12 may acquire satellite images taken at the timing indicated by the information indicating the timing of the analysis target.

[0031] Satellite images are, for example, images captured by SAR mounted on an artificial satellite. However, satellite images are not limited to images captured by SAR. Furthermore, the satellite image acquisition unit 12 may acquire satellite images captured by multiple imaging methods. For example, the satellite image acquisition unit 12 acquires satellite images captured by SAR and optical images in the visible light region. The satellite image acquisition unit 12 may also acquire satellite images in the infrared region. Satellite images captured by multiple imaging methods are not limited to satellite images that capture the same area, as long as the region to be analyzed is included in each of them.

[0032] The environmental data acquisition unit 13 acquires environmental data for the region to be analyzed. The environmental data acquisition unit 13 acquires environmental data that includes, for example, the observation location and the date and time of observation. The information of the observation location is, for example, the latitude and longitude of the location where the environmental data was observed. The information of the observation location is not limited to the above. The environmental data may also be time-series data of observed values ​​of the environment. The environmental data acquisition unit 13 acquires environmental data for the region to be analyzed from, for example, an environmental data management server 40. The environmental data acquisition unit 13 may acquire environmental data for the region to be analyzed from multiple environmental data management servers 40. The environmental data acquisition unit 13 may also acquire environmental data as environmental data from an environmental data provision server operated by the Japan Meteorological Agency or other government agencies.

[0033] Environmental data includes, for example, data on at least one item from among water depth, oceanic weather, and topographic data. Oceanic weather data includes data on at least one item from among water temperature, air temperature, wave conditions, and wind speed. Oceanic weather data is not limited to those listed above. Environmental data is also not limited to those listed above. The environmental data acquisition unit 13 may acquire environmental data for items set based on the object being analyzed. For example, if the object being analyzed is a ship, the environmental data acquisition unit 13 acquires water depth, wave conditions, and topographic data as environmental data necessary for estimating the presence or absence of a ship. The items for estimating the presence or absence of a ship are set, for example, by the person in charge of the analysis or the operator of the image analysis system 10.

[0034] The environmental data acquisition unit 13 may acquire environmental data for items set based on at least one of the region or timing of the analysis target. For example, if the analysis target is a region where marine organisms exist in winter, and the analysis is performed on satellite images taken in winter, the environmental data acquisition unit 13 acquires environmental data for items necessary for estimating the presence or absence of marine organisms, and which was observed in winter.

[0035] If the acquisition unit 11 has acquired information indicating the region to be analyzed, the environmental data acquisition unit 13 may acquire environmental data observed in the region indicated by the information indicating the region to be analyzed. If the acquisition unit 11 has acquired information indicating the timing of the analysis target, the environmental data acquisition unit 13 may acquire environmental data observed at the timing indicated by the information indicating the timing of the analysis target.

[0036] The environmental data acquisition unit 13 may acquire environmental data such that the time of satellite image acquisition corresponds to the period of observation. Corresponding satellite image acquisition time and period of observation of environmental data means that the time of satellite image acquisition falls within a period in which the fluctuations in environmental data are within the range of fluctuations that can occur under assumed conditions. For example, if the tidal current is constant during the winter, the satellite image acquired at some point in the winter and the tidal current data acquired at some point in the winter may correspond in terms of acquisition time and observation period. Also, if the average value affects the presence or absence of an object, the environmental data acquisition unit 13 acquires average value data of environmental data for a predetermined period including the time of satellite image acquisition. The predetermined period is set, for example, by the person in charge of analysis. Also, for example, in the case of data that changes from day to day, such as wave height, the environmental data acquisition unit 13 acquires wave height data such that the date and time the satellite image was acquired and the date and time the observation occurred are the same. Data that are the same includes, for example, data that differs within a range that can be considered the same in the processing of the estimation unit 15. Furthermore, if the environmental data is data that does not normally change, such as terrain, the environmental data acquisition unit 13 may acquire environmental data at any point in time when no change has occurred. The timing at which the observed environmental data is acquired is not limited to the above.

[0037] The environmental data acquisition unit 13 may acquire information that identifies vessels present in the area under analysis. The environmental data acquisition unit 13 may acquire vessel identification signals from the Automatic Identification System (AIS) in the area under analysis. The environmental data acquisition unit 13 acquires vessel identification signals from the AIS from a monitoring server that monitors vessel navigation in the area under analysis. The information that identifies vessels present in the area under analysis is not limited to vessel identification signals from the AIS.

[0038] The detection unit 14 detects the object to be analyzed in the satellite image. The detection unit 14 detects the object to be analyzed in the satellite image using, for example, an image recognition model. The detection unit 14 detects the area in the satellite image in which the object to be analyzed is located using, for example, an image recognition model. The detection unit 14 may further detect the type of the object to be analyzed in the satellite image using an image recognition model.

[0039] The image recognition model is, for example, a learning model that takes satellite images as input and estimates objects that appear in the satellite images. The detection unit 14 may use the image recognition model to detect objects other than the object being analyzed that appear in the satellite images. For example, if the object being analyzed is a ship, and marine organisms are present in the area being analyzed, the detection unit 14 may use the image recognition model to detect marine organisms that appear in the satellite images.

[0040] Image recognition models are generated, for example, by learning the relationship between satellite images and the objects depicted in those images. Image recognition models are also generated, for example, by learning the relationship between satellite images and the regions in which the object to be analyzed is depicted in the satellite images. Image recognition models may also be generated by learning the relationship between satellite images and the names of the objects depicted in those images. Image recognition models may also be generated by learning the relationship between satellite images, the regions in which objects are depicted, and the names of those objects. Image recognition models are generated, for example, by deep learning using neural networks. The training data and training algorithms used to generate image recognition models are not limited to those described above. Furthermore, image recognition models are generated, for example, in a system outside the image analysis system 10. Image recognition models may also be generated in a generation unit (not shown) provided by the image analysis system 10.

[0041] The detection unit 14 may detect the object to be analyzed based on the brightness changes in the satellite image. For example, if the satellite image only shows ocean areas and there are no changes in topography, the detection unit 14 detects the outline of the object in the satellite image based on the brightness changes between pixels in the satellite image. Then, the detection unit 14 determines whether the object in the satellite image is the object to be analyzed based on at least one of the size of the area enclosed by the outline and the shape of the outline. If the object in the image is determined to be the object to be analyzed, the detection unit 14 detects, for example, that the object to be analyzed exists in the area enclosed by the outline. The method of detecting the object to be analyzed is not limited to the above.

[0042] The estimation unit 15 estimates the accuracy of detection based on environmental data. For example, the estimation unit 15 estimates an index indicating the likelihood that the object detected by the detection unit 14 is the object to be analyzed, based on environmental data, as the accuracy of detection. For example, the estimation unit 15 estimates the accuracy of detecting the object to be analyzed in the detection result by the image recognition model, based on environmental data. That is, the estimation unit 15 estimates the likelihood of the detection result by the image recognition model, based on environmental data. For example, the estimation unit 15 estimates the presence or absence of the object to be analyzed, based on environmental data. Then, the estimation unit 15 estimates the accuracy of detection based on the estimated presence or absence of the object to be analyzed.

[0043] The estimation unit 15 may estimate the accuracy of the detection based on whether the object detected by the image recognition model is an object other than the object to be analyzed. For example, the estimation unit 15 estimates candidate objects when the object detected by the image recognition model is an object other than the object to be analyzed. The estimation unit 15 then estimates the accuracy of the detection based on, for example, the probability that the candidate object exists. For example, when the environmental data is suitable for the existence of the candidate object, the probability that the candidate object exists increases. When there is a high probability that an object other than the object to be analyzed exists, the probability that the object detected by the image recognition model is the object to be analyzed decreases. When the object to be analyzed is a ship, the estimation unit 15 estimates candidate objects other than ships when the object detected by the image recognition model is not a ship. When the object to be analyzed is a ship, and it is estimated from the environmental data of the area to be analyzed that marine organisms exist, the estimation unit 15 estimates marine organisms as candidate objects other than ships. Furthermore, when environmental data indicates a high probability of marine life being present, the estimation unit 15 estimates, for example, that the likelihood of an object detected as a ship by the image recognition model being a ship is low. In other words, the estimation unit 15 estimates that the accuracy of the detection result identified as a ship by the image recognition model is low.

[0044] The estimation unit 15 estimates the accuracy of detection based, for example, on the detection result of the image recognition model and the presence or absence of the object to be analyzed in the area to be analyzed, which is estimated using environmental data. The presence or absence of the object to be analyzed in the area to be analyzed, which is estimated using environmental data, is, for example, the possibility that the object to be analyzed may exist in the area to be analyzed, which is estimated based on environmental data. The estimation unit 15 estimates the presence or absence of the object to be analyzed in the area to be analyzed, for example, based on criteria that define the relationship between the object to be analyzed and the environmental data. Then, when the image recognition model detects the object to be analyzed, the estimation unit 15 estimates the accuracy of the detection result based on the estimation result of the presence or absence of the object to be analyzed. The criteria that define the relationship between the object to be analyzed and the environmental data are criteria for estimating the presence or absence of the object to be analyzed based on environmental data. The criteria that define the relationship between the object to be analyzed and the environmental data are criteria for determining under what environmental conditions the object to be analyzed is likely to exist. Alternatively, the criteria that define the relationship between the object to be analyzed and the environmental data may also be criteria for determining under what environmental conditions the object to be analyzed is unlikely to exist.

[0045] If the object being analyzed is a ship, the estimation unit 15 estimates the accuracy of the detection based on the result of the image recognition model detecting a ship and the probability that a ship may exist in the area being analyzed, which is estimated using environmental data. For example, if the image recognition model detects a ship, which is the object being analyzed, and the terrain data included in the environmental data indicates terrain unsuitable for anchoring, the estimation unit 15 estimates that the detected object is unlikely to be a ship because the terrain is unsuitable for anchoring. Also, for example, if the image recognition model detects a large ship, which is the object being analyzed, and the water depth data included in the environmental data indicates a depth unsuitable for navigation and anchoring, the estimation unit 15 estimates that the detected object is unlikely to be a ship because the water depth is unsuitable for navigation and anchoring. Furthermore, for example, if the image recognition model detects multiple ships approaching each other, which are the objects being analyzed, and the ocean current and wave data included in the environmental data indicates values ​​unsuitable for approaching ships, the estimation unit 15 estimates that the multiple detected objects are unlikely to be ships. The criteria defining the relationship between the object being analyzed and the environmental data are not limited to those stated above. These criteria may be set, for example, by the person performing the analysis or the operator of the image analysis system 10.

[0046] The estimation unit 15 estimates the accuracy of detection based on a score indicating the likelihood of an object existing, for example, using environmental data. The estimation unit 15 estimates a score indicating the likelihood of an object being analyzed existing, for example, based on environmental data. Then, the estimation unit 15 estimates the accuracy of detection based on the estimated score. The score indicating the likelihood of an object existing is, for example, an index indicating whether the environment is suitable for an object to exist. For example, the score indicating the likelihood of a ship existing is an index indicating whether the environment is suitable for a ship to exist. An environment in which a ship can exist is, for example, an environment in which the water depth, topography, and wave conditions are suitable for at least one of the navigation and anchoring of a ship. The estimation unit 15 calculates a score for each item of the environmental data. Then, the estimation unit 15 estimates a score indicating the likelihood of an object existing by, for example, summing the scores for each item of the environmental data. The relationship between the values ​​of the environmental data and the scores is set up, for example, as a table for each item of the environmental data. The relationship between the values ​​of the environmental data and the scores may also be set up using a function in which the values ​​of the environmental data are the explanatory variables and the scores are the dependent variables. The relationship between environmental data values ​​and scores is set, for example, by the person performing the analysis or the operator of the image analysis system 10.

[0047] The estimation unit 15 may estimate a score indicating the possibility of the presence of objects other than the object being analyzed, based on the environmental data. Objects other than the object being analyzed are, for example, objects that the image recognition model may mistakenly recognize as the object being analyzed. If a score indicating the possibility of the presence of objects other than the object being analyzed is calculated, the estimation unit 15 estimates the accuracy of detection based on the score indicating the possibility of the presence of objects other than the object being analyzed. The estimation unit 15 estimates a score indicating the possibility of the presence of objects other than the object being analyzed that may exist in the area being analyzed. "May exist in the area being analyzed" means, for example, that the object has existed in the area being analyzed in the past. "May exist in the area being analyzed" may also mean that the object is frequently present in areas where at least one of the topography and environment is the same as or similar to the area being analyzed. For example, if the object being analyzed is a ship and there is a possibility of marine life being present, the estimation unit 15 estimates a score indicating the possibility of marine life being present. For example, if the score for marine life is higher than that for the ship, the estimation unit 15 estimates that the object detected as a ship is likely to be marine life. In other words, if the score for marine organisms is higher than that for ships, the estimation unit 15 estimates that the object detected as a ship is unlikely to be a ship.

[0048] The estimation unit 15 may estimate the accuracy of detection based on the probability that the object is the target of analysis in the detection results of the image recognition model and a score indicating the likelihood of the object existing, estimated from the environmental data. For example, if the probability that the object is the target of analysis in the estimation results of the image recognition model is Pm, the score indicating the likelihood of the object existing, estimated from the environmental data, is S, and the accuracy of detection is Pd, then the estimation unit 15 estimates the accuracy of detection using the formula Pd = Pm × S. The formula for estimating the accuracy of detection is not limited to the above. Furthermore, the method of estimating the accuracy of detection is not limited to the above.

[0049] The estimation unit 15 estimates the accuracy of detection based on environmental factors that affect the accuracy of the detection results of the image recognition model estimated using environmental data. The estimation unit 15 may also estimate the accuracy of detection based on environmental factors that affect the presence of an object that the image recognition model estimated using environmental data may mistakenly identify as the object being analyzed. Environmental factors that affect the accuracy of the detection results of the image recognition model that detects the object being analyzed are, for example, items of the environmental data that have a large influence on whether or not an object is present. The estimation unit 15 estimates, for example, that the item with the highest score is the environmental factor that affects the accuracy of the detection results. The estimation unit 15 may also estimate, for example, that the item whose score meets a certain criterion is the environmental factor that affects the accuracy of the detection results. The estimation unit 15 may also estimate that the items from the top to a predetermined rank are the environmental factors that affect the accuracy of the detection results.

[0050] Furthermore, the estimation unit 15 may estimate environmental factors that affect the accuracy of the detection result of the image recognition model that detects the object to be analyzed, based on the environmental data, as reasons for estimating the accuracy of detection. For example, the estimation unit 15 estimates environmental factors that affect the probability of the existence of the object to be analyzed, estimated using the environmental data, as reasons for estimating the accuracy of detection. Also, the estimation unit 15 estimates environmental factors that affect the probability of the existence of objects other than the object to be analyzed, estimated using the environmental data, as reasons for estimating the accuracy of detection. If the object to be analyzed is a ship, the estimation unit 15 estimates the reasons for estimating the accuracy of detection based, for example, on environmental factors that affect the probability of the existence of a ship or environmental factors that affect the probability of the existence of objects other than ships.

[0051] The estimation unit 15 may estimate the accuracy of detection based on objects present in the area under analysis, which are estimated using an estimation model that estimates present objects from environmental data. The estimation model is, for example, a learning model that has learned the relationship between environmental data and present objects. The estimation model is generated, for example, by deep learning using a neural network.

[0052] The estimation unit 15 may estimate the accuracy of detection based on objects present in the area under analysis, estimated using an estimation model capable of identifying the reason for estimation. An estimation model capable of outputting the reason for estimation may be generated using, for example, a learning algorithm based on factorized asymptotic Bayesian inference. When learning using a learning algorithm based on factorized asymptotic Bayesian inference, the learner divides the data of each item of environmental data into cases using decision tree-like rules, with the existing objects as input data and the data itself as ground truth data. The learner then generates a learning model that predicts the degree of realization using a linear model that combines different explanatory variables in each case. The learner generates the learning model by sequentially performing the following processes: optimizing the data case division conditions, generating a prediction model by optimizing the combination of explanatory variables, and deleting unnecessary prediction models. The estimation model may also be a learning model that identifies the reason for estimation based on the change in the estimation result with respect to the variation amount of each item of environmental data. The learning algorithm for generating the estimation model is not limited to the above. The estimation model may also be generated in, for example, a system outside the image analysis system 10. The estimation model may also be generated in a generation unit (not shown) provided by the image analysis system 10.

[0053] The estimation unit 15 may estimate the accuracy of detection based on areas where the object to be analyzed is not found, which is estimated using environmental data. When the object to be analyzed is a ship, the estimation unit 15 estimates areas where ships are not found, for example, based on environmental data. When the object to be analyzed is a ship, the estimation unit 15 estimates that the accuracy of detection is low when a ship is found in an area where ships are estimated not to exist. The estimation unit 15 estimates a score indicating the probability of the object to be analyzed existing in each area, based on environmental data. Then, the estimation unit 15 estimates that the accuracy of detection is low when a ship is found in an area where the score is below a certain threshold.

[0054] The estimation unit 15 may estimate the accuracy of detection based on areas where no predetermined objects other than the object to be analyzed, estimated using environmental data, exist. The predetermined objects are, for example, objects that the image recognition model is expected to misidentify as the object to be analyzed. The estimation unit 15 estimates a score indicating the probability of the predetermined object existing in each area based on the environmental data. The estimation unit 15 then estimates, for example, that the predetermined object does not exist in areas where the score is below a certain threshold. The estimation unit 15 estimates, for example, that the accuracy of detection is high when the object to be analyzed is detected in an area where the predetermined object does not exist. The estimation unit 15 may estimate areas where there is a high probability that predetermined objects other than the object to be analyzed exist, based on the environmental data. The estimation unit 15 estimates, for example, that the accuracy of detection is low when the object to be analyzed is detected in an area where there is a high probability that the predetermined object exists. The estimation unit 15 estimates a score indicating the probability of the predetermined object existing in each area based on the environmental data. The estimation unit 15 then estimates, for example, that areas where the score is above a certain threshold are areas where there is a high probability that the predetermined object exists. The specified object is, for example, defined by the person performing the analysis.

[0055] The estimation unit 15 may further estimate the accuracy of detection using the ship's identification signal. For example, the estimation unit 15 estimates the accuracy of detection based on the AIS identification signal. For example, the estimation unit 15 estimates that the accuracy of detection is high when the type of ship indicated by the detection result matches the type of ship indicated by the AIS identification signal. For example, the estimation unit 15 may estimate that the accuracy of detection is high the more items that match between the detection result and the information indicated by the AIS identification signal. For example, the estimation unit 15 estimates that the accuracy of detection is higher when the type of ship and the course match between the detection result and the information indicated by the AIS identification signal than when only the type of ship matches.

[0056] The output unit 16 outputs the detection result and information indicating the accuracy of the detection. The output unit 16 outputs the detection result and information indicating the accuracy of the detection to, for example, a terminal device 20. The output unit 16 outputs, for example, a satellite image, the detection result of the object to be analyzed in the satellite image, and information indicating the accuracy of the detection. The output unit 16 outputs a detection result indicating the area in which the object to be analyzed was detected by, for example, enclosing the area in which the object to be analyzed was detected with a geometric shape on the satellite image. The geometric shape enclosing the area in which the object to be analyzed was detected is, for example, a rectangle. The geometric shape enclosing the area in which the object to be analyzed was detected is not limited to a rectangle. The output unit 16 outputs, for example, a satellite image in which the line enclosing the area in which the object to be analyzed was detected on the satellite image is changed according to the level of accuracy of the detection. The output unit 16 outputs, for example, a satellite image in which at least one of the shape, thickness, and color of the line enclosing the area in which the object to be analyzed was detected on the satellite image is changed according to the level of accuracy of the detection. The output unit 16 may output satellite images in which the shape of the figure surrounding the area where the object to be analyzed is detected is changed according to the level of detection accuracy. The output unit 16 may also output a numerical value indicating the level of detection accuracy superimposed on the satellite image. Furthermore, the output unit 16 may also output a numerical value, character, or symbol indicating the level of detection accuracy superimposed on the satellite image. The form in which the detection accuracy is output is not limited to the above.

[0057] The output unit 16 may output candidate objects other than the target object estimated by the estimation unit 15 based on environmental data. The output unit 16 outputs information on candidate objects other than the target object estimated by the estimation unit 15 based on environmental data, associating it with the region where the image recognition model detected the object. The output unit 16 outputs the names of the candidate objects other than the target object estimated by the estimation unit 15 based on environmental data, associating them with the region where the image recognition model detected the object.

[0058] The output unit 16 may further output the reason for estimating the accuracy of the detection. For example, the output unit 16 may output the reason for estimating the accuracy of the detection estimated by the estimation unit 15 based on environmental data. For example, the output unit 16 may output the reason for estimating that the detected object is the object to be analyzed. For example, the output unit 16 may output the reason for estimating that the detected object is an object other than the object to be analyzed.

[0059] The output unit 16 may output satellite images in which objects of the same type as the detected object have been identified as reference images. For example, if the detected object is a large ship, the output unit 16 may output satellite images in which large ships have been identified in past analyses as reference images. The output unit 16 may also output satellite images of objects other than the object being analyzed that may exist in the area being analyzed as reference images. For example, if the object being analyzed is a ship, the output unit 16 may output satellite images of objects other than ships.

[0060] The output unit 16 may output satellite images using a different imaging method than the satellite image being analyzed. For example, if the satellite image being analyzed is an image captured by SAR, the output unit 16 may further output an optical image in the visible light region that includes the same range as the area in which the SAR image was captured.

[0061] The output unit 16 may, for example, output an enlarged image of the area selected by the person performing the analysis. For example, if the output unit 16 detects that an area in which the object to be analyzed is detected on the display screen has been selected by the person performing the analysis, it will output an enlarged image of the selected area. The output unit 16 may also output an image of the area selected by the person performing the analysis with a higher resolution than other areas. The output unit 16 may also output a satellite image as a reference image for the area selected by the person performing the analysis, which has been identified in past analyses as containing the object to be analyzed. The output unit 16 may also output a satellite image as a reference image for the area selected by the person performing the analysis, which has been identified in past analyses as containing candidate objects other than the object to be analyzed.

[0062] Figure 7 shows an example of a display screen that shows the reason for the estimation of the detection accuracy. In the example of the display screen in Figure 7, the area enclosed by a solid rectangle is, for example, an area where the detection accuracy is above the standard. In the example of the display screen in Figure 7, the area enclosed by a dashed rectangle is, for example, an area where the detection accuracy is below the standard. In addition, in the example of the display screen in Figure 7, the area enclosed by a dashed line shows the possibility that the detected object is a "seal". In addition, in the example of the display screen in Figure 7, the reason for estimating that the object is not the object being analyzed is shown to be "temperature" and "current". In the example of the display screen in Figure 7, the observational data for temperature and current among the environmental data are suitable for the presence of seals, indicating that the possibility of it being a ship is low and the possibility of it being a seal is high.

[0063] The storage unit 17 stores, for example, data used for analyzing satellite images. The storage unit 17 stores, for example, satellite images acquired by the satellite image acquisition unit 12. The storage unit 17 stores, for example, environmental data acquired by the environmental data acquisition unit 13. The storage unit 17 stores, for example, a table showing the relationship between environmental data and a score indicating the likelihood of an object being present. The storage unit 17 stores, for example, an image recognition model. The storage unit 17 stores, for example, an estimation model. The image recognition model and the estimation model may be stored in storage means other than the storage unit 17. The storage unit 17 stores, for example, the detection results of the image recognition model. The storage unit 17 may store, for example, the reason for estimating the accuracy of the detection. The storage unit 17 may store, for example, satellite images that have been identified in past analyses as containing the object to be analyzed as reference images. The storage unit 17 may store, for example, satellite images that have been identified in past analyses as containing objects other than the object to be analyzed as reference images.

[0064] Terminal device 20 is, for example, a terminal device used by an analyst to analyze satellite images. Terminal device 20 obtains, for example, satellite images, detection results of objects in the satellite images, and detection accuracy from the output unit 16 of the image analysis system 10. Then, terminal device 20 outputs, for example, satellite images, detection results of objects in the satellite images, and detection accuracy to a display device (not shown). Terminal device 20 obtains, for example, the reason for estimating the detection accuracy from the output unit 16 of the image analysis system 10. Then, terminal device 20 outputs, for example, the reason for estimating the detection accuracy to a display device (not shown).

[0065] The terminal device 20 may, for example, acquire information about the object to be analyzed, which is input by the user. The terminal device 20 may, for example, output information about the object to be analyzed to the acquisition unit 11 of the image analysis system 10.

[0066] The terminal device 20 can be, for example, a personal computer, a tablet computer, or a smartphone. The terminal device 20 is not limited to the examples above.

[0067] The satellite image management server 30 manages images of the Earth's surface taken by an imaging device mounted on a satellite, for example. The satellite image management server 30 acquires satellite images of the Earth's surface taken by an imaging device mounted on a satellite from a ground station that communicates with the satellite, for example. The satellite image management server 30 associates the acquired satellite image with the date and time of acquisition and the location where it was acquired. The satellite image management server 30 may identify the location where the image was acquired from the satellite's position at the time of acquisition and imaging parameters attached to the satellite image. The imaging parameters include, for example, the direction from which electromagnetic waves in the frequency band used for imaging are transmitted, the transmission angle of electromagnetic waves relative to the Earth's surface, and the reception accuracy of electromagnetic waves reflected from the Earth's surface. For example, when the satellite image management server 30 receives a request for satellite images from the image analysis system 10, it outputs the requested satellite image, along with the date and time of acquisition and the acquisition location of the satellite image, to the satellite image acquisition unit 12 of the image analysis system 10. Furthermore, there may be multiple satellite image management servers 30. For example, satellite images may be stored on different servers depending on the management entity of the satellite that performed the imaging. The number of satellite image management servers 30 can be set as appropriate.

[0068] The environmental data management server 40 manages environmental data, for example. The environmental data management server 40 acquires environmental data, for example. The environmental data management server 40 stores the acquired environmental data in association with observation date and time and observation location information. The environmental data management server 40 stores observation data such as water depth, water temperature, air temperature, current, wave height, and wind speed, for example. The environmental data management server 40 also stores topographic data, for example. Furthermore, when the environmental data management server 40 receives a request for environmental data from the image analysis system 10, for example, it outputs the requested environmental data, the observation date and time, and the observation location of the environmental data to the environmental data acquisition unit 13 of the image analysis system 10. There may be multiple environmental data management servers 40. For example, each entity that observes environmental data may store it on a different server. The number of environmental data management servers 40 can be set as appropriate.

[0069] An example of the operation of the image analysis system 10 will be described. Figure 8 shows an example of the operation flow when the image analysis system 10 detects objects in satellite images and estimates the accuracy of the detection.

[0070] The satellite image acquisition unit 12 acquires satellite images of the area to be analyzed (step S11). The satellite image acquisition unit 12 acquires satellite images of the area to be analyzed from, for example, the satellite image management server 30.

[0071] Furthermore, the environmental data acquisition unit 13 acquires environmental data for the region to be analyzed (step S12). The environmental data acquisition unit 13 acquires environmental data for the region to be analyzed from, for example, the environmental data management server 40.

[0072] Once satellite images and environmental data of the area to be analyzed are acquired, the detection unit 14 detects the object to be analyzed in the satellite image (step S13). The detection unit 14 detects the object to be analyzed in the satellite image, for example, using an image recognition model.

[0073] If an object to be analyzed is detected in the satellite image (Yes in step S14), the estimation unit 15 estimates the accuracy of detecting the object to be analyzed based on the environmental data (step S15). For example, the estimation unit 15 estimates the accuracy of the detection result, which is the accuracy of the detection of the object to be analyzed by the image recognition model, based on the environmental data.

[0074] Once the accuracy of detecting the object to be analyzed is estimated, the output unit 16 outputs the detection result of the object to be analyzed and the detection accuracy (step S16). The output unit 16 outputs the detection result of the object to be analyzed and the detection accuracy to, for example, the terminal device 20.

[0075] When outputting the results of detecting the object to be analyzed and the accuracy of the detection, if there are images in which the process of detecting the object to be analyzed has not been performed (No in step S17), the process returns to step S13, and the detection unit 14 detects the object to be analyzed in the satellite images in which the object to be analyzed has not been detected.

[0076] In step S16, when the detection results of the object to be analyzed and the accuracy of the detection are output, if the process of detecting the object to be analyzed has been performed for all satellite images (Yes in step S17), the image analysis system 10 terminates the process of detecting objects in the satellite images and estimating the accuracy of the detection.

[0077] In step S14, if the object to be analyzed is not detected in the satellite image (No in step S14), or if there is a satellite image in which the process of detecting the object to be analyzed has not been performed (No in step S17), the process returns to step S13, and the detection unit 14 detects the object to be analyzed in the other satellite image in which the analysis has not been performed.

[0078] In step S14, if the object to be analyzed is not detected in the satellite image (No in step S14), and if the process of detecting the object to be analyzed has been performed for all satellite images (Yes in step S17), the image analysis system 10 terminates the process of detecting the object in the satellite image and estimating the accuracy of the detection.

[0079] The image analysis system 10 detects the target object from satellite images of the area to be analyzed. Based on environmental data of the area to be analyzed, the image analysis system 10 estimates the probability of detecting the target object. The image analysis system 10 then outputs the detection result of the target object and information indicating the probability of detection. By outputting the detection result of the target object and information indicating the probability of detection in this way, the person in charge of analyzing the satellite image can refer to the information indicating the probability of detection output as the probability of detection to confirm whether the object in the satellite image is the target object. For example, if there is a high probability of marine life being present based on environmental data, the person in charge of analysis can consider the possibility of marine life being present when confirming whether the object in the satellite image is the target object. Also, if there is a high probability of no ships being present based on topography and water currents, the person in charge of analysis can consider that if there is a ship, it is an unusual ship when analyzing the satellite image. In this way, by analyzing satellite images using the detection result of the target object and information indicating the probability of detection, it becomes easier to identify objects present in the area to be analyzed. For this reason, objects present in the area to be analyzed can be easily analyzed by using the image analysis system 10.

[0080] Furthermore, when outputting the reasons for estimating the accuracy of detection, the analyst can determine whether an object in the image is the target object by considering, for example, the reasons why it is likely to be the target object or the reasons why it is likely not to be the target object. Therefore, outputting the reasons for estimating the accuracy of detection makes it easier to analyze objects present in the area being analyzed.

[0081] Each process in the image analysis system 10 may be distributed and executed across multiple information processing devices connected via a network. For example, the processing in the detection unit 14 and the processing in the estimation unit 15 may be performed on separate information processing devices. The choice of which of the multiple information processing devices performs each process in the image analysis system 10 can be set as appropriate.

[0082] Each process in the image analysis system 10 can be implemented by executing a computer program on a computer. Figure 9 shows an example of the configuration of a computer 100 that executes the computer programs that perform each process in the image analysis system 10. The computer 100 includes a CPU (Central Processing Unit) 101, memory 102, storage device 103, input / output interface (I / F) 104, and communication interface (I / F) 105.

[0083] The CPU 101 reads and executes computer programs that perform various processes from the storage device 103. The CPU 101 may be composed of a combination of multiple CPUs. Alternatively, the CPU 101 may be composed of a combination of a CPU and another type of processor. For example, the CPU 101 may be composed of a combination of a CPU and a GPU (Graphics Processing Unit). The memory 102 is composed of DRAM (Dynamic Random Access Memory) or the like, and temporarily stores computer programs executed by the CPU 101 and data being processed. The storage device 103 stores computer programs executed by the CPU 101. The storage device 103 is composed of, for example, a non-volatile semiconductor storage device. Other storage devices such as hard disk drives may be used for the storage device 103. The input / output interface 104 is an interface that receives input from the operator and outputs display data, etc. The communication interface 105 is an interface that sends and receives data between the terminal device 20, the satellite image management server 30, the environmental data management server 40, and other information processing devices. Furthermore, the terminal device 20, the satellite image management server 30, and the environmental data management server 40 may have the same configuration as the computer 100.

[0084] The computer programs used to execute each process can also be stored and distributed on a computer-readable recording medium that non-temporarily stores data. Examples of recording media include magnetic tapes for data recording and magnetic disks such as hard disks. Optical discs such as CD-ROMs (Compact Disc Read Only Memory) can also be used as recording media. Non-volatile semiconductor memory devices may also be used as recording media.

[0085] Some or all of the above embodiments may also be described as follows, but are not limited to the following:

[0086] [Note 1] Image acquisition means for acquiring satellite images of the region to be analyzed, An environmental data acquisition method for acquiring environmental data of the region to be analyzed, A detection means for detecting objects to be analyzed that appear in satellite images, An estimation means for estimating the accuracy of the detection based on the aforementioned environmental data, An output means that outputs the results of the detection and information indicating the accuracy of the detection. An image analysis system equipped with the following features.

[0087] [Note 2] The accuracy of the detection is an indicator of the likelihood that the object detected by the detection means is the object to be analyzed. The image analysis system described in Appendix 1.

[0088] [Note 3] The estimation means estimates the accuracy of the detection based on environmental factors that affect the accuracy of the detection result of the image recognition model that detects the object to be analyzed, which is estimated using environmental data. The image analysis system described in Appendix 2.

[0089] [Note 4] The estimation means estimates the accuracy of the detection based on environmental factors that influence the presence of an object that the image recognition model, estimated using environmental data, may mistakenly identify as the object being analyzed. The image analysis system described in Appendix 3.

[0090] [Note 5] The estimation means estimates the accuracy of the detection based on the detection result of the image recognition model and the presence or absence of the object to be analyzed estimated using the environmental data. The image analysis system described in Appendix 3.

[0091] [Note 6] The object being analyzed is a ship. The estimation means estimates the accuracy of the detection based on environmental factors indicating that the object detected by the image recognition model is not the vessel. An image analysis system as described in any of the appendices 3 to 5.

[0092] [Note 7] The estimation means estimates the accuracy of the detection based on the area where the vessel is not present, which is estimated using the environmental data. The image analysis system described in Appendix 6.

[0093] [Note 8] The output means outputs satellite images relating to objects other than the vessel estimated by the estimation means. The image analysis system described in Appendix 6 or 7.

[0094] [Note 9] The aforementioned environmental data is data relating to at least one of the following: air temperature, water temperature, water flow, water depth, and topography. An image analysis system as described in any of the appendices 6 to 8.

[0095] [Note 10] The estimation means further estimates the accuracy of the detection using the ship's identification signal. An image analysis system as described in any of the appendices 6 to 9.

[0096] [Note 11] The output means further outputs the reason for the estimation of the accuracy of the detection. An image analysis system as described in any of the appendices 1 through 10.

[0097] [Note 12] The output means outputs a satellite image in which the line surrounding the area where the object to be analyzed is detected is changed according to the level of detection accuracy. An image analysis system as described in any of the appendices 1 through 11.

[0098] [Note 13] We obtain satellite images of the area to be analyzed, Obtain environmental data for the area to be analyzed, Detect the object to be analyzed in the satellite image, Based on the aforementioned environmental data, the accuracy of the detection is estimated. The system outputs the results of the detection and information indicating the accuracy of the detection. Image analysis methods.

[0099] [Note 14] The process involves acquiring satellite images of the area to be analyzed, The process of acquiring environmental data for the region to be analyzed, The process of detecting objects to be analyzed in satellite images, Based on the aforementioned environmental data, a process is performed to estimate the accuracy of the detection. A process that outputs the results of the detection and information indicating the accuracy of the detection. A recording medium that non-temporarily stores an image analysis program that causes a computer to execute.

[0100] The present disclosure has been explained above using the embodiments described above as examples. However, the present disclosure is not limited to the embodiments described above. That is, the present disclosure can be applied in various forms that can be understood by those skilled in the art within the scope of the present disclosure. [Explanation of Symbols]

[0101] 10 Image Analysis Systems 11 Acquisition Department 12 Satellite Image Acquisition Unit 13. Environmental Data Acquisition Unit 14 Detection unit 15 Estimation part 16 Output section 17 Memory section 20 Terminal devices 30 Satellite Image Management Server 40. Environmental Data Management Server 100 Computers 101 CPU 102 memory 103 Storage device 104 Input / Output Interfaces 105 Communication I / F

Claims

1. Image acquisition means for acquiring satellite images of the region to be analyzed, An environmental data acquisition method that acquires environmental data, which is data obtained by observing the natural environment of the area to be analyzed, A detection means for detecting objects to be analyzed that appear in satellite images, An estimation means for estimating the accuracy of the detection based on the aforementioned environmental data, An output means that outputs the results of the detection and information indicating the accuracy of the detection. An image analysis system equipped with the following features.

2. The accuracy of the detection is an indicator of the likelihood that the object detected by the detection means is the object to be analyzed. The image analysis system according to claim 1.

3. The estimation means estimates the accuracy of the detection based on environmental factors that affect the accuracy of the detection result of the image recognition model that detects the object to be analyzed, which is estimated using environmental data. The image analysis system according to claim 2.

4. The estimation means estimates the accuracy of the detection based on environmental factors that influence the presence of an object that the image recognition model, estimated using environmental data, may mistakenly identify as the object being analyzed. The image analysis system according to claim 3.

5. The estimation means estimates the accuracy of the detection based on the detection result of the image recognition model and the presence or absence of the object to be analyzed estimated using the environmental data. The image analysis system according to claim 3.

6. The object being analyzed is a ship. The estimation means estimates the accuracy of the detection based on environmental factors indicating that the object detected by the image recognition model is not the vessel. The image analysis system according to any one of claims 3 to 5.

7. The estimation means estimates the accuracy of the detection based on the area where the vessel is not present, which is estimated using the environmental data. The image analysis system according to claim 6.

8. The output means outputs satellite images relating to objects other than the vessel estimated by the estimation means. The image analysis system according to claim 6.

9. We obtain satellite images of the area to be analyzed, We obtain environmental data, which is data obtained by observing the natural environment of the area to be analyzed. Detect the object to be analyzed in the satellite image, Based on the aforementioned environmental data, the accuracy of the detection is estimated. The system outputs the results of the detection and information indicating the accuracy of the detection. Image analysis methods.

10. The process involves acquiring satellite images of the area to be analyzed, The process involves obtaining environmental data, which is data obtained by observing the natural environment of the area being analyzed, and The process of detecting objects to be analyzed in satellite images, Based on the aforementioned environmental data, a process is performed to estimate the accuracy of the detection. A process that outputs the results of the detection and information indicating the accuracy of the detection. An image analysis program that causes a computer to perform the following actions.

Images