Learning model, signal processing device, flying object, and program
Patent Information
- Application Number
- JP2021207261
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2021-12-21
- Publication Date
- 2024-11-13
- Estimated Expiration
- 2041-06-18
AI Technical Summary
【0025】 本発明によれば、観測対象物の状況あるいは観測対象物の周囲の環境の適切な観測を可 能にする、学習モデル、信号処理装置、飛翔体、及びプログラムを提供することができる 。
Abstract
Description
Technical Field
[0001] The present invention relates to a learning model, a signal processing device, a flying object, and a program.
Background Art
[0002] Observations of the state of the earth's surface, including land and sea, using flying objects such as artificial satellites, airplanes, or drone devices, are widely carried out. Observation methods using artificial satellites include observation methods performed by acquiring optical images, observation methods performed by acquiring so-called SAR images, which are radar images obtained using synthetic aperture radar (SAR) technology, or observation methods performed by acquiring both optical images and SAR images and combining the two images. Patent Document 1 discloses a method for generating a synthetic image in which a radar image such as a SAR image and an optical image are combined to facilitate the discrimination of ground objects.
[0003]
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0004] An SAR image is generated based on a signal (hereinafter referred to as a received signal) corresponding to an electromagnetic wave (microwave) irradiated from an artificial satellite equipped with a radar device to an observation target and reflected by the observation target. The SAR image is generated by performing predetermined compression processing on the received signal.
[0005] In the compression process for generating SAR images, the received signal is filtered in the frequency domain. Filtering removes some of the data from the received signal. This filtering process then compresses the data. This reduces the amount of data in the received signal, thus easing the computational burden in the compression process. Removing the received signal through filtering results in the loss of detectable information from the received signal or information False positives may occur. Information loss or false positives may be due to the state of the observed object or its surroundings. This affects the accuracy or precision of environmental observations.
[0006] Therefore, the present invention provides high-precision or high-precision measurement of the state of the object being observed or the environment surrounding the object being observed. A learning model, signal processing device, projectile, and program that enable accurate observation. The purpose is to provide. [Means for solving the problem]
[0007] A learning model according to one aspect of the present invention is a learning model in which electromagnetic waves irradiated onto a first target region are reflected. A first received signal based on emitted electromagnetic waves is used as input, and a predetermined item corresponds to the first received signal. The system is trained using training data with the first metadata as output, and the electromagnetic field irradiated onto the second target area is measured. A second received signal based on the reflected electromagnetic wave from the wave is input, and corresponding to the second received signal, Outputs a second set of metadata containing specific items.
[0008] According to this embodiment, the learning model responds to the input of a received signal with respect to the received signal. It is trained and operates to output metadata. By using this trained model, For example, metadata that includes an item such as the total number of moving objects and buildings in the target area, It becomes possible to acquire data based on the received signal. From the received signal, for example, via SAR images... This allows metadata to be obtained without loss, thus preventing information contained in the received signal from being lost. Therefore, it is possible to observe metadata indicating the status of the object being observed with high accuracy or precision. Yes.
[0009] In the above embodiment, the learning model is generated based on the first received signal and the first received signal. The first generated signal is used as input and the first metadata is output as training data, and the second The second received signal and the second generated signal generated based on the second received signal are input, and the second metadata You may output the report.
[0010] According to this embodiment, the training data in the learning model is generated based on the first received signal. A first generated signal is further included. The first generated signal is, for example, based on the first received signal. This is a signal for generating SAR images (SAR signal). Like the SAR signal, the received signal Using a learning model that further includes the data generated based on this as input and outputs metadata In cases where the SAR signal can appropriately observe metadata of the target object, etc., observation This makes it possible to observe the condition of an object with high precision or accuracy.
[0011] In the above embodiment, the learning model indicates the first received signal and the environment in the first target region. The system is trained using training data that takes information as input and outputs first metadata, and the second received signal Information indicating the environment in the second target area may be input, and second metadata may be output.
[0012] According to this embodiment, the training data in the learning model includes the environment in the first target domain. The information to be shown is further included. The information indicating the environment in the first target area is, for example, meteorological conditions such as the weather in the first target area, or environmental conditions caused by artificial factors such as smoke. By using the input including the information indicating the environment in the second target area when outputting the second meta-information, it becomes possible to observe the second meta-information with high precision or high accuracy. In the above aspect, the learning model is trained using teacher data with the output of the first meta-information including the information indicating the environment in the first target area, and may output the second meta-information including the information indicating the environment in the second target area. By using the learning model according to this aspect, it becomes possible to obtain the information indicating the environment in the second target area, which is the environment around the observation object. Therefore, it becomes possible to observe the environment around the observation object with high precision or high accuracy. Also, in another aspect, the signal processing device includes a storage unit storing the learning model of the above aspect, a signal acquisition unit for acquiring the second received signal, and an estimation unit for inputting the second received signal to the learning model and estimating the second meta-information.
[0013] According to this aspect, it becomes possible to perform signal acquisition and estimation of the second meta-information using the learning model with a single signal processing device. Thereby, for example, even in an environment where there are certain restrictions on communication with the outside, such as the space above the earth, it becomes possible to estimate the meta-information using the learning model, and the signal processing device can observe the situation of the observation object with high precision or high accuracy.
[0014]
[0015] <In the above embodiment, the signal processing device includes an estimation unit which determines the second received signal at the first time and The second received signal at time 2 is input to the learning model in the above configuration, and at time 1 The second metadata at the first time point corresponding to the second received signal and the said at the second time point The second metadata at the second time point corresponding to the second received signal is estimated, and at the first time point... Based on the second metadata at the second time point and the second metadata at the second time point, in the second target area The system may further include a change detection unit that determines the change.
[0018] According to this embodiment, the second metadata of the second target region at the first time step is When a change occurs, the presence or absence of that change can be detected. Based on metadata, the signal The processing device observes the state of the object being observed or the environment surrounding the object being observed with high precision or accuracy. Furthermore, it is possible to determine changes in the second target area.
[0019] In the above embodiment, the signal processing device, when the determined change satisfies predetermined conditions, The system may further include a change information output unit that outputs change information indicating a change.
[0020] According to this embodiment, the signal processing device determines whether or not a change has occurred when a change satisfies predetermined conditions. In addition, the content of the changes is output as change information. This allows the details of the changes to be sent to an external device. Therefore, it becomes possible to acquire the information. The signal processing device changes the information as needed based on the conditions. By outputting information, the amount of communication between the signal processing device and the outside and the communication within the signal processing device are determined. This makes it possible to reduce unnecessary power consumption.
[0021] In another embodiment, the flying object includes a memory unit in which the learning model of the above embodiment is stored, and 2. A signal acquisition unit that acquires the received signal, and a learning model that receives the second received signal and the second metadata An estimation unit that estimates the information, and a signal output unit that outputs an output signal based on the second metadata to the outside, It is equipped with.
[0022] According to this embodiment, the flying object can independently estimate second metadata using a learning model. This makes it possible for flying objects placed in environments with certain restrictions on communication with the outside world to... This makes it possible to estimate the second metadata without external communication. The amount and power consumption required for communication can be reduced. In addition, the flying object will use the second metadata. Based on this, for example, output information including the second metadata itself and information indicating changes in the second metadata. The data can be output externally. This allows for the output of large data, such as SAR data, to external sources. Instead of transmitting, it sends a second set of metadata equivalent to the metadata observed based on the SAR data to an external source. This makes it possible to communicate. This will allow communication between the flying object and the outside world, and communication by the flying object. This can reduce the amount of electricity needed.
[0023] In another embodiment, the program is programmed on a computer to record the learning model of the above embodiment. A signal acquisition process that acquires the second received signal input to the stored memory unit, and a learning model that acquires the second The system receives two received signals and performs an estimation process to estimate the second metadata. This allows the system to perform the following: To enable a computer to observe metadata indicating the state of the object being observed with high accuracy or precision. This becomes possible.
[0024] In the above embodiment, the program sends an output signal based on the second metadata to the computer. Further processing of the signal output to be output externally may be performed. This allows, for example, the p In a flying object equipped with a computer on which a program is recorded, the amount of communication between the flying object and the outside and This can reduce the power consumption required for communication by flying objects. [Effects of the Invention]
[0025] According to the present invention, it is possible to appropriately observe the condition of the object being observed or the environment surrounding the object being observed. It is possible to provide a learning model, signal processing device, flying object, and program to enable this. . [Brief explanation of the drawing]
[0026] [Figure 1] This is a block diagram of the observation system according to this embodiment. [Figure 2] This figure illustrates the learning model according to this embodiment. [Figure 3] This diagram illustrates the learning process of the learning model according to this embodiment. [Figure 4] This figure illustrates an example of the information used to train the learning model according to this embodiment. [Figure 5] This figure illustrates an example of the information used to train the learning model according to this embodiment. [Figure 6] This diagram illustrates the correspondence between the training data used to train the learning model according to this embodiment. [Figure 7] This is a flowchart illustrating the processing in the flying object according to this embodiment. [Figure 8] This figure illustrates the estimation of metadata by the signal processing device according to this embodiment. [Figure 9] This figure illustrates an example of metadata estimated by the signal processing device according to this embodiment. [Figure 10] This diagram illustrates the determination of changes in metadata by the signal processing device according to this embodiment. [Figure 11] This is a flowchart illustrating the change detection process by the signal processing device according to this embodiment. [Figure 12] This diagram illustrates other aspects of learning and estimation of a learning model. [Figure 13] This is a block diagram showing other embodiments of the observation system. [Modes for carrying out the invention]
[0027] A preferred embodiment of the present invention will be described with reference to the attached drawings. Those bearing the same reference numeral have the same or similar configuration.
[0028] Figure 1 shows a block diagram of the observation system 10 according to this embodiment. 10 comprises a flying object 100 and an observation device 200. The flying object 100 is positioned in the space above the Earth. The observation device 200 is positioned on Earth. In this embodiment, the flying object 100 is positioned on the Earth's surface. The target area D is observed by radar, and the observation signal O processed by the flying object 100 is observed It is transmitted to the measuring device 200. The observation signal O is, for example, acquired by the flying object 100, as described later. This is the received signal or metadata corresponding to said received signal.
[0029] The flying object 100 is equipped with a communication antenna 101, a radar device 102, and a signal processing device 103. Prepare. The flying object 100 is an artificial satellite capable of acquiring and processing received signals, and is in outer space. It is positioned and orbits the Earth. Note that the flying object 100 may be a geostationary satellite. The flying object 100 is an aircraft, helicopter, or drone, positioned in the airspace above the Earth. Any device capable of doing so would suffice.
[0030] The communication antenna 101 communicates with external devices installed on Earth or in outer space when the flying object 100 is in contact with them. It is an antenna used for communication.
[0031] The radar device 102 uses electromagnetic waves EM1, such as microwaves, to target a specific area D on the Earth's surface. The electromagnetic wave EM1 was reflected by the object being observed in the target region D when the object was irradiated. This is a device for acquiring electromagnetic waves EM2. The radar device 102 is, for example, a synthetic aperture radar (S AR) The reflected electromagnetic wave EM2 is handled by the radar device 102 on the flying object 100. It is processed and recorded as a received signal (RAW data) based on electromagnetic wave fluctuations, which is possible. The received signal is recorded, for example, as the signal strength for each predetermined coordinate in the target area D. Device 102 transmits the observation signal O to the observation device 200 via the communication antenna 101.
[0032] The radar device 102 includes a processor for controlling the process of acquiring received signals and the control It includes a memory device that stores the necessary programs.
[0033] The signal processing device 103 processes the received signal acquired by the radar device 102. It is a signal processing device. The signal processing device 103 has a storage area such as memory, and stores in the storage area In a computer, the processor executes a program that has been created to perform a predetermined process. be.
[0034] The signal processing device 103 has a storage unit 104 and a control unit 105. The storage unit 104 is, for example, For example, semiconductor memory such as RAM, or optical discs. The storage unit 104 is a signal processing unit 1 It stores various types of information used in the processing in step 03.
[0035] The learning model 1041 is stored in the memory unit 104. The learning model 1041 is used to store the received signal This program is trained to take a number as input and output metadata corresponding to the received signal. Details of the metadata and learning model 1041 will be described later.
[0036] The control unit 105 performs signal processing in the signal processing device 103. Furthermore, the control unit 105 also performs signal processing in the signal processing device 103. The control unit 105 controls the transmission of the processing results of the received signal through the flying object 100. The unit consists of an acquisition unit 1051, an estimation unit 1052, a signal output unit 1053, a change determination unit 1054, and a change information unit. It has a notification output unit 1055.
[0037] The signal acquisition unit 1051 acquires the received signal from the radar device 102.
[0038] The estimation unit 1052 uses the received signal acquired by the signal acquisition unit 1051 to learn the model 10 Input to 41 to obtain metadata corresponding to the received signal.
[0039] The signal output unit 1053 transmits the metadata acquired by the estimation unit 1052 via the communication antenna 101. The information is output to the observation device 200 as an observation signal O. Furthermore, the signal output unit 1053 is meta Along with the information, the received signal corresponding to the metadata may be output as the observation signal O.
[0040] The change determination unit 1054 corresponds to the received signals acquired from the target area D at different times. Based on multiple metadata, the status of the observed object in the target area D or the observed object It determines changes in the environment surrounding an object.
[0041] The change information output unit 1055 outputs that the change determined by the change determination unit 1054 is determined to be a predetermined condition. If the conditions are met, change information indicating the change will be output. Change information includes, for example, metadata. Among these, information from which changed metadata has been extracted, and the areas within target area D where changes have occurred. This is information indicating the range. Processing by the change determination unit 1054 and the change information output unit 1055. I will explain that later.
[0042] The observation device 200 controls the observation of the target area D by the flying object 100. This device transmits to 100 and acquires the observation signal O from the flying object 100. Observation device 2 00 has a communication unit 201 which includes an antenna and a control unit that controls communication by the antenna. Information is transmitted and received with the projectile 100 via the communication unit 201.
[0043] The signal processing unit 202 processes the observation signal O from the flying object 100. Based on the observation signal O acquired from the flying object 100, for example, the observation results in the target region D The results are visualized using images.
[0044] The learning process of the learning model 1041 according to this embodiment will be explained with reference to Figures 2 to 6. ru.
[0045] Figure 2 is a diagram illustrating the learning and estimation of the learning model 1041. Model 1041 is trained using the training data LD as the training data. The training data LD contains a The received signal R0 (first received) is obtained by irradiating the target area (first target area) with electromagnetic waves. It includes a set of metadata MD0 (first metadata) corresponding to the signal (transmit signal) and the received signal R0. The training model 1041 is trained using the received signal R0 as input and metadata MD0 as output.
[0046] For mapping metadata MD0 to the received signal R0, the received signal R0 is SAR data It is converted to a U-type signal. The received signal R0 will not be converted to a U-type signal by the observation device 200 unless it undergoes a predetermined conversion process. The data does not become understandable information. The conversion process of SAR data is, for example, performed by observation device 200 This is done by performing analysis and visualization processing based on SAR data. The user will be able to understand the observation results based on the signal R0.
[0047] SAR data has multiple levels depending on the content of the transformation applied to the received signal R0. For AR data, for example, range compression and single-look AR are applied to the received signal R0. There is first-level SAR data that has undergone mass compression. The first-level SAR data is multi This is prime number information, and information on the amplitude and phase of the reflected electromagnetic wave EM2 in the target region D. It is included. By visualizing the first level of SAR data as a SAR image, This allows for understanding the contents of the received signal R0.
[0048] Other SAR data includes range compression and multi-look azimuth processing for the received signal R0. There is also second-level SAR data that has been compressed. This allows for the visualization of the received signal R0 using a SAR image in a geometrically corrected state.
[0049] Other SAR data include range compression and single-look azimuth reduction for the received signal R0. There is also third-level SAR data that has undergone compression and orthorectification. This allows us to obtain SAR images that can be overlaid with optical images.
[0050] As described above, the received signal R0 is converted into SAR data and visualized as a SAR image. This allows the user to understand the observation results in the target region D. This involves associating metadata MD0 with the received signal R0 as information that indicates the meaning of the observation results. Yes, it is possible. Alternatively, a learning model that associates SAR images with metadata can be used to perform the task. The data then associates metadata with the SAR image and then matches the received signal with the metadata. I'm willing to accommodate.
[0051] Metadata M0, for example, in observations of ships at sea, relates to the ship and sea conditions. It is information. Specifically, as ship information, it includes information corresponding to the items of location, overall length, and type. It is obtained from SAR images. In addition, phase information is obtained from SAR data that has complex number components. Using this, it becomes possible to obtain the speed of a ship. In addition, as information about sea conditions, The wind direction and speed above can be obtained based on SAR data. Wind direction and speed at sea This is based on the backscatter coefficient of SAR data and the actually measured wind direction and wind speed data. It is estimated based on the calculated correlation. The backscattering coefficient is the coefficient of scattering at the surface of the target region. This coefficient is based on the intensity of electromagnetic waves that return in the direction of irradiation from the irradiated electromagnetic waves.
[0052] The mapping of metadata MD0 to the received signal R0 requires additional information other than SAR data. It can be used. For example, the metadata MD0 for a ship may include AIS (Automa A system called an automatic identification system. This includes AIS information, such as location, ship name, flag, and overall length, obtained from the TEM. This is also acceptable. Based on the AIS information at the time the received signal R0 was acquired, metadata MD0 is generated. It may also be done. In addition, regarding sea conditions, wind direction and wind speed may be obtained from buoys placed on the ocean. , and weather information may also be used as metadata MD0.
[0053] Metadata MD0 has various items depending on the observed object. For example, for land-based objects such as cars and animals. Metadata for a moving object includes information about the object's movement trajectory. This is obtained from changes in SAR data due to changes in the interference of reflected electromagnetic waves caused by this. In this case, the metadata includes the movement trajectory based on the GPS device attached to the mobile object. That's fine.
[0054] If the area being observed is a flooded area during a disaster, the extent of the flooded area should be included in the SAR image as metadata. It may also be acquired as a low-brightness area. In this case, the metadata may include aircraft and artificial satellites. The extent of flooding may include optical images obtained by stars.
[0055] For use as metadata in crop management, the growth stage of the crop may also be used as metadata. The degree of growth of an object is determined by the backscattering coefficient of the SAR data and the actual observed degree of crop growth. It is estimated based on the correlation calculated based on the following. In this case, the metadata includes optical instruments The data may include spectral reflectance characteristics obtained through measurement, as well as information on the actual height of the crops measured.
[0056] In detecting buildings, information about newly constructed buildings may be used as metadata. Information regarding new buildings includes the backscatter coefficient of SAR data and the newly observed data. It is estimated based on correlations calculated from information about the building. In this case, light Building information detected from image data, or information about new buildings obtained from map data, may be used. Furthermore, even if information regarding the type of building is obtained based on maps, etc., and used as metadata, good.
[0057] In determining the correspondence between metadata and received signals, electromagnetic wave simulations are used. It is also possible to simulate the irradiation and reflection of electromagnetic waves in the simulation model. , it can generate received signals and SAR data. In this case, the simulation model The conditions in the log, such as the number of ships, the shape of the ships, and metadata such as the trajectory of the moving object, are used. This becomes metadata associated with the received signal and SAR data.
[0058] Based on the electromagnetic waves irradiated onto another target area (second target area) by the flying object 100, The received signal R1 (second received signal) acquired by the aircraft 100 is used by the learned model 1041. When input is received, the learning model 1041 generates metadata MD1 corresponding to the received signal R1. Output (second metadata).
[0059] The learning model 1041, in addition to the received signal R1, also receives additional information when the received signal R1 is acquired. The report may be input and metadata MD1 may be output. For example, in the example of the ship described above, AIS information and information from buoys placed at sea may be additional information. Here, learning Model 1041 uses SAR data, AIS information, and information from buoys placed at sea. Assume that the model was trained using metadata MD0 based on the additional information it contains. In this case, the trained model When estimating metadata MD1 using Ru1041, AIS information or buoys installed at sea Any of the information from is input to the learning model 1041 as additional information. By including additional information in addition to the SAR data in the input of 1041, the metadata is improved. The observation may be performed with higher accuracy or higher precision.
[0060] The relationship between the received signal R0 and metadata MD0 will be explained with reference to Figures 3 to 6. In terms of implementation, it enables the prediction of oceanographic conditions, which are the environment surrounding ships at sea. Let's explain using a learning model as an example.
[0061] In the example in Figure 3, based on the received signal R0 corresponding to the target region D, S in the target region D is calculated. AR image IG1 is generated. In SAR image IG1, ships S31, S32, S3 3 has been detected. In this case, as shown in Figure 4, the first ship information and Metadata MD01, which includes first oceanographic information about ocean conditions, is obtained from the SAR image IG1. Metadata MD01 is associated with the received signal R0. Note that the first ship information and the following The "Ship ID" field in the second ship information identifies the ship in the metadata MD01. This is arbitrary information used for that purpose.
[0062] Furthermore, when the received signal R0 corresponding to the target area D is acquired, the AIS in the target area D Based on the information, metadata MD02 is obtained as shown in Figure 5. Metadata MD0 Section 2 includes second-level ship information concerning vessels and second-level sea conditions information concerning sea conditions. The information and the second oceanographic information have different items than the first ship information and the second oceanographic information. The data information MD02 is associated with the received signal R0.
[0063] Based on AIS information, as shown in the correct image IG2 in Figure 3, ship S31 In addition to S33, the presence of the vessel S34 in the target area D is recorded in metadata MD02. Therefore, it is shown.
[0064] The metadata MD0, which is associated with the received signal R0 as learning data LD, is metadata It is prepared based on MD1 and metadata MD2. Metadata MD0 contains information about the ship, The third ship information is included based on the first ship information, so that it has the same items as the first ship information. Furthermore, the metadata MD0 contains items related to ocean conditions similar to those in the second ocean condition information. Therefore, the third oceanographic information is included based on the second oceanographic information. In this way, the model is used The metadata MD0 is based on metadata MD01, which is based on SAR images, and information from other devices. Meta information MD02 can be generated by combining it. The metadata of either report MD01 or metadata MD02 is used as metadata MD0. That's good too.
[0065] The learning model 1041 uses the training data LD prepared as described above, for example, Learned through common machine learning methods such as those using neural networks. The learning model 1041 may be configured as a single learning model, or multiple learning models may be configured. It may be configured as a learning model that combines different models.
[0066] The processing by the flying object 100 will be explained with reference to Figures 7 to 9.
[0067] In step S701 of Figure 7, the radar device 102 emits electromagnetic waves E to the target region D. Irradiate with M1. The irradiation timing is controlled by the observation device 200. This may also be at a predetermined timing in the flying object 100.
[0068] In step S702, the signal acquisition unit 1051 receives the reflection detected by the radar device 102. The received signal R1 based on the electromagnetic wave EM2 is acquired from the radar device 102.
[0069] In step S703, the estimation unit 1052 receives the received signal R1 and assigns it to the learning model 1041. Input. At this time, the estimation unit 1052 also considers the received signal R1 if the received signal R1 is Additional information obtained may also be input into the learning model 1041.
[0070] In step S704, the estimation unit 1052 calculates the metadata MD corresponding to the received signal R1. The value 1 is obtained from the learning model 1041.
[0071] In step S705, the signal output unit 1053 outputs a signal based on the metadata MD1. This is output to the observation device 200. The output signal based on the metadata MD1 is the total of the metadata MD1. It is a signal that transmits part of the metadata MD1. Alternatively, the output signal is a signal that transmits part of the metadata. Conversely, it may also be a signal that transmits information as a result of information processing.
[0072] Refer to Figure 8 to explain the estimation of metadata by the learning model 1041. Here, Figure A learning model 104 trained using training data LD having the correspondences shown in 6. Let's explain how to use option 1.
[0073] The received signal R1 can be converted into SAR data that shows the state indicated in the SAR image IG3. This is information. In the metadata generated based on the SAR image IG3, the vessel S81 to S8 This includes information related to point 3. On the other hand, the actual situation is as shown in the correct image IG4. This is assumed to be the case. In other words, information regarding the ship S84 is obtained by converting the received signal R1 into SAR data. That's missing.
[0074] In this case, the signal processing device 103 inputs the received signal R1 to the learning model 1041, as shown in Figure 9. Metadata MD1 as shown can be obtained. Metadata MD1 contains information on four ships. The information shown is included. In other words, the signal processing device 103 detects from the SAR data. This makes it possible to detect metadata that would otherwise be impossible to find.
[0075] By using the received signal and the learning model 1041, the effects of SAR data conversion can be reduced. Metadata can be estimated without any effort. This allows for, for example, SA (Synthetic Analysis) to be used, although this is not shown in the diagram. In R images, false detections can occur where objects that should not be detected are detected as false images. Even so, metadata can be appropriately estimated.
[0076] Refer to Figures 10 and 11 to explain the detection of changes in metadata by the flying object 100. ru.
[0077] Figure 10 illustrates the detection of changes in metadata related to a building. At a certain time T1( At time point 1, the buildings are positioned as shown in the layout image IG5. At another time point At T2 (second time step), the buildings are arranged in the same region as shown in the arrangement image IG6. The placement images IG5 and IG6 show, for example, a building being viewed from the air or space. These are optical or SAR images observed from [location]. The arrangement images IG5 and IG6 are predetermined It has regions A01 to A04. As shown in the arrangement image IG6, at time T2 Then, a new building, NB, is being constructed in area A03.
[0078] An example of how the above changes are detected using the learning model 1041 is explained. Here, the learning model Unit 1041, in response to the input of the received signal, provides building information, which is the number of units in each area of the building. Assume that it has been trained to output this information as data.
[0079] Figure 11 shows a flowchart of the processing performed by the flying object 100.
[0080] In step S1101, the radar device 102 transmits an electromagnetic signal to the target area at time T1. It emits waves.
[0081] In step S1102, the signal acquisition unit 1051 receives the signal detected by the radar device 102. The received signal R3 based on the emitted electromagnetic wave is acquired from the radar device 102. The received signal R3 is stored in the memory unit. It may be stored in 104.
[0082] In step S1103, the estimation unit 1052 uses the received signal R3 to learn the model 1041 Enter it here.
[0083] In step S1104, the estimation unit 1052 calculates metadata M corresponding to the received signal R3. D3 is obtained from the learning model 1041. The metadata MD3 is stored in the memory unit 104. That's good too.
[0084] In step S1105, the radar device 102 transmits an electromagnetic signal to the target area at time T2. It emits waves.
[0085] In step S1106, the signal acquisition unit 1051 receives the signal detected by the radar device 102. The received signal R4 based on the emitted electromagnetic wave is acquired from the radar device 102.
[0086] In step S1107, the estimation unit 1052 uses the received signal R4 to learn the model 1041 Enter it here.
[0087] In step S1108, the estimation unit 1052 calculates metadata M corresponding to the received signal R4. D4 is obtained from the learning model 1041.
[0088] In step S1109, the change determination unit 1054 determines the metadata MD3 and metadata MD Based on 4, the change in the target area is determined. In the case of Figure 10, the metadata MD04 is In area A03, the number of buildings has increased to four. Therefore, the change determination unit 1054 This indicates a change in the number of buildings.
[0089] In step S1110, the change information output unit 1055 outputs when the change satisfies a predetermined condition. This determines whether or not there is a change. Here, the predetermined conditions are the result of the determination of whether or not there is a change, and the specific details of the change. These are conditions related to volume. For example, in the example in Figure 10, the specific content of the change is in a certain area. The conditions include the presence of change, or a change in the number of buildings, whether increasing or decreasing, within one of these areas. This could be the case.
[0090] Here, we assume that an increase in the number of buildings occurs in one of the domains. At this time, the change information output unit 1055 determines that the change satisfies predetermined conditions.
[0091] If it is determined that the change satisfies the predetermined conditions, in step S1111, change information The output unit 1055 outputs information indicating the change as an output signal to the observation device 200. If certain conditions are not met, the process terminates. The output signal is, for example, metadata MD3 and This is the signal that transmits MD4. Alternatively, the output signal relates to region A03 where a change occurred. It is a signal that extracts metadata and transmits the extracted metadata. Or, if there is a change... Within the determined region A03, we need coordinate information that shows the parts where changes have occurred in more detail. It can also be a signal being transmitted.
[0092] The projectile 100, based on accurate metadata derived from the received signal, determines the target area Changes in this area can be detected. Therefore, the accuracy or precision of change detection is also improved.
[0093] The flying object 100 determines the change and, only for the parts that have changed, it retrieves the metadata itself and By transmitting information indicating the range of the change to the observation device 200, This reduces the amount of data transmitted. As a result, the flying object 100 can suppress power consumption. This is possible because the available power is limited by the environment of the space above the Earth. This would be advantageous for projectile 100.
[0094] Figure 12 is a diagram illustrating the learning and estimation of the learning model 1041A. The Dell 1041A uses the received signal R0 and metadata MD0 as learning data LD, in addition to the following: Includes SAR data (first generated signal) and additional information generated based on the received signal R0. By preparing the training data LD in this way, the training model 1041A receives the received signal R1 , SAR data (second generated signal) generated based on the received signal R1 and the received signal R1 This enables the estimation of metadata MD1 using corresponding additional information as input.
[0095] For example, additional information could include the environment in the target region when the received signal R0 was acquired. The weather information for the target area can be used as data. (Using the learning model 1041A) When estimating metadata MD1, the environment in the target region at the time the received signal R1 was acquired is shown. This information can be obtained from other devices and included in the input of the learning model 1041A. This allows for more accurate or precise estimation of the metadata MD1.
[0096] Figure 13 shows a block diagram illustrating another embodiment: observation system 10A. As shown, the signal processing device 103 is provided to be included in the observation device 200A. It is also possible that the control unit 1301 of the flying object 100A receives the data acquired by the radar device 102. The signal is transmitted to the observation device 200A. The signal processing device 103 of the observation device 200A processes the signal. It is also possible to perform the above-mentioned processing.
[0097] The embodiments described above are provided to facilitate understanding of the present invention and do not limit the present invention. It is not intended to be interpreted in that way. The elements and conditions of each embodiment are illustrative examples. It is not limited to those things, and can be changed as appropriate. Also, different configurations can be combined. It is possible to substitute or combine them in parts. [Explanation of Symbols]
[0098] 10... Observation system, 100... Projectile, 101... Communication antenna, 102... Radar equipment, 103...Signal processing unit, 104...Storage unit, 1041...Learning model, 105...Control unit, 10 51...Signal acquisition unit, 1052...Estimation unit, 1053...Signal output unit, 1054...Change detection unit, 1055... Change information output unit, 200... Observation device, 201... Communication unit, 202... Signal processing unit
Claims
1. learning is performed using teacher data in which a first synthetic aperture radar reception signal based on a reflected electromagnetic wave that is irradiated to a first target area and a first generated signal generated based on the first synthetic aperture radar reception signal are input, and first meta information corresponding to the first synthetic aperture radar reception signal and having predetermined items is output; A learning model that causes a computer to function in response to input of a second synthetic aperture radar reception signal based on a reflected electromagnetic wave that is irradiated to a second target area and a second generated signal generated based on the second synthetic aperture radar reception signal, to output second meta information that corresponds to the second synthetic aperture radar reception signal and has the specified items.
2. learning is performed using teacher data in which a first synthetic aperture radar reception signal based on a reflected electromagnetic wave that is irradiated onto a first target area and information indicating an environment in the first target area are input, and first meta information corresponding to the first synthetic aperture radar reception signal and having predetermined items is output; A learning model that causes a computer to function in response to input of a second synthetic aperture radar received signal based on a reflected electromagnetic wave that is irradiated onto a second target area and information indicating the environment in the second target area, to output second meta information that corresponds to the second synthetic aperture radar received signal and has the specified items.
3. The learning model according to claim 1 or 2, The learning model is learned using the teacher data that outputs the first meta information including information indicating the environment in the first target area, and causes the computer to function to output the second meta information including information indicating the environment in the second target area.
4. A storage unit in which the learning model according to any one of claims 1 to 3 is stored; A signal acquisition unit that acquires the second synthetic aperture radar received signal; and an estimation unit that inputs the second synthetic aperture radar received signal to the learning model and estimates the second meta information.
5. A storage unit in which the learning model according to any one of claims 1 to 3 is stored; A signal acquisition unit that acquires the second synthetic aperture radar received signal; an estimation unit that inputs the second synthetic aperture radar received signal to the learning model and estimates the second meta information; A flying object comprising: a signal output unit that outputs an output signal based on the second meta information to the outside.
6. On the computer, A signal acquisition process for acquiring the second synthetic aperture radar received signal input to a storage unit in which the learning model according to any one of claims 1 to 3 is stored; A program that executes an estimation process in which the second synthetic aperture radar received signal is input to the learning model and the second meta information is estimated.
7. The program according to claim 6, The computer includes: a signal output process for outputting an output signal based on the second meta information to an outside source.
