Information processing device, information processing method, and recording medium

Abstract

An information processing device according to the present invention partitions a plurality of points that are included in a radar image into a plurality of point groups that include a portion that overlaps at least one other point group. For each of the plurality of point groups, the information processing device creates a model for predicting a phase that precedes the wrapping of a first phase that is associated with meta-information for the plurality of points that constitute the point group. For each of a plurality of point pairs selected for each of the plurality of point groups, the information processing device uses a predicted phase difference that is a difference for the phase predicted using the model and a first phase difference that is a phase difference for the first phase to calculate a corrected first phase difference that is the result of the unwrapping of the first phase difference. For each of the plurality of points, the information processing device calculates a second phase difference that is a phase difference for a second phase that is a prediction value for a phase that is the result of the unwrapping of the first phase so as to reduce a residual between the corrected first phase difference and the second phase difference.

Description

Information processing device, information processing method, and recording medium

[0001] This invention relates to an information processing apparatus, an information processing method, and a recording medium.

[0002] SAR images are used to detect and measure minute changes occurring on the ground, such as ground subsidence and building tilt. SAR images are obtained by observing the backscattered signals of radar waves transmitted from a SAR (Synthetic Aperture Radar), and their pixel values ​​include complex numerical values ​​representing amplitude and phase. The phase included in SAR images may be folded (wrapped) into values ​​within 2π [rad], such as -π [rad] to π [rad]. Unwrapping is sometimes performed to unfold the phase and return it to its original phase with a wide range of values ​​in order to obtain the total amount of delay, distance, etc. from the wrapped phase.

[0003] For example, the signal processing device described in Patent Document 1 approximates the real and imaginary parts constituting a set of acquired signals with spline functions, and performs phase unwrapping using the polynomials of each interval of the obtained spline functions. In this phase unwrapping, a sequence of polynomials is calculated by applying the Euclidean algorithm to the polynomials of the spline functions. Then, the number of sign changes in the sequence obtained by arranging the values ​​of that polynomial sequence at the point where the phase of each interval is to be determined is examined, and the indeterminate part, which is an integer multiple of π, is determined based on the number of changes.

[0004] International Publication No. 2013 / 157299

[0005] Generally, for example, if the phase is proportional to the distance between the object and the SAR, and the object changes distance rapidly with respect to time, the complex signal obtained by observing that object will be observed as a high-frequency signal. In such cases, the spline function described in Patent Document 1 may not be able to accurately approximate the high-frequency signal, and the prediction accuracy of the unwrapped phase may decrease. Furthermore, if the intervals between the sample points approximated by the spline function are unequal, the spline function described in Patent Document 1 may not be able to accurately approximate such high-frequency signals, and the prediction accuracy of the unwrapped phase may decrease even further.

[0006] One of the challenges of this disclosure is to accurately predict the unwrapped phase from the wrapped phase.

[0007] The information processing device in this disclosure includes: a division means for dividing a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; a model creation means for creating a model for each of the plurality of point clouds for predicting the phase before a first phase associated with metadata of the plurality of points constituting the point cloud is wrapped; a pair selection means for selecting a plurality of point pairs from a plurality of points constituting the point cloud for each of the plurality of point clouds; a phase difference calculation means for each of the plurality of point pairs for calculating a corrected first phase difference obtained by unwrapping the first phase difference based on a predicted phase difference which is the phase difference predicted using the model of the point cloud to which the point pair belongs and a first phase difference which is the phase difference of the first phase; and a second phase calculation means for calculating the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference which is the phase difference of the second phase which is the predicted value of the phase obtained by unwrapping the first phase is small.

[0008] The information processing method in this disclosure involves one or more computers dividing a plurality of points included in at least one radar image into a plurality of point clouds that include portions that overlap with at least one other point cloud; creating a model for each of the plurality of point clouds to predict the phase before the first phase associated with the metadata of the plurality of points constituting the point cloud is wrapped; selecting a plurality of point pairs from the plurality of points constituting the point cloud for each of the plurality of point clouds; calculating a corrected first phase difference for each of the plurality of point pairs based on a predicted phase difference, which is the phase difference predicted using the model of the point cloud to which the point pair belongs, and a first phase difference, which is the phase difference of the first phase; and calculating the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference, which is the phase difference of the second phase, which is the predicted value of the phase after the first phase has been unwrapped, is small.

[0009] The recording medium in the present disclosure causes one or more computers to divide a plurality of points included in at least one radar image into a plurality of point groups including portions overlapping with at least one other point group, create a model for predicting a phase before wrapping for each of the plurality of point groups, where the phase before wrapping is associated with meta information of a plurality of points constituting the point group, select a plurality of point pairs from the plurality of points constituting each of the plurality of point groups, calculate a corrected first phase difference obtained by unwrapping the first phase difference for each of the plurality of point pairs based on a predicted phase difference, which is a difference in phase predicted using the model of the point group to which the point pair belongs, and a first phase difference, which is a difference in the first phase, and calculate the second phase such that a residual between the corrected first phase difference and a second phase difference, which is a difference in phase between the second phase, which is a predicted value of the phase obtained by unwrapping the first phase, and the plurality of points included in the at least one radar image, is reduced. A program for causing the above to be executed is recorded.

[0010] According to the present disclosure, it becomes possible to accurately predict an unwrapped phase from a wrapped phase.

[0011] It is a block diagram showing a configuration example of a first information processing apparatus according to the present disclosure. It is a flowchart showing an example of a processing operation of the first information processing apparatus according to the present disclosure. It is a block diagram showing a configuration example of a first information processing system according to the present disclosure. It is a block diagram showing a detailed configuration example of the first information processing apparatus according to the present disclosure. It is a flowchart showing a detailed processing operation example of the first information processing apparatus according to the present disclosure. It is a diagram showing an example of dividing a plurality of time-series points regarding the same observation point into a plurality of point groups. It is a diagram showing an example of a point pair selected for point group 1 according to the present disclosure. It is a diagram showing an example of a point pair selected for point group 2 according to the present disclosure. It is a diagram showing an example of a point pair selected for point group 3 according to the present disclosure. It is a block diagram showing a configuration example of a second information processing apparatus according to the present disclosure. It is a flowchart showing an example of a processing operation of the second information processing apparatus according to the present disclosure. It is a block diagram showing a physical configuration example of the information processing apparatus according to the present disclosure.

[0012] Hereinafter, in the present disclosure, the drawings are associated with one or more embodiments. Also, in all the drawings, the same components are denoted by the same reference numerals, and the description thereof will be omitted as appropriate.

[0013] [Embodiment 1] (Overview) As shown in FIG. 1, the information processing apparatus 100 includes a division unit 120, a model creation unit 130, a pair selection unit 140, a phase difference calculation unit 150, and a second phase calculation unit 160.

[0014] The division unit 120 divides a plurality of points included in at least one radar image into a plurality of point groups including a portion overlapping with at least one other point group.

[0015] The model creation unit 130 creates a model for predicting the phase before the first phase associated with the meta-information of the plurality of points constituting the point group is wrapped for each of the plurality of point groups.

[0016] The pair selection unit 140 selects a plurality of point pairs from the plurality of points constituting the point group for each of the plurality of point groups.

[0017] The phase difference calculation unit 150 calculates a corrected first phase difference obtained by unwrapping the first phase difference based on a predicted phase difference, which is a difference in phases predicted using the model of the point group to which the point pair belongs, and a first phase difference, which is a difference in the first phases, for each of the plurality of point pairs.

[0018] The second phase calculation unit 160 calculates the second phase so that the residual between the corrected first phase difference and a second phase difference, which is a difference in phases between the corrected first phase difference and a predicted value of the phase obtained by unwrapping the first phase, is minimized for the plurality of points included in at least one radar image.

[0019] According to this information processing apparatus 100, the second phase obtained by unwrapping the first phase of the plurality of points can be calculated so as to be consistent with the model for each of the plurality of point groups. Therefore, it becomes possible to accurately predict the unwrapped phase (second phase) from the wrapped phase (first phase).

[0020] The information processing apparatus 100 executes information processing as shown in the flowchart of FIG. 2.

[0021] The division unit 120 divides a plurality of points included in at least one radar image into a plurality of point clouds that include portions that overlap with at least one other point cloud (step S120).

[0022] The model creation unit 130 creates a model for each of the multiple point clouds to predict the phase before the first phase associated with the metadata of the multiple points constituting the point cloud is wrapped (step S130).

[0023] The pair selection unit 140 selects multiple pairs of points from multiple points that constitute each of the multiple point clouds (step S140).

[0024] The phase difference calculation unit 150 calculates a corrected first phase difference for each of the plurality of point pairs, based on the predicted phase difference, which is the phase difference predicted using a model of the point cloud to which the point pair belongs, and the first phase difference, which is the phase difference of the first phase (step S150).

[0025] The second phase calculation unit 160 calculates the second phase at multiple points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference, which is the phase difference of the second phase that is the predicted value of the phase obtained by unwrapping the first phase, becomes small (step S160).

[0026] This information processing method allows for the calculation of a second phase, which is obtained by unwrapping the first phase of multiple points, in a manner consistent with the model for each of the multiple point clouds. Therefore, it becomes possible to accurately predict the unwrapped phase (second phase) from the wrapped phase (first phase).

[0027] The following describes a detailed example of the information processing device 100.

[0028] (Detailed example) The information processing device 100 is provided, for example, in an observation system SYS for observing a target area based on the backscatter signal of radar waves. The observation system SYS comprises a flying object 10 and the information processing device 100, as shown in Figure 3, for example. Note that there may be multiple flying objects 10.

[0029] (Flying object 10) Flying object 10 irradiates the target area with radar waves and observes the backscatter signal. Flying object 10 transmits observation data based on the observed backscatter signal. Flying object 10 is, for example, a SAR (Synthetic Aperture Radar) satellite that irradiates the Earth's surface with radar waves and observes the backscatter signal. Note that flying object 10 is not limited to a SAR satellite, but may be, for example, a drone.

[0030] The observation data may include the observation period, observation location, and radar wave irradiation method. The observation period is information indicating the time of observation. The observation period may be, for example, a date, but its representation is not limited to a date; for example, it may be a date and time. The observation location is, for example, the position of the flying object 10 at the time of observation.

[0031] (About the information processing device 100) The information processing device 100 is a device for obtaining a second phase φ from a wrapped first phase θ for each of a plurality of points included in at least one radar image. The second phase φ is the phase obtained by unwrapping the first phase θ. Note that the second phase φ is a value calculated (estimated) from the first phase θ, as will be explained below, and may differ from the true value.

[0032] As shown in Figure 4, for example, the information processing device 100 includes a radar image acquisition unit 110, a division unit 120, a model creation unit 130, a pair selection unit 140, a phase difference calculation unit 150, and a second phase calculation unit 160.

[0033] The information processing device 100 performs information processing as shown in Figure 5, for example.

[0034] The radar image acquisition unit 110 acquires at least one radar image (step S110). Steps S120 to S160 described above are then performed.

[0035] (Regarding the radar image acquisition unit 110) As described above, the radar image acquisition unit 110 acquires at least one radar image.

[0036] For example, the radar image acquisition unit 110 acquires observation data transmitted from the flying object 10. The radar image acquisition unit 110 may acquire observation data via various relay devices. The radar image acquisition unit 110 may generate at least one radar image based on the acquired observation data.

[0037] A radar image is image information that includes, for example, complex numerical values ​​representing amplitude and phase in the pixel value of each pixel (point), and is an interferometric SAR image, but is not limited to this.

[0038] The phase included in the radar image is, for example, the wrapped phase, i.e., the first phase θ. The wrapped phase is the phase in which the original phase is expressed as a value within 2π [rad] based on the periodicity of the phase.

[0039] In the following explanation, we will use an example where the first phase, which is the wrapped phase, is represented in the range of -π[rad] to π[rad]. Note that the method of representing the value within 2π[rad] is not limited to -π[rad] to π[rad], but may also be, for example, 0[rad] to 2π[rad].

[0040] Metadata may be associated with the radar image. The metadata may include at least one of the following: observation time, observation location, angle of incidence of the radar wave, and pixel position in the image (e.g., row number and column number). The observation location is, for example, the position of the flying object 10. The angle of incidence is, for example, the angle at which the radar wave emitted from the flying object 10 entered the observation area.

[0041] The following explanation uses an example where at least one radar image is actually multiple radar images observed at different times.

[0042] The radar image acquisition unit 110 may, for example, generate multiple radar images at different observation times based on observation data observed over a long observation period. The radar image acquisition unit 110 may, for example, acquire multiple radar images at different observation times from an image storage unit (not shown) that stores the generated radar images. More specifically, for example, the multiple radar images may be time-series radar images relating to the same observation area, that is, time-series radar images based on observation data that observed the same observation area.

[0043] Here, "the same" means substantially the same, and may include cases where the difference is within a predetermined range due to errors, etc., and the same applies hereafter.

[0044] (Regarding the division unit 120) As described above, the division unit 120 divides a plurality of points included in at least one radar image into a plurality of point clouds that include portions that overlap with at least one other point cloud.

[0045] For example, suppose at least one radar image is actually multiple radar images that include the same observation area but are measured at different times. In this case, the division unit 120 divides, for example, multiple points in chronological order relating to the same observation point included in the multiple radar images into multiple point clouds. The multiple points may span multiple radar images. That is, the multiple points may be the number of points obtained by taking one point from each of the multiple radar images that have captured the same area, for example.

[0046] Figure 6 shows an example of dividing multiple time-series points related to the same observation point into multiple point clouds. The horizontal axis of the figure represents time corresponding to the observation period, and the vertical axis represents the phase (first phase) included in the radar images for the multiple points. The multiple points in the figure are time-series points related to the same observation point, for example, points corresponding to the same observation point included in multiple radar images in chronological order. The dotted line in the figure shows an example of the phase before it is wrapped into the first phase.

[0047] The example in the figure shows a case where multiple time-series points related to the same observation point are divided into three point groups: point group 1, point group 2, and point group 3. Specifically, point group 1 consists of points observed from observation time d1 to d5, point group 2 consists of points observed from observation time d3 to d7, and point group 3 consists of points observed from observation time d5 to d9.

[0048] In this example, point group 1 includes observation periods d3 to d5 as a portion that overlaps with point group 2. Point group 2 includes observation periods d3 to d5 as a portion that overlaps with point group 1, and includes observation periods d5 to d7 as a portion that overlaps with point group 3. Point group 3 includes observation periods d5 to d7 as a portion that overlaps with point group 2. In other words, in this example, multiple points d1 to d9 in chronological order are divided into multiple point groups such that each point group includes a portion that overlaps with each of the adjacent point groups in chronological order.

[0049] The number of points to be divided into is not limited to three; it could be two, four, or more. Furthermore, the overlapping portion is not limited to two or more points; it could be just one point.

[0050] (Regarding the model creation unit 130) The model creation unit 130 creates the model as described above.

[0051] The model is, for example, a model for predicting the phase before the first phase θ is wrapped to a value in the range of -π [rad] to π [rad]. As described above, the information processing device 100 obtains the second phase φ, which is obtained by unwrapping the first phase θ. Therefore, the value predicted using the model is not used as the second phase φ, which is ultimately calculated as a predicted value obtained by unwrapping the first phase θ.

[0052] A model is created, for example, for each of the multiple point clouds created in the division unit 120. Also, for example, each model is a model for predicting the phase before the first phase θ of the multiple points constituting the corresponding point cloud is overlapped.

[0053] The first phase θ is associated with metadata, for example, in radar images, as described above. For example, the first phase of a time series may be associated with the observation time included in the metadata. In this case, each model may be represented by a function that includes the observation time as a parameter.

[0054] When the phase predicted using the model is denoted as the predicted phase Φ, the following equation (1) can be given as an example of an equation representing the model.

[0055] Here, v in equation (1) represents, for example, the phase change rate. ｉ is a parameter representing the observation time i. In equation (1), a represents a constant. As shown in equation (1), the model is a function whose parameters are elements included in the metadata for multiple points that make up a point cloud. The model exemplified in equation (1) is a model that approximates the time-series change in the position (e.g., height) of an observation point, and assumes that the change in the position (displacement) of the observation point is linear. Note that the model is not limited to the example in equation (1), and may be expressed as, for example, an appropriate periodic function, a multidimensional function, etc.

[0056] The model creation unit 130 creates a model for predicting the phase (predicted phase Φ) in which the first phase θ is unwrapped, using the first phase θ of multiple points that make up the point cloud, for example, for each point cloud. In the example of equation (1) above, for example, a model corresponding to each point cloud is created by searching for v and a that minimize S, which is expressed by the following equation (2), for each point cloud.

[0057]

[0058] Here, θi in equation (2) is the first phase θ at observation time i. v, d in equation (2) ｉ , a is the same as in equation (1).

[0059] Function W is a wrapping function that wraps the phase to values ​​within 2π [rad], for example, W(α) = α + 2πn (where n is an integer satisfying -π < α + 2πn ≤ π). Function f is a function that decreases from -π [rad] to 0 [rad] and increases from 0 [rad] to π [rad].

[0060] By calculating the model parameters that minimize this function S for each point cloud, it is possible to create a model that minimizes the residual between the predicted phase Φ predicted using the model and the first phase θ across all points constituting the point cloud.

[0061] The model creation unit 130 may calculate the predicted phase Φ using the created model.

[0062] Note that the model, function S, function f, etc., are not limited to the examples given above.

[0063] The model represented by equation (1) above is a model for the case where the observation point is displaced at a constant velocity. This model is an example of a model based on physical knowledge regarding the displacement velocity of the observation point. The model may also use, for example, a model based on physical knowledge regarding thermal expansion, seasonal changes, and changes in viewpoint (e.g., the position of the flying object 10).

[0064] For example, an observation point may be displaced roughly linearly with respect to temperature changes due to thermal expansion. In a model relating to thermal expansion, for example, in equation (1), v, which corresponds to the displacement rate, is a quantity proportional to the expansion coefficient, and d, which corresponds to the observation time, is in equation (1). ｉ It may also be expressed as an equation in which is replaced with the temperature on the observation day.

[0065] For example, observation points may shift in a roughly one-year cycle due to seasonal changes. A model relating to seasonal changes is, for example, "Φ = v ｃｏｓ cos(2πt / 365)+v ｓｉｎ It can also be expressed as the formula sin(2πt / 365) + a.

[0066] For example, the observation point may be displaced in proportion to the angle of incidence and the height of the observation point due to a change in viewpoint (a change in the relative positional relationship with the flying object 10). In equation (1), v, which corresponds to the displacement velocity, is an example of a value that is proportional to the height of the observation point. Therefore, the model relating to the change in viewpoint is, for example, d, which corresponds to the observation time in equation (1). ｉ It may also be expressed as an equation in which is replaced with the angle of incidence on the observation day.

[0067] The model may be based on physical findings that combine one or more of the following: displacement velocity of the observation point, thermal expansion, seasonal changes, viewpoint changes, etc. The model may also be represented by an appropriate function, such as a polynomial other than a linear function.

[0068] The function f may be -cos(x), and in this case, due to its periodicity, the function W may not be used. The function S may be, for example, a weighted sum corresponding to the observation time i.

[0069] (Regarding the pair selection unit 140) As described above, the pair selection unit 140 selects multiple pairs of points from multiple points that constitute each of the multiple point clouds. The multiple pairs of points may be selected such that all of the multiple points that constitute the point cloud are included in at least one of the paired points.

[0070] Figures 7 to 9 show examples of point pairs selected for each of the point clouds 1 to 3.

[0071] Figure 7 shows an example in which seven pairs of points are selected from point group 1, which consists of multiple points corresponding to observation times d1 to d5. These seven pairs of points correspond to the combinations of observation times d1 and d2, observation times d2 and d3, observation times d3 and d4, observation times d4 and d5, observation times d1 and d3, observation times d2 and d4, and observation times d3 and d5.

[0072] Figure 8 shows an example in which seven pairs of points are selected from a point group 2, which consists of multiple points corresponding to observation times d3 to d7. These seven pairs of points correspond to the combinations of observation times d3 and d4, d4 and d5, d5 and d6, d6 and d7, d3 and d5, d4 and d6, and d5 and d7.

[0073] Figure 9 shows an example in which seven pairs of points are selected from a point group 3, which consists of multiple points corresponding to observation times d5 to d9. These seven pairs of points correspond to the combinations of observation times d5 and d6, d6 and d7, d7 and d8, d8 and d9, d5 and d7, d6 and d8, and d7 and d9.

[0074] The method for selecting point pairs is not limited to the example described above. For example, for a point group 1 consisting of multiple points corresponding to observation times d1 to d5, there are a total of 10 possible combinations of point pairs. The pair selection unit 140 may select one or more of these 10 possible point pairs, or it may select point pairs such that all of the multiple points corresponding to observation times d1 to d5 are included in at least one point pair.

[0075] (Regarding the phase difference calculation unit 150) As described above, the phase difference calculation unit 150 calculates the corrected first phase difference Δe for each of the plurality of point pairs using a model. The plurality of point pairs are, for example, the point pairs selected by the pair selection unit 140. The corrected first phase difference Δe is a value obtained by unwrapping the first phase difference Δθ, which is the phase difference of the first phase θ, using a model.

[0076] The corrected first phase difference Δe is represented by, for example, the following equations (3) and (4).

[0077]

[0078]

[0079] Here, θ in Equation (3) t and θ ｓ respectively represent the first phases θ of one and the other of the two points constituting the point pair e. θ t represents, for example, the first phase θ of the point corresponding to the observation time d ｔ of one of the two points constituting the point pair e. θ ｓ represents, for example, the first phase θ of the point corresponding to the observation time d ｓ of the other of the two points constituting the point pair e. That is, θ t - θ ｓ is the difference in the first phase between the point pairs and is the first phase difference Δθ.

[0080] D in Equation (3) ｅ represents the predicted phase difference D ｅ obtained by Equation (4). d in Equation (4) ｔ and d ｓ respectively represent the observation times d ｔ and the observation time d ｓ v and a in Equation (4) are the same as those in Equation (1) respectively and are values included in the model created by the model creation unit 130. The predicted phase difference D ｅ can be said to be the difference between the predicted phases Φt and Φs at the observation times dt and ds respectively.

[0081] By Equations (3) and (4), the corrected first phase difference Δe of the point pair e can be calculated. That is, the predicted phase difference D ｅBased on this, a corrected first phase difference, which is corrected to best match the predicted phase Φ among the unwrapped phases, is obtained for each point pair. Therefore, for each of the multiple point pairs included in at least one radar image, the corrected first phase difference Δe for each of the multiple point pairs can be calculated, for example, by using equations (3) and (4).

[0082] Furthermore, the formula for calculating the corrected first phase difference Δe (i.e., the first phase difference Δθ, which is the phase difference of the first phase θ) obtained by unwrapping it using the model is not limited to the example described above.

[0083] (Regarding the second phase calculation unit 160) As described above, the second phase calculation unit 160 calculates the second phase φ obtained by unwrapping the first phase θ.

[0084] The second phase φ can also be described as a predicted value of the phase obtained by unwrapping the first phase θ of each of the multiple points.

[0085] The second phase calculation unit 160 calculates the second phase φ such that the residual ε between the corrected first phase difference Δe and the second phase difference Δφ is small for the entire set of points included in at least one radar image.

[0086] The second phase difference Δφ is the phase difference of the second phase φ.

[0087] For example, suppose at least one radar image is a set of multiple radar images taken at different measurement times. In this case, the second phase calculation unit 160 may calculate the second phase φ such that the residual ε between the corrected first phase difference Δe and the second phase difference Δφ is small for the entire set of multiple points in the time series relating to the same observation point in the multiple radar images.

[0088] For example, the second phase calculation unit 160 calculates the second phase φ that minimizes the value of E, which is expressed by the following equation (5).

[0089]

[0090] Here, φ in equation (5) t and φ ｓ These represent the second phase φ of one of the two points that constitute the pair e, and the other point.

[0091] φt For example, observation time d is one of the two points that make up point pair e. ｔ This represents the second phase φ of the point corresponding to . ｓ For example, observation time d is the other point of the two points that make up point pair e. ｓ This represents the second phase φ of the point corresponding to .

[0092] φ t -φ ｓ This corresponds to the second phase difference Δφ between point pairs e. The second phase calculation unit 160 minimizes the value E obtained by, for example, equation (5) φ. t and φ ｓ The second phase φ may be calculated as the second phase. The second phase calculation unit 160 gradually decreases the value of E obtained by equation (5) φ t and φ ｓ By searching for the second phase φ that minimizes the value of E obtained by equation (5), t ､φ ｓ You may also calculate φ. t and φ s This may be set gradually. The second phase calculation unit 160 is φ t and φ ｓ As an initial value for searching for the predicted phase Φ, for example, the model created by the model creation unit 130 is used to calculate the predicted phase Φ. t , Φ ｓ You may also use this. Note that this initial value is the predicted phase Φ t , Φ ｓ It is not limited to that; predetermined values ​​may also be used.

[0093] Since Δe in equation (5) is the first corrected phase difference Δe, the function g is the residual ε(= |Δe - (φ) of each point group e). t -φ ｓ The variable contains ) | ).

[0094] The function g is, for example, a function defined for a value greater than or equal to 0, and is a monotonically increasing function. Using such equation (5), the second phase calculation unit 160 calculates the second phase φ that minimizes E. i This calculates the second phase φ such that the residual ε is small. ｉ The observation time d ｉ This is the second phase φ.

[0095] Equation (5) is an example of an evaluation formula for evaluating the residual ε. Thus, the second phase φ may be a phase calculated using the residual ε evaluation formula such that the residual ε of the entire set of point pairs e included in at least one radar image is small.

[0096] For example, as shown in Figures 7-9, suppose there are 21 selected point pairs for 9 points corresponding to observation times d1-d9. For example, the second phase calculation unit 160 uses equation (5) to calculate the value of function g for each of the 21 point pairs, and then calculates a second phase φ that reduces the sum E of the values ​​of the 21 point pairs. ｉ The second phase φ corresponding to each of the observation times d1 to d9 is calculated. １ ~φ ９ However, the first phase θ corresponds to each of the observation times d1 to d9. １ ~θ ９ It is calculated as the unwrapped value.

[0097] Furthermore, this is not the only method for calculating the second phase φ in a way that minimizes the residual ε.

[0098] The second phase calculation unit 160 may output a second phase φ in which the first phase θ has been unwrapped. This allows the second phase φ to be used, for example, for analyzing radar images. Such processing using the second phase φ, such as analysis functions, may be performed by, for example, the information processing device 110, or by an external device (not shown). The second phase calculation unit 160 may output information contained in the radar image, information associated with the radar image, etc., along with the second phase φ.

[0099] So far, we have described an example in which, when at least one radar image consists of multiple radar images taken at different observation times, the second phase φ is calculated for multiple points in chronological order related to the same observation point. The multiple radar images may include multiple points in chronological order for each of the multiple observation points, in which case the above processing may be performed for each observation point. This makes it possible, for example, to calculate the second phase φ, in which the first phase θ of the time series is unwrapped, for the entire common observation area included in multiple radar images taken at different observation times.

[0100] As can be seen from the above explanation, by correcting the second phase φ predicted using the model with the corrected first phase difference Δe, Δφ, the residuals that may occur in the second phase φ predicted from the model can be reduced. Therefore, it becomes possible to accurately predict the phase after the wrapped first phase θ is unwrapped as the second phase φ.

[0101] Furthermore, by replacing the time series in the above explanation with a spatial direction, the method of unwrapping the first phase θ of multiple points in a time series can be applied to unwrapping the first phase θ of multiple points contained in a single radar image (for example, points arranged one-dimensionally or two-dimensionally in the observation region). Therefore, the above unwrapping technique can be extended to multiple dimensions.

[0102] For example, in time-series analysis of interferometric SAR, the shape of a mountain can be understood by unwrapping the two-dimensional first phase θ. When the surface of a mountain is divided into small regions, it can be considered as a plane with a constant gradient, so a model with a certain constant gradient can be applied to each of the multiple point clouds contained in the overlapping regions. As a result, the mountain as a whole can be treated as a region composed of multiple planes with non-constant gradients, and a more accurately unwrapped phase φ can be obtained.

[0103] (Function and Effects) As described above, according to this embodiment, the information processing device 100 comprises a division unit 120, a model creation unit 130, a pair selection unit 140, a phase difference calculation unit 150, and a second phase calculation unit 160.

[0104] The division unit 120 divides a plurality of points included in at least one radar image into a plurality of point clouds, each containing a portion that overlaps with at least one other point cloud. For each of the plurality of point clouds, the model creation unit 130 creates a model for predicting the phase before the first phase θ, which is wrapped in association with the metadata of the plurality of points constituting the point cloud, is wrapped.

[0105] The pair selection unit 140 selects multiple point pairs from multiple points constituting each of the multiple point clouds. The phase difference calculation unit 150 calculates a corrected first phase difference Δe for each of the multiple point pairs, based on a predicted phase difference De, which is the phase difference predicted using a model of the point cloud to which the point pair belongs, and a first phase difference Δθ, which is the phase difference of the first phase θ, with the first phase difference Δθ unwrapped. The second phase calculation unit 160 calculates a second phase φ for multiple points included in at least one radar image, such that the residual between the corrected first phase difference Δe and the second phase difference Δφ, which is the phase difference of the second phase φ, which is the predicted value of the phase with the first phase unwrapped, is minimized.

[0106] This allows us to calculate the second phase φ, which is obtained by unwrapping the first phase of multiple points, in a way that is consistent with the model for each of the multiple point clouds. Therefore, it becomes possible to accurately predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ).

[0107] According to this embodiment, at least one radar image is a plurality of radar images observed at different times. The division unit 120 divides a plurality of points in chronological order relating to the same observation point included in the plurality of radar images into a plurality of point clouds.

[0108] This allows for accurate unwrapping of the first phase θ for multiple points in chronological order. Consequently, it becomes possible to accurately analyze temporal changes.

[0109] According to this embodiment, the pairs of points are selected such that all of the points constituting the point cloud are included in at least one of the pairs.

[0110] This allows the second phase to be calculated using the corrected first phase difference of each point in each point cloud. As a result, the second phase can be calculated based on more points than when using the corrected first phase difference of only some points in each point cloud. Therefore, it becomes possible to predict the phase (second phase φ) with high accuracy.

[0111] [Embodiment 2] In Embodiment 1, an example of unwrapping the first phase θ using points included in the radar image was described. In this embodiment, an example of unwrapping the first phase θ using representative points that represent the created model will be described. In this embodiment, in order to simplify the explanation, explanations that overlap with other embodiments will be omitted as appropriate.

[0112] As shown in Figure 10, the information processing device 200 includes a radar image acquisition unit 110, a division unit 120, and a model creation unit 130 similar to those in Embodiment 1, a representative point calculation unit 270, a pair selection unit 240, a phase difference calculation unit 150 similar to those in Embodiment 1, and a second phase calculation unit 260.

[0113] The information processing device 200 performs the information processing shown in Figure 11.

[0114] Steps S110 to S130 described above are executed.

[0115] The representative point calculation unit 270 calculates the predicted phase Φ of the representative point set for each of the multiple point clouds using a model (step S270).

[0116] The pair selection unit 240 selects multiple point pairs from multiple points that constitute each of the multiple point clouds, including a representative point, for each of the multiple point clouds (step S240).

[0117] Step S150 described above is performed.

[0118] The second phase calculation unit 260 calculates a second phase φ by correcting the second phase φ so that the residual between the corrected first phase difference Δe and the second phase difference Δφ, which is the phase difference obtained from the second phase φ, is small for multiple points included in at least one radar image (step S260).

[0119] (Regarding the representative point calculation unit 270) The representative point calculation unit 270 calculates the predicted phase of the representative point set for each of the multiple point clouds using a model.

[0120] A representative point is a point that represents the model created for each of the multiple point clouds. The representative point may be a hypothetical point other than the points included in the radar image.

[0121] For example, a model M created for a certain point group k. ｋ However, v ｋ ×d ｉ +a ｋ Let's assume that this is the case. ｋ and a ｋ This represents the values ​​determined for the point cloud k for each of v and a included in equation (1). ｉ is d in equation (1) ｉ Similarly, it is a parameter that represents the observation period i.

[0122] Representative points may, for example, be points corresponding to specific values ​​related to parameters included in the metadata. These specific values ​​may be, for example, the mean or median of the observation period i included in the point cloud k, or they may be predetermined constants (e.g., 0). Representative points corresponding to the mean or median of the observation period i are examples of representative points corresponding to phases with reduced noise due to smoothing. The predicted phase of a representative point is calculated, for example, by substituting specific values ​​into the parameters included in the model (observation period in equation (1) above).

[0123] (Regarding the pair selection unit 240) The pair selection unit 240 selects multiple pairs of points from multiple points that constitute the point cloud, including a representative point, for each of the multiple point clouds.

[0124] For example, the pair selection unit 240 may select multiple point pairs for each of the multiple point clouds such that at least one of the multiple point pairs includes the representative point. That is, at least one of the multiple point pairs may include the representative point.

[0125] For example, the pair selection unit 240 may select multiple point pairs for each of the multiple point clouds such that each of the multiple point pairs includes a representative point. That is, each of the multiple point pairs may include a representative point.

[0126] (Regarding the second phase calculation unit 260) The second phase calculation unit 260 calculates the second phase φ in the same manner as the second phase calculation unit 160 described in Embodiment 1, for example. For example, the second phase calculation unit 260 calculates the second phase φ for each of the multiple points including the representative point such that the residual ε between the corrected first phase difference Δe and the second phase difference Δφ is small for the entire set of multiple points including the representative point.

[0127] (Function and Effects) As described above, according to this embodiment, the information processing device 200 includes a representative point calculation unit 270 that calculates the predicted phase of a representative point set for each of a plurality of point clouds using a model.

[0128] This allows the second phase φ to be calculated using a representative point less susceptible to noise. Consequently, it becomes possible to predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ) with greater accuracy.

[0129] Furthermore, by using a representative point corresponding to, for example, the mean or median of the observation period i (a parameter), the phase with reduced noise due to smoothing can be used as the phase of the representative point. Consequently, it becomes possible to predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ) with even greater accuracy.

[0130] Furthermore, representative points can also be unwrapped simultaneously. If the representative point is one that has a noise-reduced phase due to the smoothing process described above, the unwrapped, smoothed phase can be obtained as the output. That is, for example, by outputting the result of the representative point, an unwrapped phase that has been smoothed and is less affected by noise can be obtained. As a result, it becomes possible to accurately obtain positional changes, etc.

[0131] According to this embodiment, each of the multiple pairs of points includes a representative point.

[0132] This allows us to calculate the second phase φ by including a representative point less susceptible to noise in each of multiple point pairs. Consequently, it becomes possible to predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ) with even greater accuracy.

[0133] [Embodiment 3] In Embodiment 1, equation (5) was given as an example of an evaluation formula used to calculate the second phase φ. In this embodiment, other examples of evaluation formulas will be described. In this embodiment, in order to simplify the explanation, explanations that overlap with other embodiments will be omitted as appropriate.

[0134] Instead of equation (5), for example, equation (6) below may be used as the evaluation formula for calculating the second phase φ.

[0135]

[0136] Here, w in equation (6) ｅ This represents a predetermined weight for each point group e.

[0137] In other words, the second phase φ calculated using equation (6) is calculated using an evaluation formula that is weighted according to the expected size of the residual ε. Residual ε (= |Δe - (φ) t -φ ｓ )|) is the difference between the first corrected phase difference Δe and the second phase difference Δφ.

[0138] Weight lol ｅ For example, the value of d can be small enough that the residual ε is likely to be large. t and d s As the distance from the model increases, the deviation from the model also increases, which tends to increase the residual, so the weight may be set to a small value. Alternatively, a small value may be set for the weight, as a small coherence indicates the amount of noise contained in the observed phase value, and the smaller the coherence, the larger the residual tends to be. Alternatively, based on the deviation between the model and the first phase calculated in the model creation unit 130, the larger the deviation, the smaller the value may be set. Note that the weight w ｅ This is not limited to this, and may be determined as appropriate.

[0139] (Effects) As described above, according to this embodiment, the second phase φ is calculated using an evaluation formula that is weighted according to the magnitude of the residual ε.

[0140] This allows for the calculation of the second phase φ in a way that minimizes the residuals by applying weighting according to the magnitude of the residuals. Consequently, it becomes possible to predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ) with greater accuracy.

[0141] According to this embodiment, weight w ｅ The larger the residual ε, the smaller the value of ε.

[0142] This reduces the impact of points where the model doesn't fit well, allowing for the calculation of the second phase φ. Consequently, it becomes possible to predict the unwrapped phase (second phase φ) from the wrapped phase (first phase θ) with greater accuracy.

[0143] (Example of physical configuration of information processing devices 100, 200) Physically, the information processing devices 100, 200 have, for example, a bus 1010, a processor 1020, a memory 1030, a storage device 1040, a network interface 1050, an input interface 1060, and an output interface 1070, as shown in Figure 12.

[0144] Bus 1010 is a data transmission path for the processor 1020, memory 1030, storage device 1040, network interface 1050, input interface 1060, and output interface 1070 to send and receive data to and from each other. However, the method of connecting the processor 1020 and the other components to each other is not limited to bus connection.

[0145] Processor 1020 is a processor implemented using components such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit).

[0146] Memory 1030 is a main memory device implemented as RAM (Random Access Memory), etc.

[0147] The storage device 1040 is an auxiliary storage device implemented as an HDD (Hard Disk Drive), SSD (Solid State Drive), memory card, or ROM (Read Only Memory). The storage device 1040 stores program modules for realizing the functions of the device equipped with it. The processor 1020 reads each of these program modules into the memory 1030 and executes them, thereby realizing the functions corresponding to those program modules.

[0148] The network interface 1050 is an interface for connecting a device equipped with it to a communication network.

[0149] The input interface 1060 is an interface for the user to input information. The input interface 1060 consists of, for example, a touch panel, a keyboard, a mouse, and the like.

[0150] The output interface 1070 is an interface for presenting information to the user. The output interface 1070 is composed of, for example, a liquid crystal panel, an organic EL (Electro-Luminescence) panel, and the like.

[0151] Thus, the functions of the information processing devices 100 and 200 can be realized by the collaborative execution of a software program by each physical component. Therefore, the present invention may be realized as a software program, or as a non-temporary storage medium on which the program is recorded. Furthermore, the information processing device may be physically composed of multiple devices (for example, computers).

[0152] Although the present disclosure has been described above with reference to embodiments, the present disclosure is not limited to the embodiments described above. Various modifications to the structure and details of the present disclosure can be made as can be understood by those skilled in the art within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0153] Furthermore, while the flowcharts used in the above description show multiple steps (processes) in sequence, the execution order of the steps performed in each embodiment is not limited to the order in which they are described. In each embodiment, the order of the illustrated steps can be changed to the extent that it does not impair the content.

[0154] Some or all of the above embodiments may also be described as follows, but are not limited to the following: 1. Information processing apparatus comprising: division means for dividing a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; model creation means for each of the plurality of point clouds to create a model for predicting the phase before a first phase associated with metadata of a plurality of points constituting the point cloud is wrapped; pair selection means for each of the plurality of point clouds to select a plurality of point pairs from a plurality of points constituting the point cloud; phase difference calculation means for each of the plurality of point pairs to calculate a corrected first phase difference in which the first phase difference is unwrapped, based on a predicted phase difference which is the phase difference predicted using the model of the point cloud to which the point pair belongs and a first phase difference which is the phase difference of the first phase; and second phase calculation means for the plurality of points included in at least one radar image to calculate the second phase such that the residual between the corrected first phase difference and the second phase difference which is the phase difference of the second phase which is the predicted value of the phase after the first phase has been unwrapped is small. 2. 1. The information processing device according to 1. The information processing device according to 1. The information processing device according to 2. The information processing device according to 1. The information processing device according to 2. The information processing device according to 1. The information processing device according to 2. The information processing device according to 1. The information processing device according to 3. The information processing device according to 1. The information processing device according to 3. The information processing device according to any one of 1. to 3. The information processing device further comprises a representative point calculation means for each of the multiple point group, which calculates the predicted phase of a representative point set for the point group using the model, and at least one of the multiple point group includes the representative point. 5. The information processing device according to 4. The information processing device according to 4. The information processing device according to any one of 1. to 5. The information processing device according to 1. to 5. The information processing device according to 6. The information processing device according to 6. The information processing device according to 6. The information processing device according to 6. The information processing device according to 1. The information processing device according to 1.8. An information processing method comprising: one or more computers dividing a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; creating a model for each of the plurality of point clouds to predict the phase before the first phase associated with the metadata of the plurality of points constituting the point cloud is wrapped; selecting a plurality of point pairs from the plurality of points constituting the point cloud for each of the plurality of point clouds; calculating a corrected first phase difference for each of the plurality of point pairs based on a predicted phase difference, which is the phase difference predicted using the model of the point cloud to which the point pair belongs, and a first phase difference, which is the phase difference of the first phase, by unwrapping the first phase difference; and calculating the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference, which is the phase difference of the second phase, which is the predicted value of the phase after unwrapping the first phase, is small. 9. The information processing method according to 8., wherein the at least one radar image is a plurality of radar images observed at different times, and the division means divides the plurality of points in chronological order relating to the same observation point included in the plurality of radar images into the plurality of point clouds. 10. The information processing method according to 8. or 9., wherein the plurality of point pairs are selected such that all of the plurality of points constituting the point cloud are included in at least one of the point pairs. 11. The information processing method according to any one of 8. to 10., further comprising calculating the predicted phase of a representative point set for each of the plurality of point clouds using the model, wherein at least one of the plurality of point pairs includes the representative point. 12. The information processing method according to 11., wherein each of the plurality of point pairs includes the representative point. 13. The information processing method according to any one of 8. to 12., wherein the second phase is calculated using an evaluation formula weighted according to the magnitude of the residual. 14. The weight is smaller the larger the residual, as described in 13. Information processing method.15. A program to cause one or more computers to perform the following: divide a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; create a model for each of the plurality of point clouds to predict the phase before the first phase associated with the metadata of the plurality of points constituting the point cloud is wrapped; select a plurality of point pairs from the plurality of points constituting the point cloud for each of the plurality of point clouds; calculate a corrected first phase difference for each of the plurality of point pairs based on a predicted phase difference, which is the phase difference predicted using the model of the point cloud to which the point pair belongs, and a first phase difference, which is the phase difference of the first phase, by unwrapping the first phase difference; and calculate the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference, which is the phase difference of the second phase, which is the predicted value of the phase after unwrapping the first phase, is small. 16. The program according to 15., wherein the at least one radar image is a plurality of radar images observed at different times, and the division means divides the plurality of points in chronological order relating to the same observation point included in the plurality of radar images into the plurality of point clouds. 17. The program according to 15. or 16., wherein the plurality of point pairs are selected such that all of the plurality of points constituting the point cloud are included in at least one of the point pairs. 18. Furthermore, for each of the plurality of point clouds, the program calculates the predicted phase of a representative point set for the point cloud using the model, and at least one of the plurality of point pairs includes the representative point. 19. The program according to 18., wherein each of the plurality of point pairs includes the representative point. 20. The program according to 15. to 19., wherein the second phase is calculated using an evaluation formula weighted according to the magnitude of the residual. 21. The weight is smaller the larger the residual. 20. The program described in 22. A recording medium on which the program described in any one of 22.15. to 22.21. is recorded.

[0155] This application claims priority based on Japanese Patent Application No. 2024-161017, filed on 18 September 2024, and incorporates all of its disclosures herein.

[0156] SYS Observation System 100, 200 Information Processing Unit 110 Radar Image Acquisition Unit 120 Segmentation Unit 130 Model Creation Unit 140, 240 Pair Selection Unit 150 Phase Difference Calculation Unit 160, 260 Second Phase Calculation Unit 270 Representative Point Calculation Unit

Claims

1. Information processing apparatus comprising: division means for dividing a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; model creation means for creating a model for each of the plurality of point clouds for predicting the phase before a first phase associated with metadata of the plurality of points constituting the point cloud is wrapped; pair selection means for selecting a plurality of point pairs from a plurality of points constituting the point cloud for each of the plurality of point clouds; phase difference calculation means for each of the plurality of point pairs for calculating a corrected first phase difference obtained by unwrapping the first phase difference based on a predicted phase difference which is the phase difference predicted using the model of the point cloud to which the point pair belongs and a first phase difference which is the phase difference of the first phase; and second phase calculation means for calculating the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference which is the phase difference of the second phase which is the predicted value of the phase obtained by unwrapping the first phase is small.

2. The information processing device according to claim 1, wherein the at least one radar image is a plurality of radar images observed at different times, and the division means divides the plurality of points in chronological order relating to the same observation point included in the plurality of radar images into the plurality of point groups.

3. The information processing apparatus according to claim 1 or 2, wherein the plurality of point pairs are selected such that all of the plurality of points constituting the point group are included in at least one of the point pairs.

4. The information processing apparatus according to any one of claims 1 to 3, further comprising representative point calculation means for each of the plurality of point clouds, which calculates the predicted phase of a representative point set for the point cloud using the model, wherein at least one of the plurality of point pairs includes the representative point.

5. The information processing apparatus according to claim 4, wherein each of the plurality of point pairs includes the representative point.

6. The information processing apparatus according to any one of claims 1 to 5, wherein the second phase is calculated using an evaluation formula weighted according to the magnitude of the residual.

7. The information processing apparatus according to claim 6, wherein the weight is smaller as the residual is larger.

8. An information processing method comprising: one or more computers dividing a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; creating a model for each of the plurality of point clouds to predict the phase before a first phase associated with metadata of the plurality of points constituting the point cloud is wrapped; selecting a plurality of point pairs from the plurality of points constituting the point cloud for each of the plurality of point clouds; calculating a corrected first phase difference for each of the plurality of point pairs based on a predicted phase difference which is the phase difference predicted using the model of the point cloud to which the point pair belongs, and a first phase difference which is the phase difference of the first phase, and calculating the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference which is the phase difference of the second phase which is the predicted value of the phase after the first phase has been unwrapped is small.

9. A recording medium on which a program is recorded for causing one or more computers to perform the following actions: divide a plurality of points included in at least one radar image into a plurality of point clouds including portions that overlap with at least one other point cloud; create a model for each of the plurality of point clouds to predict the phase before a first phase associated with metadata of the plurality of points constituting the point cloud is wrapped; select a plurality of point pairs from the plurality of points constituting the point cloud for each of the plurality of point clouds; calculate a corrected first phase difference for each of the plurality of point pairs based on a predicted phase difference which is the phase difference predicted using the model of the point cloud to which the point pair belongs, and a first phase difference which is the phase difference of the first phase, and calculate the second phase for the plurality of points included in at least one radar image such that the residual between the corrected first phase difference and the second phase difference which is the phase difference of the second phase which is the predicted value of the phase after the first phase has been unwrapped is small.

Images