Measurement parameter estimation device and its program

Abstract

To provide a measurement parameter estimation device and a program that can estimate a measurement parameter of an interference SAR in an emergency.SOLUTION: A parameter measurement estimation device 1 includes: measurement parameter estimation means 21 for estimating a measurement parameter; SARDSM calculation means 22 for calculating an SARDSM of a GCP candidate; histogram generation means 23 for generating a differential height histogram; fitting means 24 for fitting the differential height histogram; normal distribution determination means 25 for determining the number of normal distributions which fit the differential height histogram the most; false GCP candidate extraction means 26 for extracting a false GCP candidate; and control means 27 for commanding repeating processing.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to a measurement parameter estimation device and program for estimating measurement parameters of interferometric SAR. [Background technology]

[0002] Synthetic Aperture Radar (SAR) is used in remote sensing to measure the height of terrain, buildings, trees, and other structures. SAR can take measurements regardless of day or night or weather conditions, and is expected to be useful in emergencies such as natural disasters.

[0003] Currently, interferometric scanning (ASR) is the most widely used method for measuring altitude using SAR. In ASAR, observation signals (radar waves) are transmitted to the same object from multiple locations separated by only a short baseline length, and the height (and horizontal position) of the object is measured from the phase difference of the reflected observation signals. At this time, the position measurement accuracy of ASAR depends heavily on measurement parameters such as the baseline length, so it is important to estimate accurate measurement parameters.

[0004] The following describes conventional methods for estimating measurement parameters. In measurement parameter estimation methods, multiple GCPs (Ground Control Points) for comparison are selected from SAR image data, and the measurement parameters are inversely estimated by comparing the phase difference measured by interferometric SAR at each GCP with reference altitude information (Non-Patent Literature 1). For example, the reference altitude information can be the digital elevation model of the base map information. In addition, there is a known conventional technique that utilizes the fringe pattern that appears in phase interferometry images (images representing the phase difference of observed signals) (Non-Patent Literature 2).

[0005] Two perspectives are key when evaluating measurement parameter estimation methods. The first evaluation item is whether the entire process, including the estimation of measurement parameters and the generation of altitude maps (images representing the height of measured objects) from phase interferometry images, can be automated without human intervention. As mentioned above, SAR is often expected to be urgent, so this first evaluation item is important from the perspective of social benefit.

[0006] The second evaluation item is whether altitude information (ground model, DTM: Digital Terrain Model) that does not include the height of objects such as buildings and trees can be used as a reference for measuring measurement parameters. SAR measurement data is altitude information (digital surface model, DSM: Digital Surface Model) that includes the height of objects, and a typical application of this is measuring the height of objects. In this application, the difference between the altitude information shown in the SAR measurement data (hereinafter, "SARDSM") and the reference DTM is calculated. In this case, it is preferable that the reference used for estimating the measurement parameters and the DTM used for the difference processing are the same. If the two do not match, the possibility of an altitude difference between the two cannot be ruled out, and there is a risk of offset errors in the height of objects. Therefore, whether or not a DTM can be used as the reference altitude information is also an important evaluation item.

[0007] Therefore, the estimation method for measurement parameters requires that it be capable of automated processing and that DTM be used as reference altitude information. In order to realize such an estimation method, for SARDSM, for which altitude has not yet been determined, it is necessary to select a portion corresponding to the ground surface as the GCP. This is because if rooftops of buildings, which are not on the ground surface, are selected as GCPs and compared with the reference DTM, the rooftop and the ground surface will be treated as having the same height, leading to the estimation of incorrect measurement parameters.

[0008] Here, several conventional methods for extracting the portion corresponding to the ground surface from SAR DSMs have been proposed (Non-Patent Documents 3 to 5). In the method described in Non-Patent Document 3, the ground surface is extracted by combining SAR DSMs with satellite photo data and geospatial vector data. Further, in the method described in Non-Patent Document 4, a pre-learned Markov random field is used to remove buildings and the like, and the ground surface is extracted from the remaining area on the condition that the height is the lowest, etc. Further, in the method described in Non-Patent Document 5, the ground surface is extracted using a plurality of overlapping SAR DSMs.

Prior Art Documents

Non-Patent Documents

[0009]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

[0010] The methods described in Non-Patent Documents 3 and 5 may not be able to prepare the necessary data in emergencies such as natural disasters. Furthermore, while the method described in Non-Patent Document 4 has been demonstrated to be able to remove structures, it is unclear whether it can remove trees and other vegetation. Thus, no conventional technology has been proposed that can extract the portion corresponding to the ground surface from SARDSM and estimate the measurement parameters of interferometric SAR with high accuracy, even in emergencies.

[0011] Therefore, the object of the present invention is to provide a measurement parameter estimation device and program that can estimate the measurement parameters of interferometric SAR with high accuracy, even in emergency situations. [Means for solving the problem]

[0012] To solve the aforementioned problems, the measurement parameter estimation device according to the present invention is a measurement parameter estimation device that receives predetermined GCP candidates and reference DTMs as input and estimates the measurement parameters of interferometric SAR from the input GCP candidates and reference DTMs, and comprises a measurement parameter estimation means, a SARDSM calculation means, a histogram generation means, a fitting means, a normal distribution determination means, a false GCP candidate extraction means, and a control means.

[0013] In this configuration, the measurement parameter estimation means estimates the measurement parameters from GCP candidates using a convergence algorithm. The SARDSM calculation means calculates the SARDSM for the GCP candidate based on the measurement parameters estimated by the measurement parameter estimation means. The histogram generation means generates a differential height histogram representing the difference between the reference DTM and the SARDSM. The fitting means fits the difference height histogram generated by the histogram generation means to each of the normal distributions ranging from one to a predetermined number.

[0014] The normal distribution determination means determines, based on an information criterion, the number of normal distributions that best fit the difference height histogram from among the given number of normal distributions. The false GCP candidate extraction means extracts false GCP candidates that exceed a predetermined upper limit of the distribution in the number of normal distributions determined by the normal distribution determination means. The control means satisfies the pre-set termination conditions. The system determines whether or not the condition is met and continues using the measurement parameter estimation means, SARDSM calculation means, histogram generation means, fitting means, normal distribution determination means, and false GCP candidate extraction means until the termination condition is met. Command to repeat the process. The measurement parameter estimation means, SARDSM calculation means, histogram generation means, fitting means, normal distribution determination means, and false GCP candidate extraction means repeatedly execute processing until the termination condition is met. The measurement parameter estimation means excludes false GCP candidates from the GCP candidates when estimating the measurement parameters.

[0015] In this way, the measurement parameter estimation device can accurately estimate interferometric SAR measurement parameters even in emergency situations by extracting surface GCP candidates from SARDSM while excluding false GCP candidates other than those on the surface.

[0016] Furthermore, the present invention can also be implemented using a program that causes a computer to function as the aforementioned measurement parameter estimation device. [Effects of the Invention]

[0017] According to the present invention, measurement parameters of interferometric SAR can be estimated with high accuracy even in emergency situations. [Brief explanation of the drawing]

[0018] [Figure 1] This is an explanatory diagram illustrating interferometric SAR in an embodiment. [Figure 2] This is an explanatory diagram illustrating the error of interferometric SAR in an embodiment. [Figure 3] This is an explanatory diagram illustrating a GCP candidate in an embodiment. [Figure 4] In this embodiment, the graph shows an example of a differential height histogram. [Figure 5] This is a block diagram showing the configuration of a measurement parameter estimation device according to an embodiment. [Figure 6] Figure 5 is a flowchart showing the operation of the measurement parameter estimation device. [Figure 7]This graph shows an example of a differential height histogram in the Nth iteration of the embodiment. [Figure 8] In this embodiment, it is a graph illustrating the fitting of the differential height histogram. [Figure 9] This is a graph illustrating the extraction of false GCP candidates in the embodiment. [Figure 10] This graph shows an example of a differential height histogram in the N+1th iteration of the embodiment. [Figure 11] In the embodiment, (a) is an example image of a difference map obtained by radar measurement, (b) is an example image of a difference map obtained by a measurement parameter estimation device, and (c) is an example image of a difference map obtained by the prior art. [Modes for carrying out the invention]

[0019] Embodiments of the present invention will be described below with reference to the drawings. However, the embodiments described below are intended to embody the technical concept of the present invention, and unless otherwise specified, the present invention is not limited to these embodiments. In addition, the same reference numerals are used for the same means, and their descriptions may be omitted.

[0020] (Embodiment) [Principles of measurement parameter estimation] The following describes the principle of measurement parameter estimation, followed by a description of a measurement parameter estimation device according to an embodiment. Figure 1 illustrates an example of an interferometric SAR (SAR9) mounted on an aircraft. In an aircraft-mounted SAR, two antennas 90 are mounted on the underside of the aircraft 91. m ,90 s ) is used, and the measurement data corresponds to a DSM that includes the height of object 92.

[0021] Two antennas 90 (90 m ,90 s ) are separated by a predetermined baseline length B. Thereafter, one is connected to the master antenna 90 mLet one be the master antenna 90 and the other be the slave antenna 90 s Let it be so. The inclination angle of the line segment connecting the master antenna 90 m and the slave antenna 90 s with respect to the horizontal is defined as the baseline angle γ. Let the position of the object 92 be represented by P = P(x, y), P(x) be the horizontal position of the object 92, and P(z) be the height of the object 92. Let the inclination angle of the line segment connecting the master antenna 90 m and the position P with respect to the vertical be θ. Also, let the distance between the master antenna 90 m and the position P be s mp and the distance between the slave antenna 90 s and the position P be s sp Let it be so. Also, let the height from the ground surface to the master antenna 90 m be H.

[0022] In the example of FIG. 1, the distance S s between the slave antenna 90 sp and the position P is longer than the distance S m between the master antenna 90 mp and the position P by the distance difference Δs. Therefore, in an InSAR, as shown in the following equations (1) and (2), the distance difference Δ m , 90 s between the two antennas 90 S is measured as the phase coherence difference φ obs . In equation (1), φ offset represents the phase difference offset. Also, wrap is a function that converts the argument inside the parentheses to the range between 0 and 2π.

[0023]

Equation

[0024] Next, in an InSAR, the inclination angle θ is obtained from the phase coherence difference φ obs , and the position P(x, y) is measured as shown in the following equations (3) and (4). At this time, the measurement result of the position P(x, y) is quite sensitive to the errors of the measurement parameters, i.e., the baseline length B, the baseline angle γ, and the phase difference offset φ offset .

[0025]

number

[0026] In interferometric SAR, it is common practice to reference ground control points (GCPs) to estimate measurement parameters. Specifically, in interferometric SAR, as shown in equations (5) and (6) below, the tilt angle θ and distance s are estimated from the GGP. mp After calculating the above, the measurement parameters are estimated using equations (1) and (2).

[0027]

number

[0028] For example, the Geospatial Information Authority of Japan's DEM (Data Element) is available as a source of GCP (Ground Control Point) information and represents elevation information (DTM) that does not include the height of objects such as buildings and trees. On the other hand, SAR (Surface Alignment) measurement results represent elevation information (DSM) that includes the height of objects. Therefore, if an object's location is incorrectly selected as a GCP, it can cause measurement errors in the location P.

[0029] Specifically, in interferometric SAR, as shown in Figure 1, the master antenna 90 m Points P and P' at the same distance are measured superimposed on the same pixel. Typically, as shown in Figure 2, the reflection intensity of the observed signal is higher on the side P of the building than on the ground surface P'. As a result, the position of the side P of the building is mistakenly identified as the position of the ground surface P', causing the measurement result of position P to be shifted in the height and horizontal directions (illustrated by dashed lines).

[0030] Therefore, to suppress the measurement results of position P, it is considered that using GCPs on the ground surface would be effective. As shown in Figure 2, the height of the building's footprint (the point where the building touches the ground) F always coincides with the height of the ground surface P', so the object's footprint F is an optimal candidate for a GCP. In other words, the object's footprint F should be extracted as a candidate for a GCP (Countermeasure 1).

[0031] Prior experiments revealed that a certain percentage of the GCP candidates extracted in Countermeasure 1 included false GCP candidates that did not match the ground surface height. Therefore, the GCP candidates were refined using histogram fitting to exclude the false GCP candidates (Countermeasure 2).

[0032] In training GCP candidates, we will utilize the following two properties of SARDSM. As the first property, as shown in Figure 3, SARDSM indicates GCP candidate 93, which represents the height of the ground surface. A (93 A1 ~93 A5 ) and 93 GCP candidates that indicate the height of objects such as buildings and trees. B (93 B1 ~93 B4 It is composed of ). The second property is that GCP candidate 93 indicates the height of the ground surface. A This is GCP candidate 93, which indicates the height of the object. B It will be lower than this. Based on these properties, after the measurement parameters have been correctly estimated, the difference in height between SARDSM and the reference DTM will represent the difference altitude histogram shown in Figure 4. That is, in the difference altitude histogram of Figure 4, the GCP candidate 93 indicated by a circle A The smaller distribution 94 A Formed, GCP candidate 93 illustrated with a square. B The larger distribution 94 B This forms the following. Taking advantage of this, we can estimate the measurement parameters using GCP candidates and reference DTMs so that a differential elevation histogram like the one in Figure 4 is formed.

[0033] As shown in Figure 4, the differential altitude histogram has the differential altitude on the horizontal axis and the frequency (number of GCP candidates) on the vertical axis. Also, for the sake of clarity, Figure 4 shows 93 GCP candidates. A ,93 B This is illustrated.

[0034] [Configuration of the measurement parameter estimation device] Referring to Figure 5, the configuration of the measurement parameter estimation device 1 will be explained. The measurement parameter estimation device 1 estimates the measurement parameters of interferometric SAR from GCP candidates and reference DTMs. As shown in Figure 5, it comprises a GCP candidate extraction means 10, a GCP candidate training means 20, and an altitude map generation means 30. The measurement parameter estimation device 1 receives SARDSM data from the interferometric SAR 9 in a state where measurement parameter estimation is not yet complete, along with a reference DTM. The GCP candidate extraction means 10 corresponds to countermeasure 1, and the GCP candidate training means 20 corresponds to countermeasure 2.

[0035] The GCP candidate extraction means 10 extracts footprints, which are the contact points of objects, from SARDSM in a state where the estimation of measurement parameters has not yet been completed, as GCP candidates. For example, the GCP candidate extraction means 10 extracts footprints from SARDSM using the methods described in References 1 and 2. Specifically, the GCP candidate extraction means 10 extracts the lowest pixel position as the footprint in the region representing an object such as a building or tree included in the SAR image data. The GCP candidate extraction means 10 then outputs the extracted footprint as a GCP candidate to the GCP candidate training means 20.

[0036] Reference 1: Japanese Patent Publication No. 2016-090361 Reference 2: Uemoto, J., 2021. Refinement of interferometric SAR parameters using digital terrain model as an external reference. ISPRS J. Photogramm. Remote Sens. 175, 34-43.

[0037] The GCP candidate training means 20 trains GCP candidates and uses the trained GCP candidates to estimate measurement parameters. As shown in Figure 5, the GCP candidate training means 20 includes a measurement parameter estimation means 21, a SARDSM calculation means 22, a histogram generation means 23, a fitting means 24, a normal distribution determination means 25, a false GCP candidate extraction means 26, and a control means 27.

[0038] The measurement parameter estimation means 21 estimates measurement parameters from the GCP candidate extraction means 10 using a convergence algorithm. For example, the measurement parameter estimation means 21 estimates measurement parameters using the Levengerg-Marquard method as the convergence algorithm. In this case, the measurement parameter estimation means 21 excludes false GCP candidates from the GCP candidates when estimating measurement parameters.

[0039] The SARDSM calculation means 22 calculates the SARDSM for the GCP candidate based on the measurement parameters estimated by the measurement parameter estimation means 21. The method for calculating SARDSM is as described above, so its explanation is omitted (see Figure 1). The histogram generation means 23 generates a differential height histogram that represents the difference between the reference DTM and the SARDSM.

[0040] The fitting means 24 fits the differential height histogram generated by the histogram generation means 23 to each of the normal distributions ranging from one to a predetermined number. In this embodiment, the fitting means 24 fits the differential height histogram generated by the histogram generation means 23 to one normal distribution and two normal distributions. For example, the fitting means 24 fits the differential height histogram using a Gaussian mixture model method.

[0041] The normal distribution determination means 25 determines, based on an information criterion, the number of normal distributions that best fit the difference height histogram from among a given number of normal distributions. In this embodiment, the normal distribution determination means 25 determines the number of normal distributions that best fit the difference height histogram from among one or two normal distributions. For example, the normal distribution determination means 25 determines the number of normal distributions using the BIC (Bayesian information criterion) information criterion or the AIC (Akaike's information criterion) information criterion.

[0042] The false GCP candidate extraction means 26 extracts false GCP candidates that exceed a preset upper limit of the distribution in the number of normal distributions determined by the normal distribution determination means 25. In this embodiment, the false GCP candidate extraction means 26 presets μ+σ of the normal distribution that has the smallest μ among the normal distributions as the upper limit of the distribution.

[0043] The control means 27 repeatedly instructs each of the GCP candidate training means 20 to perform processing until a preset termination condition is met. In this embodiment, the control means 27 pre-sets the termination condition as the number of GCP candidates used for estimating measurement parameters not changing. Furthermore, when the termination condition is met, the control means 27 outputs the measurement parameters estimated by the measurement parameter estimation means 21 to the altitude map generation means 30. The details of each of the GCP candidate training means 20 will be explained in the operation of the measurement parameter estimation device 1.

[0044] The altitude map generation means 30 generates an altitude map using measurement parameters input from the GCP candidate training means 20. Specifically, the altitude map generation means 30 generates an altitude map from a phase interferometry image (phase difference map) using the measurement parameters. The method for generating the altitude map itself is known, so its explanation is omitted.

[0045] [Operation of the measurement parameter estimation device] Referring to Figure 6, the operation of the measurement parameter estimation device 1 will be explained. As shown in Figure 6, in step S1, the GCP candidate extraction means 10 extracts footprints, which are the contact points of an object, from SARDSM in a state where the estimation of measurement parameters has not yet been completed, as GCP candidates.

[0046] In step S2, the measurement parameter estimation means 21 uses a convergence algorithm to determine the measurement parameters (baseline length B, baseline angle γ, phase difference offset φ) from the GCP candidates extracted in step S1. offse We estimate ). In step S3, the SARDSM calculation means 22 calculates the SARDSM at the GCP candidate based on the measurement parameters estimated in step S2.

[0047] In step S4, the histogram generation means 23 generates a differential altitude histogram representing the difference between the reference DTM and SARDSM. Figure 7 shows an example of the differential altitude histogram in the Nth iteration. Also, Figure 7 shows the GCP candidate 93, which indicates the height of the ground surface, similar to Figure 4. A And, GCP candidate 93 indicating the height of the object. B Distribution 94 A ,94 B This was illustrated in the diagram.

[0048] In step S5, the fitting means 24 fits the difference height histogram generated in step S4 with one normal distribution and two normal distributions. Figure 8 shows an example of the result of fitting the difference height histogram of Figure 7. As shown in Figure 8, when fitted with one normal distribution 95, the peak of the normal distribution 95 is the same as that of the two normal distributions 94 A ,94 B It falls between (illustrated by a dashed line). Also, two normal distributions 96 A ,96 B When fitted to this (illustrated by the solid line), the normal distribution is 96 A 93 GCP candidates A Corresponding to the normal distribution 96 B 93 GCP candidatesA It corresponds to.

[0049] In step S6, the normal distribution determination means 25 determines the number of normal distributions that best fit the difference height histogram, from among one normal distribution or two normal distributions. In the example in Figure 8, the normal distribution determination means 25 determines that two normal distributions 96 are better than one normal distribution 95. A ,96 B Since this fits the difference height histogram, we determine that the number of normal distributions is 2.

[0050] In step S7, the false GCP candidate extraction means 26 extracts GCP candidates that exceed the upper limit of the distribution μ+σ in the normal distribution of the number of elements determined in step S6 as false GCP candidates. In the example in Figure 9, the false GCP candidate extraction means 26 extracts two normal distributions 96 A ,96 B Among them, the normal distribution with a small mean μ is 96 A μ+σ is set in advance as the upper limit of the distribution. Then, the false GCP candidate extraction means 26 selects GCP candidates whose difference elevation exceeds the upper limit of the distribution μ+σ, that is, GCP candidates 93 located to the right of the upper limit of the distribution μ+σ. A4 ,93 A5 ,93 B Extract it as a fake GCP candidate.

[0051] In step S8, the control means 27 determines whether the termination condition is met, which is that the number of GCP candidates used for estimating the measurement parameters does not change. If the termination condition is not met (N in step S8), the control means 27 returns to the process in step S2.

[0052] In the N+1th iteration, the measurement parameter estimation means 21 excludes the false GCP candidates extracted in the Nth iteration from the GCP candidates when estimating the measurement parameters (where N is an integer greater than or equal to 1). Subsequently, the measurement parameter estimation device 1 executes the processes from step S3 onward. Figure 10 shows an example of a differential elevation histogram in the N+1th iteration. In the differential elevation histogram of Figure 10, the position and shape of the distribution change compared to Figure 7 because the false GCP candidates are excluded when estimating the measurement parameters. Note that in the above-described iteration, when calculating SARDSM in step S3, it is not necessary to exclude the false GCP candidates.

[0053] If the termination condition is met (Y in step S8), the control means 27 proceeds to the process in step S9. In step S9, the altitude map generation means 30 generates an altitude map using the measurement parameters estimated in step S2.

[0054] [Effects / Effects] As described above, the measurement parameter estimation device 1 extracts ground surface GCP candidates from the SARDSM while excluding false GCP candidates other than those on the ground surface, thereby enabling highly accurate estimation of interferometric SAR measurement parameters even in emergency situations. As a result, a highly accurate SARDSM can be calculated using the measurement parameter estimation device 1.

[0055] Furthermore, the measurement parameter estimation device 1 does not require any external data other than the SARDSM and reference DTM in which the estimation of measurement parameters is not yet complete, making it usable in a variety of scenarios. In addition, since the measurement parameter estimation device 1 can automatically process up to the generation of the SARDSM, even operators who are not skilled in handling SAR data can generate the SARDSM, and a quick response is possible in emergencies. Moreover, since the measurement parameter estimation device 1 is not dependent on specific hardware, it is also highly convenient.

[0056] For example, the measurement parameter estimation device 1 can be used to acquire height data of objects such as buildings and trees in bulk. It can also be used to identify landslide locations and to estimate the amount of soil runoff. Furthermore, it can be used to detect areas where land may have been developed (including unregistered illegal development). Additionally, it can be used as a basis for determining whether or not to update high-precision DTMs, which have high costs per unit area. Finally, it can be used to improve the accuracy of past SARDSMs.

[0057] (modified version) Although embodiments have been described in detail above, the present invention is not limited to the embodiments described above, and also includes design modifications that do not depart from the spirit of the present invention.

[0058] In the embodiments described above, the measurement parameter estimation device was described as including a GCP candidate extraction means and executing countermeasure 1, but it is not limited to this. In other words, the measurement parameter estimation device does not need to include a GCP candidate extraction means and does not need to execute countermeasure 1. For example, when measuring the height of a residential area using interferometric SAR, the measurement parameter estimation device can estimate highly accurate measurement parameters by inputting GCP candidates and executing only countermeasure 2.

[0059] In the embodiments described above, it was assumed that there were two normal distributions, but there may be three or more normal distributions. For example, when low-rise buildings such as houses and high-rise buildings such as office buildings are mixed within the measurement range of interferometric SAR, using three normal distributions allows for high-precision fitting of the differential height histogram.

[0060] In the embodiments described above, the termination condition was explained as the condition that the number of GCP candidates used for estimating the measurement parameters does not change, but the invention is not limited to this. For example, the termination condition may be a condition that the process is repeated a predetermined number of times or for a predetermined period of time.

[0061] In the embodiments described above, the measurement parameter estimation device was described as an independent piece of hardware, but the present invention is not limited thereto. For example, the present invention can also be implemented by a program that causes hardware resources such as a CPU, memory, and hard disk of a computer to function as the measurement parameter estimation device described above. This program may be distributed via a communication line, or it may be written to a recording medium such as a CD-ROM or flash memory and distributed. [Examples]

[0062] Referring to Figure 11, the measurement accuracy of the measurement parameter estimation device will be explained as an example. Figure 11(a) shows the difference map between the reference DTM and DSM obtained by radar measurement, as the true value. The light areas (white) represent the exposed parts of the ground surface, and the dark areas (red) represent objects such as buildings and trees. Figure 11(b) shows the difference map between the SARSDM calculated by the measurement parameter estimation device and the reference DTM. The difference map in Figure 11(b) is similar to the true value in Figure 11(a), indicating that the measurement parameter estimation device can calculate a highly accurate SARDSM. Figure 11(c) shows the difference map between the SARSDM calculated by a conventional method and the reference DTM. In the difference map in Figure 11(c), the blue area at the bottom is more prominent compared to Figure 11(b), indicating that an error has occurred. [Explanation of Symbols]

[0063] 1. Measurement parameter estimation device 10 GCP candidate extraction method 20 GCP Candidate Training Methods 21 Measurement parameter estimation means 22 SARDSM Calculation Method 23 Histogram generation means 24 Fitting means 25 Normal distribution determination means 26 Fake GCP candidate extraction means 27 Control means 30 Advanced map generation means

Claims

1. A measurement parameter estimation device that receives predetermined GCP candidates and reference DTMs as input and estimates the measurement parameters of interference SAR from the input GCP candidates and reference DTMs, A measurement parameter estimation means for estimating the measurement parameters from the GCP candidates using a convergence algorithm, A SARDSM calculation means calculates the SARDSM in the GCP candidate based on the measurement parameters estimated by the measurement parameter estimation means, Histogram generation means for generating a differential height histogram representing the difference between the reference DTM and the SARDSM, A fitting means for fitting the difference height histogram generated by the histogram generation means to each of the normal distributions ranging from one to a predetermined number of elements, A normal distribution determination means that determines the number of normal distributions that best fit the difference height histogram among the various normal distributions based on an information criterion, A false GCP candidate extraction means extracts false GCP candidates from the number of normal distributions determined by the normal distribution determination means, which include a false GCP candidate extraction means that extracts GCP candidates that exceed a preset upper limit of the distribution, The system includes a control means that determines whether a pre-set termination condition is met and instructs the measurement parameter estimation means, the SARDSM calculation means, the histogram generation means, the fitting means, the normal distribution determination means, and the false GCP candidate extraction means to repeat processing until the termination condition is met. The measurement parameter estimation means, the SARDSM calculation means, the histogram generation means, the fitting means, the normal distribution determination means, and the false GCP candidate extraction means repeatedly perform processing until the termination condition is met. The measurement parameter estimation means is characterized by excluding false GCP candidates from the GCP candidates when estimating the measurement parameters.

2. The GCP candidate further comprises a GCP candidate extraction means for extracting footprints, which are the contact points of an object, from the SARDSM in a state where the estimation of the measurement parameters has not been completed. The measurement parameter estimation device according to claim 1, characterized in that the measurement parameter estimation means estimates the measurement parameters from the GCP candidates extracted by the GCP candidate extraction means.

3. The measurement parameter estimation device according to claim 1 or 2, characterized in that the false GCP candidate extraction means pre-sets the μ+σ of the normal distribution that has the smallest mean μ among the normal distributions as the upper limit of the distribution.

4. The fitting means fits the difference height histogram generated by the histogram generation means to one normal distribution and two normal distributions. The measurement parameter estimation device according to any one of claims 1 to 3, characterized in that the normal distribution determination means determines the number of normal distributions that best fit the difference height histogram from among the one normal distribution or the two normal distributions.

5. The measurement parameter estimation device according to any one of claims 1 to 4, characterized in that the control means pre-sets a condition as the termination condition that the number of GCP candidates used for estimating the measurement parameters does not change.

6. A program for causing a computer to function as a measurement parameter estimation device according to any one of claims 1 to 5.

Images