JP2016090361A - Extraction method of vertical structure from sar interferogram - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an extraction method of vertical structures from an interferogram in a rough shape, with loosened restriction on HOA (Height Of Ambiguity) without requiring elevation information.SOLUTION: A method for extracting vertical structures from output data of interferometric SAR (synthetic Aperture Radar) includes the steps of: 1) setting a vertical structure extraction filter with a size of a predetermined breadth and the length according to an offnadir angle in an observation in response to a position; 2), while scanning the vertical structure extraction filter on the interferometric image data, determining that layover/foreshortening and an observation phase difference are in a linear dependence relationship on image pixels in the vertical structure extraction filter at respective positions, by using a cosine transformation and sine transformation of the linear dependence relationship, and recording that the scanning positions have vertical structures; 3) grouping the vertical structures within a predetermined range as a series; 4) organizing them as a pat of vertical structure image; and 5) transforming it into the heights of the vertical structures.SELECTED DRAWING: Figure 6

Description

  The present invention relates to a method for automatically extracting a vertical structure from an SAR interferogram that can find a vertical structure from an interferogram (Interferogram) obtained by an interferometric synthetic radar (InSAR). .

  SAR (Synthetic Aperture Radar: Synthetic Aperture Radar) is a type of imaging radar that generates an image of the ground surface using electromagnetic waves. SAR has a higher spatial resolution than other imaging radars due to aperture synthesis technology using the phase of electromagnetic waves. SAR is mounted on a moving object (platform) such as an artificial satellite or an aircraft to observe the ground surface. The platform irradiates the microwave while moving, and measures the intensity and phase of the reflected wave backscattered on the ground surface.

  The amplitude and phase information of the received signal express the characteristics of the received wave, and these can be used to interfere with a plurality of data. InSAR has been proposed that uses this interference to measure terrain fluctuation and elevation. In InSAR, the altitude of the terrain is observed based on the phase difference calculated from the interference between two data in the same area observed using the SAR. Since the phase difference is used, the phase of the generated interference fringes has an indefiniteness of 2π. Even when the ground surface is flat, interference fringes called orbital fringes occur.

  Altitude information represented by DEM (Digital Elevation Model) can be obtained from optical observation, laser observation, etc., but extraction of altitude information using SAR observation data from the merit that can be observed regardless of day / night / weather Technology is regarded as important. Up to now, altitude information extraction based on SAR observation data has often been targeted for natural topographical altitudes with large spatial scale changes. Research and development of application methods to man-made structures in urban areas with small scales are becoming popular. The location of artificial structures such as power transmission towers, radio towers, and high-rise buildings and their altitude information or their changes are based on the safety flight of low-flying vehicles, as well as grasping the power infrastructure network, radio wave propagation obstacles, and city landscape Etc. are also important.

  As a method of obtaining altitude information from SAR observation, interferometry is often used. In this method, a phase difference image is generated by interfering observation data of two receiving antennas arranged perpendicular to the traveling direction. The phase difference (δφ ′) from which orbital fringes have been removed (pre-processed) and the height of the object to be observed (δh) have a relationship of δh = αIFδφ ′, so that altitude information can be obtained from the phase difference image. (Non-Patent Document 1). Here, αIF is a parameter uniquely determined by observation conditions such as a distance between receiving antennas.

  However, since the phase difference (δφ ′) obtained from interferometry inherently has an indefiniteness of 2π, in order to uniquely determine the altitude, a phase connection process that eliminates the indefiniteness ( A phase unwrapping process is inevitable. In areas such as urban areas where many artificial structures are mixed, automatic altitude information extraction by the interferometry method is often difficult because there are many phase discontinuities that are undesirable for the phase unwrapping process. (Non-patent document 2).

  Conventional techniques relating to the extraction of artificial structures derived from interferometry include (A) use of layover phenomenon, (B) use of linearity, (C) use of spectral shift, and (D) image shape recognition. .

(A) Utilization of layover phenomenon In SAR observation, the position of an object is measured by the time from radio wave irradiation to reception. In addition, SAR observation is performed from an oblique direction, not from directly above. Therefore, a high altitude object has a short radio wave round trip time between the radar and the object, resulting in a phenomenon that the object falls to the radar side. This phenomenon is called layover. A method of automatically extracting an artificial structure using this layover phenomenon has been reported by Non-Patent Document 3 and the like. In Non-Patent Document 3, an artificial structure is automatically extracted by utilizing the property that the occurrence of layover mixes a plurality of received signals in one pixel of an observed image.

(B) Use of linearity and artificial structure model (model knowledge) Unlike natural objects, artificial structures contain many linear components as constituent components. In addition, since the corner reflector structure that efficiently reflects radio waves toward the antenna is formed by the artificial structure perpendicular to the ground, a strong bright line is generated at the foot of the artificial structure. A method for automatically extracting an artificial structure by using these linear components and the like and repeatedly combining a model in which the shape of the artificial structure is represented by a set of rectangles and observation data is described in Non-Patent Document 4 etc. It is reported by. Similarly, Patent Document 1 describes a technique for extracting and removing a feature based on the linearity of the feature and its strength value.

(C) Use of spectral shift When interferometry observation is performed on a slope having an inclination angle β in the radar irradiation direction, a spectral shift depending on the inclination angle β occurs, and the coherence (coherence) depends on the shift amount. A decrease occurs. A technique for extracting an artificial structure by obtaining the inclination angle β of a target object from a phase difference image using the property of reducing the coherence has been reported in Non-Patent Document 5 and the like. Note that Non-Patent Document 6 reports a technique combining Non-Patent Document 3 and this technique.

(D) Image shape recognition Non-patent document 7 reports a method of automatically extracting power transmission towers using an image shape recognition technique from interferometry observation results converted into altitude values in advance.

  In addition, as a method for extracting high-level information without using interferometry, a method of formulating the amount of collapse as in Patent Document 2 has been reported.

JP 2008-164481 A JP 2008-185375 A

Richards M. A., A beginner ’s guide to interferometric SAR concepts and signal processing, IEEE Aerosp. Electron. Syst. Mag., Vol. 22, no. 9, pp.5 -29, 2007. Tison C., Tupin F., Nicolas J.M., Maitre H., Validation of a feature fusion scheme for urban DSM retrieval from high resolution SAR interferogram, IEEE International Geoscience and Remote Sensing Symposium Proceedings, 2005. Rossi C., Fritz T., Eineder M., Detecting Building Layovers in a SAR Interferometric Processor Without External References, EUSAR proceedings, 2014. Soergel U., Thoennessen U., Stilla U., RECONSTRUCTION OF BUILDINGS FROM INTERFEROMETRIC SAR DATA OF BUILT-UP AREAS, ISPRS Archives, Vol. XXXIV, Part 3 / W8, Munich, 17.-19. Sept. 2003. Petit D., Adragna F., Durou J.D., The filtering of layover areas in high-resolution IFSAR for the building extraction, Proc.SPIE 4173, SAR Image Analysis, Modeling, and Techniques III, 230, 2000. Rossi C., Eineder M., Duque S., Fritz T., Parizzi A., PRINCIPAL SLOPE ESTIMATION AT SAR BUILDING LAYOVERS, IGARSS proceedings, 2014. Woods D., Folley C., Kwan Y.T., Houshmand B., Automatic Extraction of Vertical Obstruction Information from Interferometric SAR Elevation Data, IGARSS proceedings, 2004. Uemoto Junpei, Kobayashi Tatsuharu, Satake Makoto, Kojima Shoichiro, Umehara Toshihiko, Matsuoka Kenshi, Uratsuka Kiyomine, Automatic Extraction Method of Vertical Structures from SAR Interferogram, The 56th Annual Conference of the Remote Sensing Society of Japan, 2014 .

  As described above, several automatic extraction methods for artificial structures have been reported, but each has advantages and disadvantages. First, in the above methods (A) and (C), the height difference of the artificial structure to be extracted is a height difference that produces a phase difference of 2π, which is a parameter of the distance between the two antennas. ), The use scene is limited in the case of the SAR observation apparatus of Non-Patent Document 3 (HOA = about 50 m). Non-Patent Document 3 states that H ≦ HOA / 2, and Non-Patent Document 5 states that H ≧ 2HOA. In Non-Patent Document 4 of (B), since an altitude map is generated and used from interferometry data in the process of extracting an artificial structure, an unwrapping process is required although not explicitly stated. Such a high artificial structure may not be applicable. Similarly, in Patent Document 1, an altitude map is assumed. The artificial structure that can be extracted depends on the shape of the structure model assumed in the calculation process. In (D), altitude information that is difficult to convert from a phase difference is required in advance in areas such as urban areas where many artificial structures are mixed. Further, it is stated that the technique of Patent Document 2 is not a technique focusing on an artificial object, and further requires a GCP (Ground Control Point) representing the shape of the ground for the amount of fall / altitude conversion.

  Therefore, in the present invention, the minimum detection specified by the user from the interferogram is not restricted except that the restriction on the HOA is relaxed, altitude information is not required in advance, and the shape to be extracted is not restricted except that it is vertical. A new algorithm was developed to extract vertical structures higher than altitude.

The extraction method of the vertical structure from the SAR interferogram of the present invention is as follows:
A method of extracting a vertical structure from output interference image data or interferometry data of an interferometric synthetic aperture radar,
Setting the vertical structure extraction filter according to the slant range position so as to have a predetermined lateral width and a vertical length that is large on the front side of the slant range and small on the far side according to the cosine of the off-nadir angle at the time of observation;
While scanning the vertical structure extraction filter on the output interference image data or interferometry data, the layover collapse amount and the observation phase difference are linearly dependent on the image pixels in the vertical structure extraction filter at each position. Determining that there is a cosine transformation of the linear dependency relationship and a sine transformation of the linear dependency relationship, and recording that the scan position is a pixel of a vertical structure image candidate;
Grouping vertical structure image candidate pixels within a predetermined range as a series of vertical structures to form vertical structure image candidates;
Arranging the vertical structure image candidates as vertically existing vertical images;
Obtaining a height of the vertical structure from the vertical structure image obtained by organizing;
It is characterized by including.

  In the present invention, the determination using the cosine transformation and sine transformation of the linear dependency relationship is performed by the standard deviation of the cosine function and the standard deviation of the sine function of the linear dependency relationship between the layover collapse amount and the observed phase difference. It is characterized in that a threshold value determination is performed for.

  In the present invention, the vertical structure image pixels corresponding to the series of vertical structures by the grouping are unwrapped in phase as the series of vertical structure image pixels are continuous. .

  Further, the present invention is characterized in that, in the vertical structure image candidate, each image pixel point showing the maximum or minimum layover collapse amount is set to the highest altitude point or the minimum altitude point of the vertical structure.

  Further, the present invention uses an intensity distribution of the output interference image data or interferometry data of the interferometric synthetic aperture radar applied to the intensity distribution of the intensity distribution using the determination of the linear dependence relationship. It is effective that the determination of an image pixel whose intensity exceeds a predetermined criterion.

  Further, the present invention is characterized in that a median filter obtained by applying a median filter to the phase distribution of the output interference image data or interferometry data of the interference synthetic aperture radar is used for the determination of the linear dependence relationship. .

  By using the present invention, the minimum restriction specified by the user from the interferogram is not restricted, except that the restriction on the HOA is relaxed, altitude information is not required in advance, and the shape to be extracted is not restricted except that it is vertical. It is possible to extract a vertical structure whose height is higher than the detection altitude.

It is a conceptual diagram of single-pass interferometry observation of a vertical structure. It is a figure which shows generation | occurrence | production of the discontinuity of the observation phase difference on an interferogram. A discontinuity occurs in the phase difference on the interferogram at the point of θ = 0. It is a figure which shows the example of application (minimum detection height 31m) of this invention. It is a figure which shows the 2nd application example (minimum detection altitude 31m) of this invention, (a) is an interferogram, (b) is about 140m in height on the mountain of about 80m as a result of application of this invention. An example of extracting a TV tower surrounded by white squares. It is a figure which shows the example of a flowchart of an interferogram production | generation. It is a figure which shows the example of a flowchart of application 1 of a vertical structure extraction filter. It is a figure which shows the example of a flowchart of grouping. It is a figure which shows the example of a flowchart of application 2 of a vertical structure extraction filter.

  Embodiments of the present invention will be described below in detail with reference to the drawings. In the embodiment described below, a vertical structure extraction filter that uses the relationship between the amount of collapse of the observation target in the slant range direction due to layover and its phase change is formed and applied to the interferogram. The application means to the filter and interferometry, and application examples are shown below.

<Vertical structure extraction filter>
FIG. 1 shows a conceptual diagram of single-pass interferometry observation in a vertical structure. In this example, the radio waves radiated from the antenna (Ant) 1 are reflected by each, and the reflected radio waves are received by Ant 2 and Ant 1. In this case, the preprocessed observation phase difference δφ observed by the interferometry is represented by a value on a two-dimensional plane between the azimuth AZ coordinate axis and the slant range R coordinate axis, and the phase difference δφ associated with the altitude δh of the vertical structure. The sum of unrelated phase differences α is expressed as shown in Equation 1 regardless of 'and altitude δh. Here, the phase difference α is typically a phase difference associated with the background topography height at the foot of the vertical structure, and is a constant value independent of the height of the vertical structure. Note that the phase difference at point Q in FIG. 1 is also reflected in the observed phase difference due to the occurrence of layover, but it is ignored because the reflection intensity at point Q is weaker than that of the vertical structure.

  From Non-Patent Document 1, the phase difference δφ ′ is expressed by the following formula 2.

  On the other hand, from FIG. 1, the altitude δh of the vertical structure and the amount of tilt δR of the observation object due to the layover are expressed by the following formula 3.

  By removing δh from Equations 2 and 3, the following Equation 4 is obtained.

Regarding Equation 4, Non-Patent Document 5 reports a generalized equation of the inclination of the observation object. The following equation 5 is derived from equations 1 and 4. This equation shows that δR and δφ have a linear relationship.

  Considering a vertical structure having a height H, the amount of layover collapse δR corresponding to the height H is uniquely determined from Equation 3. Assuming that the vertical structure is continuously present, the observed phase difference δφ satisfies Equation 5 in this retraction range. Therefore, the vertical structure on the interferogram can be automatically extracted by searching the interferogram continuously for the region of the observation phase difference δφ that satisfies Equation 5. However, if this expression is applied to an arbitrary point on the interferometry image as it is, the 2π indefiniteness of the observation phase difference δφ causes discontinuity at the switching point of the period having the period [0, 2π], for example. As with the prior art, it is subject to HOA restrictions. Therefore, instead of Equation 5, the sine (Equation 6) and cosine (Equation 7) are applied as a vertical structure extraction filter. By considering sine and cosine, continuity is maintained even at a switching point of a cycle having a period of [0, 2π], for example (FIG. 2).

<Application procedure to interferogram>
FIG. 3 shows the flow of the application procedure. The application procedure has four stages A to D. Before the application is performed, the user designates the minimum detection altitude H as an input parameter in advance. In the application example of FIG. 3, the minimum detection altitude is 31 m, which is a threshold value for high-rise buildings of the Fire Service Act. Note that the threshold appearing in the following procedure is a variable threshold because it is considered to depend on the observation specifications of the SAR sensor to be used. Also, in FIG. 4, a TV tower surrounded by a white square having a height of about 140 m on a mountain having an altitude of about 80 m, which is a vertical structure of (b), extracted from the interferogram of (a) as an application result of the present invention. An example is shown.

FIG. 5 shows a flow of procedure A. This procedure is often used in the production of SAR interferogram images. It is also easy to output as interference image data or interferometry data.
[Procedure A]
[A01] The data of the receiving antenna 1 and [A02] the receiving antenna 2 are read, and [A03] registration (correction of misalignment between data) is performed. [A04] [A05] The respective signal intensity average values are obtained, and [A06] the average intensity value of the receiving antenna 2 is matched with the average intensity value of the receiving antenna 1 (that is, normalization is performed with the average intensity value). Thereafter, [A07] plane phase difference unrelated to the topographic altitude is removed (preprocess), and [A08] an interferogram image (preprocessed phase difference image) is generated.

FIG. 6 shows a flow of the procedure B. This procedure B relates to a vertical structure extraction filter.
[Procedure B]
[B01] For all pixels on the image, the signal intensity value is converted to 0 to 1 (that is, the signal intensity is normalized), and [B02] cosine transformation and sine transformation are performed on the observed phase difference δφ. Next, the median filter is applied to each of the pixel arrays of [B03] signal intensity value, [B04] sine phase difference (sine value of observed phase difference δφ), and cosine phase difference (cosine value of observed phase difference δφ). Here, the signal intensity value is used in the following B06, B07, B08, B10, and B11. Thereafter, [B05] the vertical width Ws of the application region (search window) of the vertical structure extraction filter corresponding to the predetermined minimum detection height H is calculated from Equation 3, and the search window size is, for example, vertical × horizontal = Ws. X3.
[B06] In the above Ws × 3 range, for each pixel whose signal intensity value is a threshold TA (for example, 0.8) or more, the average value Af of each of the sine phase difference and the cosine phase difference (that is, Af, cos, Af, sin), and within the range of [B07] Ws × 3, for the pixel whose signal intensity value is equal to or greater than the threshold TA, the observation phase difference δφ and the search window for the observation object An average value A (that is, Acos, Asin) of the values on the left side of Equations 6 and 7 is calculated from the vertical pixel distance from the lower end. [B09] When the number of pixels having the signal intensity value equal to or greater than the threshold TA is 40% or more, for example, [B10] Standard deviations Sf (ie, Sf, cos) of the sine phase difference and cosine phase difference. , Sf, sin) and [B11] the linear dependence of the observed phase difference δφ and the vertical pixel distance from the lower end of the search window, that is, the standard deviation S (ie, Scos, Ssin) of the values on the left side of Equations 6 and 7. Calculate each. [B12] When the respective standard deviations S are smaller than the standard deviation Sf, [B13] The absolute values of the differences between the left side values of Equations 6 and 7 and the average value A are both threshold values TB (for example, 0 .15) For the number of pixels that are less than 15, the vertical width Ws pixel is counted for, for example, a sub-region divided horizontally into four. [B14] When the density of the points obtained in B13 is 0.4 or more in all the divided sub-regions, a point satisfying the condition of B13 in this search window is adopted as the candidate point 1. In addition, when this procedure is applied to a vertical structure having a height exceeding the minimum detection height H, each pixel is continuously detected in the vertical direction by the vertical structure extraction filter having the minimum detection height H. As a result, In terms of technique, it is extracted as a candidate point of the vertical structure without any height restriction.

[Procedure C]
FIG. 7 shows a flow of the procedure C. This procedure C relates to grouping.
The candidate points extracted up to step B are grouped. [C01] Among the candidate points 1, the candidate point 1 located at the lower right end is set as, for example, group 0, and [C02] is searched for all candidate points 1 on the image and grouped according to the following rule. [C03] It is confirmed whether the found candidate point 1 is inside the existing group boundary. If the distance between the candidate points is smaller than the vertical threshold value TCV and the horizontal threshold value TCH, they are regarded as the same group, and the process returns to C01. If not, [C04] It is confirmed whether the found candidate point 1 is within 10 pixels in the vertical direction and 5 pixels in the horizontal direction, for example. [C04] If it is within that range, the existing group boundary is expanded so as to include the found candidate point 1 as one point of the rectangular boundary. [C06] Otherwise, it is regarded as another group and the found candidate point 1 is registered as a new group. [C07] These operations are performed until the grouping of all candidate points 1 is completed. Note that the group region is formed as a rectangular region whose apex is the candidate point that is present on the outermost side in the group.

  [C08] When the boundary point of the created group is included in another group area, they are integrated. [C09] The integrated group is divided into strip-like subgroups having a width of 5 pixels, for example. At that time, the vertical range of the subgroup is shortened according to the upper and lower ends of the candidate point 1 included. [C10] The maximum phase difference of candidate point 1 that can be taken from the vertical length of the subgroup is obtained. [C11] The obtained maximum phase difference is divided into, for example, 8 bins (sorting boxes), and the number of candidate points 1 included in each bin is obtained. [C12] When the number of candidate points 1 included in the bin having the maximum number is larger than, for example, 35% of the number of all candidate points 1, the subgroup is excluded from the candidates. This is due to the following reason. In other words, if each subgroup includes a vertical structure, it should have a phase difference distribution corresponding to the length of the subgroup region. Therefore, if the phase difference distribution of candidate points included in each subgroup is biased, it is excluded from the vertical structure candidates at this point.

[Procedure D]
FIG. 8 shows a flow of the procedure D. This procedure D relates to the application of the vertical structure extraction filter to the subgroup.
The vertical structure extraction filter is applied again to the subgroups that have not been excluded until step C. However, in this procedure, application is performed with the vertical length of the subgroup, not the preset minimum detection altitude. [D01] For the remaining subgroups up to the above, calculate the values of the left sides of Equations 6 and 7 for candidate point 1 in the subgroup, and evaluate the standard deviation. When both the standard deviations are equal to or less than a threshold value TE (for example, 0.15), the subgroup is determined as a vertical structure. If there is one or more subgroups that are considered vertical structures within each group, the group is considered to contain vertical structures. [D02] Within each group, the maximum length or maximum phase difference in the number of vertical pixels in the subgroups remaining up to the above is highly converted by Equation 3 or Equation 2, The height (ie height) of the vertical structure of the group.

  In the above description, the artificial vertical structure has been mainly described, but the present invention can also be applied to a case of terrain with a large step such as a cliff. Various methods have been reported for phase unwrapping in the interference SAR, but the present invention is different from those methods, and as described above, the interference data of an object having an artificial vertical structure is divided into small regions. Then, the presence / absence of a vertical structure in the small area is determined and grouped to perform phase connection processing (phase unwrapping process), thereby recognizing a single vertical structure. This method is not limited to an artificial vertical structure as long as it has a step having continuity, and can also be applied to a normal topographic map.

Claims (6)

A method of extracting a vertical structure from output interference image data or interferometry data of an interferometric synthetic aperture radar,
Setting the vertical structure extraction filter according to the slant range position so as to have a predetermined lateral width and a vertical length that is large on the front side of the slant range and small on the far side according to the cosine of the off-nadir angle at the time of observation;
While scanning the vertical structure extraction filter on the interference image data and interferometry data output data, the layover collapse amount and the observation phase difference are linearly dependent on the image pixels in the vertical structure extraction filter at each position. And using the cosine transformation of the linear dependency relationship and the sine transformation of the linear dependency relationship, and recording that the scanning position is a pixel of a vertical structure image candidate;
Grouping vertical structure image candidate pixels within a predetermined range as a series of vertical structures to form vertical structure image candidates;
Arranging the vertical structure image candidates as vertically existing vertical structure images; and
Obtaining a height of the vertical structure from the vertical structure image obtained by organizing;
A method for extracting a vertical structure from a SAR interferogram.   The above judgment using the cosine transformation and sine transformation of the linear dependence relationship is the threshold judgment for the standard deviation of each cosine function of the linear dependence relationship between the layover collapse amount and the observed phase difference and the standard deviation of the sine function. The method for extracting a vertical structure from a SAR interferogram according to claim 1, wherein:   3. The vertical structure image pixels corresponding to the series of vertical structures by the grouping are unwrapped in phase as the series of vertical structure image pixels are continuous. The method for extracting a vertical structure from a SAR interferogram according to any one of the above.   4. The vertical structure image according to claim 1, wherein each image pixel point showing a maximum or minimum amount of layover collapse is set as a highest altitude point or a minimum altitude point of the vertical structure. 5. A method for extracting a vertical structure from a SAR interferogram according to claim 1.   The intensity distribution of the output interference image data or interferometry data of the interferometric synthetic aperture radar is applied with a median filter to determine that the linear dependence is present, and the intensity of the intensity distribution is a predetermined reference The method for extracting a vertical structure from a SAR interferogram according to any one of claims 1 to 4, wherein the determination of an image pixel exceeding 1 is made effective.   6. The method according to claim 1, wherein a median filter obtained by applying a median filter to a phase distribution of the output interference image data or interferometry data of the interference synthetic aperture radar is used for determining that the linear dependence relationship is present. The method for extracting a vertical structure from a SAR interferogram according to any one of the above.