Patent Grant
7750902
The present invention relates to the field of topography, and, more particularly, to a system and related methods for generating topographical models.
Topographical models of geographical areas may be used for many applications. For example, topographical models may be used in flight simulators and for planning military missions. Furthermore, topographical models of man-made structures (e.g., cities) may be extremely helpful in applications such as cellular antenna placement, urban planning, disaster preparedness and analysis, and mapping, for example.
Various types and methods for making topographical models are presently being used. One common topographical model is the digital elevation map (DEM). A DEM is a sampled matrix representation of a geographical area which may be generated in an automated fashion by a computer. In a DEM, coordinate points are made to correspond with a height value. DEMs are typically used for modeling terrain where the transitions between different elevations (e.g., valleys, mountains, etc.) are generally smooth from one to a next. That is, DEMs typically model terrain as a plurality of curved surfaces and any discontinuities therebetween are thus “smoothed” over. Thus, in a typical DEM no distinct objects are present on the terrain.
One particularly advantageous 3D site modeling product is RealSite® from the present Assignee Harris Corp. RealSite® may be used to register overlapping images of a geographical area of interest, and extract high resolution DEMs using stereo and nadir view techniques. RealSite® provides a semi-automated process for making three-dimensional (3D) topographical models of geographical areas, including cities, that have accurate textures and structure boundaries. Moreover, RealSite® models are geospatially accurate. That is, the location of any given point within the model corresponds to an actual location in the geographical area with very high accuracy. The data used to generate RealSite® models may include aerial and satellite photography, electro-optical, infrared, and light detection and ranging (LIDAR).
Another advantageous approach for generating 3D site models is set forth in U.S. Pat. No. 6,654,690 to Rahmes et al., which is also assigned to the present Assignee and is hereby incorporated herein in its entirety by reference. This patent discloses an automated method for making a topographical model of an area including terrain and buildings thereon based upon randomly spaced data of elevation versus position. The method includes processing the randomly spaced data to generate gridded data of elevation versus position conforming to a predetermined position grid, processing the gridded data to distinguish building data from terrain data, and performing polygon extraction for the building data to make the topographical model of the area including terrain and buildings thereon.
While the above-noted approaches provide exceptional 3D models of urban areas with accurate and realistic cultural (e.g., building) feature detail, in some applications it may be desirable to produce a topographical model of a geographical area of interest without the cultural features otherwise present in the area of interest. Yet, once the cultural features are identified and extracted from the terrain data, there may be voids left in the resulting DEM. Moreover, in some situations it may be desirable to focus on cultural features from an area of interest, but foliage, etch, may obscure portions of one or more cultural features that will similarly result in voids in the cultural feature when the foliage is extracted.
Various interpolation techniques are generally used for filling in missing data in a data field, One such technique is sinc interpolation, which assumes that a signal is band-limited. While this approach is well suited for communication and audio signals, it may not be well suited for 3D data models. Another approach is polynomial interpolation. This approach is sometimes difficult to implement because the computational overhead may become overly burdensome for higher order polynomials, which may be necessary to provide desired accuracy.
One additional interpolation approach is spline interpolation. While this approach may provide a relatively high reconstruction accuracy, this approach may be problematic to implement in a 3D data model because of the difficultly in solving a global spline over the entire model, and because the required matrices may be ill-conditioned. One further drawback of such conventional techniques is that they tend to blur edge content, which may be a significant problem in a 3D topographical model.
Another approach for filling in regions within an image is set forth in U.S. Pat. No. 6,987,520 to Criminisi et al. This patent discloses an examplar-based filling system which identifies appropriate filling material to replace a destination region in an image and fills the destination region using this material. This is done to alleviate or minimize the amount of manual editing required to fill a destination region in an image. Tiles of image data are “borrowed” from the proximity of the destination region or some other source to generate new image data to fill in the region. Destination regions may be designated by user input (e.g., selection of an image region by a user) or by other means (e.g., specification of a color or feature to be replaced). In addition, the order in which the destination region is filled by example tiles may be configured to emphasize the continuity of linear structures and composite textures using a type of isophote-driven image-sampling process.
Another way in which geospatial model data can end up with voids therein is when the data is collected in the frequency domain, as is the case with Synthetic Aperture Radar (SAR) data collection. That is, a SAR returns a map or representation of radar reflectivity including both amplitude and phase over a plurality of different frequencies. However, due to interference from existing signal sources, during some SAR scans certain frequency bands may experience interference in the resulting SAR data. Moreover, the operator of the SAR may have to intentionally omit or block certain frequency bands in certain geographical areas from the scan to avoid interfering with such communication sources. Further, hardware malfunctions may result in pulse dropouts. In each of these cases, the result is that the frequency domain representation of the area of interest will have gaps or voids therein, which when converted to the spatial domain cause the resulting geospatial model image to be distorted.
Generally speaking, various approaches have been used to address the effects of interference in frequency domain data. One approach is to use linear interpolation. Super resolution and/or iterative convolution techniques have also been used which assume a point like target in the image. Moreover, hardware approaches have also been implemented to alter mode hopping to avoid interference or listening on pilot pulses to characterize the interference.
Another approach to interference suppression in SAR images is set forth in an article entitled “Interference Suppression in Synthesized SAR Images” by Reigber et al., IEEE Geoscience and Remote Sensing Letters, vol. 2, no. 1, January 2005. This article proposes an interference suppression approach that relies on the transformation of synthesized SAR images into a representation where common raw-data interference filtering methods can be applied. More particularly, this approach uses a posteriori filtering.
Despite the advantages such prior art approaches may provide in certain applications, further advancements may be desirable for filling voids in geospatial model data.
In view of the foregoing background, the present disclosure presents a geospatial modeling system and related methods which may advantageously account for voids within geospatial models and related methods.
A geospatial modeling system may include a geospatial model database and a processor. More particularly, the processor may cooperate with the geospatial model database for inpainting data into at least one void in geospatial model cultural feature data based upon propagating contour data from outside the at least one void into the at least one void.
More particularly, the geospatial model cultural feature data may include elevation data. As such, the processor may inpaint by propagating elevation contour data from outside the at least one void along a direction of lines of constant elevation contour from outside the at least one void into the at least one void. Moreover, the processor may iteratively propagate the contour data from outside the at least one void into the at least one void.
The processor may receive as an input geospatial model data containing both geospatial model foliage data and geospatial model cultural feature data, and extract the geospatial model cultural feature data. More specifically, the processor may determine the at least one void in the geospatial model cultural feature data based upon extraction of the geospatial model foliage data, and the geospatial model cultural feature data may include at least one building boundary.
In addition, the processor may receive as an input geospatial model data based upon a single reflective surface data capture for a given area. The geospatial model data may also include digital elevation model (DEM) data having a resolution of less than 30 meters, for example. Furthermore, the processor may perform inpainting based upon at least one turbulent fluid flow modeling equation. More particularly, the at least one turbulent fluid flow modeling equation may be Navier-Stokes equations. Also, the geospatial modeling system may further include a display coupled to the processor for displaying geospatial model cultural feature data produced thereby.
A geospatial modeling method aspect may include providing geospatial model cultural feature data. The method may further include inpainting data into at least one void in the geospatial model cultural feature data based upon propagating contour data from outside the at least one void into the at least one void.
The present invention will now be described more fully hereinafter with reference to the accompanying drawings, in which preferred embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Like numbers refer to like elements throughout, and prime and multiple prime notation are used to indicate similar elements in alternative embodiments.
Referring initially to
Turning additionally to
In the illustrated example, a single reflective surface data capture is performed to provide the 3D data of the geographical area of interest, at Block 31. The “raw” data provided from the collection will typically include terrain, foliage, and/or cultural features (e.g., buildings). The processor 22 uses this raw data to generate a geospatial model (i.e., DEM) of the elevation verses position data based upon the known position of the collectors, etc., at Block 32, using various approaches which are known to those skilled in the art. Of course, in other embodiments the DEM may be generated by another computer and stored in the geospatial model database 21 for processing by the processor 22. The DEN data may have a relatively high resolution, for example, of greater than about thirty meters to provide highly accurate image detail, although lower resolutions may be used for some embodiments, if desired. In some embodiments, resolutions of one meter or better may be achieved.
In many instances it is desirable to separate or extract one of the above-noted types of data from a geospatial model. For example, in some cases it may be desirable to remove the cultural features from a DEN so that only the terrain and/or foliage remains, at Block 33. In particular, the extraction process may include a series of DEM re-sampling, null filling, DEM subtraction, and null expanding steps, as will be appreciated by those skilled in the art. Yet, extracting the cultural features would ordinarily leave holes or voids within the DEM. A DEM 40a is shown in
When features have been extracted from the geospatial model, this makes determination of voids to be filled (Block 34) relatively straightforward, as these voids will occur where the cultural feature or other data has been extracted. However, in some embodiments the voids may result from causes other than data extraction, such as a blind spot of a collector, clouds over a geographical area or interest, etc. The approach described herein may also be used to correct such voids as well.
Generally speaking, the voids 41a are inpainted by propagating contour data from outside a given void into the given void, at Block 35. More particularly, the processor 22 inpaints by propagating elevation contour data from outside the given void along a direction of lines of constant elevation contour from outside the given void into the void, as seen in
More particularly, the processor 22 propagates elevation information from outside the void along a direction of iso-contour, as represented by the following equation:
where ∇L is a discrete Laplacian transform. An iso-contour direction N is obtained by taking a 90 degree rotation of the DEM gradient, as will be appreciated by those skilled in the art. An inpainting equation for performing the above-noted propagation is as follows:
Hn+1(i,j)=Hn(i,j)+ΔtHtn(i,j),∀(i,j)εΩ (2)
The above-noted propagation is performed a certain number of iterations to “shrink” the void to a desired size as seen in
Generally speaking, the above-described approach essentially treats a DEM as an incompressible fluid, which allows fluid mechanics techniques to be used for filling in the voids. That is, the partial differential equations outlined above are used to estimate how the boundaries directly adjacent a void in the 3D model would naturally flow into and fill the void if the DEM were considered to be an incompressible fluid, as will be appreciated by those skilled in the art.
This approach advantageously allows for autonomous reconstruction of bare earth in places where buildings or other cultural features have been removed, yet while still retaining continuous elevation contours. Moreover, the non-linear interpolation technique of inpainting allows for accurate propagation of data from the area surrounding a void boundary Further, the DEM may advantageously be iteratively evolved until a steady state is achieved, and the speed of propagation may be controlled to provide a desired tradeoff between accuracy of the resulting geospatial data and the speed so that the processing overhead burden does not become undesirably large, as will be appreciated by those skilled in the art.
The above-described approach may similarly be used to reconstruct other features besides terrain. More particularly, it may be used to perform inpainting on voids in a cultural feature (e.g., building) resulting from foliage, etc., that obscures part of the cultural feature. Turning now additionally to
As discussed above, the processor 22 inpaints by iteratively propagating elevation contour data from outside the voids 51a in data portions 52a, 62a along a direction of lines of constant elevation contour from outside the voids into the voids, at Blocks 35′-36′, to produce the final “repaired” data portions 52b, 62b in which building edges 55b′, 65b′ are now complete and continuous. The inpainting process is further illustrated in
where ∇H is the DEM gradient and ∇PH is the iso-contour direction to produce the repaired data section 72b (
Turning additionally to
More particularly, a frequency domain data map 100 illustrated in
However, the above-described inpainting techniques may also advantageously be used for repairing such voids in geographical model frequency domain data. More particularly, the processor 22″ cooperates with the geospatial model database 21″ for inpainting data into the missing data bands 102 (i.e., voids) based upon propagating contour data from outside the voids into the voids, at Block 82. More particularly, the propagation occurs along a direction of lines of constant contour from outside the voids into the voids. Yet, rather than being based on elevation contour data as in the above-described examples, here the contour data corresponds to the phase and amplitude values of the data surrounding the voids. Here again, the propagation is preferably iteratively performed a desired number of iterations (Block 83), or until a steady state is achieved, as will be appreciated by those skilled in the art.
Once again, this approach is based upon reconstructing data for frequencies that are missing from a frequency domain representation of a geographical area of interest by modeling the spectral signatures that are present in the data surrounding the voids as a turbulent (i.e., fluid) flow. That is, each individual known frequency is treated as a particle in an eddy flow, which are small turbulence fields inside of a general turbulence field. As such, the known “eddies” in the frequency domain data can therefore be modeled to interpolate the missing values.
Generally speaking, the processor 22″ performs inpainting based upon one or more turbulent fluid flow modeling equations. By way of example, Navier-Stokes fluid mechanics equations/relationships may be used with some modification for K-space. More particularly, the stream function will have two components rather than one as follows:
Ψ=A(kx,ky)e=φ(k
where the functions A, R, and Q are four times differentiable, and z=√{square root over (−1)}. Thus, looking at the derived equations with respect to image intensities results in the following:
A similar Navier-Stokes approach may also be used for the terrain/cultural feature void inpainting operations described above, as will be appreciated by those skilled in the art.
After the iterative propagation is completed using the above-described approach, the K-space map 100b is “repaired” with the missing data bands 102a no longer present (or substantially diminished), as shown in
A computer-readable medium is also provided which may have computer-executable modules including a geospatial model database module for storing geospatial model cultural feature data, and a processing module for inpainting data into at least one void in the geospatial model cultural feature data based upon propagating contour data from outside the at least one void into the at least one void. The geospatial model cultural feature data may include elevation data. Moreover, the processing module may propagate elevation contour data from outside the at least one void along a direction of lines of constant elevation contour from outside the at least one void into the at least one void. Additionally, the processing module may iteratively propagate the contour data from outside the at least one void into the at least one void, as discussed further above.
Many modifications and other embodiments of the invention will come to the mind of one skilled in the art having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is understood that the invention is not to be limited to the specific embodiments disclosed, and that modifications and embodiments are intended to be included within the scope of the appended claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 6654690 | Rahmes et al. | Nov 2003 | B2 |
| 6987520 | Criminisi et al. | Jan 2006 | B2 |
| 7142984 | Rahmes et al. | Nov 2006 | B2 |
| 7502521 | Sun et al. | Mar 2009 | B2 |
| 20040153128 | Suresh et al. | Aug 2004 | A1 |
| 20050169533 | Paragyios | Aug 2005 | A1 |
| 20060013442 | McDowall et al. | Jan 2006 | A1 |
| 20060013443 | McDowall et al. | Jan 2006 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20080285885 A1 | Nov 2008 | US |
