Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T5/00—Image enhancement or restoration
  • G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/10—Segmentation; Edge detection
  • G06T7/11—Region-based segmentation
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10032—Satellite or aerial image; Remote sensing
  • G06T2207/10044—Radar image
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20004—Adaptive image processing
  • G06T2207/20012—Locally adaptive
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/20—Special algorithmic details
  • G06T2207/20076—Probabilistic image processing
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30181—Earth observation

Definitions

• the present invention relates to a process for identifying statistically homogeneous pixels in SAR images acquired on the same area.
• a synthetic aperture radar (or SAR) system produces a two-dimensional image.
• One dimension of the image is called the range and is the measurement of the distance in a view line from the radar of the object that is being illuminated.
• the other dimension is called the azimuth and is perpendicular to the range.
• the SAR radar operates at a frequency that is generally comprised between 400 Mhz and 10 Ghz, and is usually installed on aeroplanes or satellite platforms orbiting at a height of between 250 and 800 Km.
• the antenna of the radar points to earth orthogonally to the direction of motion of the platform (aeroplane or satellite) at an angle known as the "offhadir" angle comprised between 20 and 80 degrees compared with the nadir direction, i.e. perpendicularly to the earth.
• the SAR is a consistent sensor and therefore the images are matrices of complex numbers the amplitude values of which are linked to the power that is backscattered by the illuminated objects (i.e. to the radar cross section thereof), whilst the step is determined by the nature of the target and the distance of the target from the radar.
• the SAR images are suitable for different applications; amongst these, the applications linked to the identification and classification of targets, "change detection" and interferometry applications are of primary importance. The latter are usually aimed at obtaining digital elevation models and/or the analysis of surface deformation of the terrain from sets of multitemporal SAR data.
• a general SAR image gathers data coming from targets of various natures: natural (woods, rocks, meadows, etc) or artificial (manufactured products, metal structures, motor vehicles, etc).
• the electromagnetic features can thus vary significantly even between adjacent pixels of a radar image.
• the object of the present invention is to provide a process for identifying statistically homogeneous pixels of SAR images acquired on the same area, i.e. characterised by similar electromagnetic properties, on which it is possible to make reliable statistical estimates, on a support selected in an adaptive manner.
• this object is achieved by a process for identifying statistically homogeneous pixels in images acquired on the same area by means of a synthetic aperture radar (SAR sensor) comprising the following steps: (a) acquiring a plurality of N radar images by means of a SAR sensor on the same area with acquisition geometries that are suitable to consent the common grid resampling of data,
• the process according to the invention also enables the response of the punctiform radar targets to be preserved.
• figure 1 shows a series of SAR images that are acquired and resampled on a common grid
• figure 2 show schematically an estimation window with pixels that are homogenous with the sample pixel, but are not connected to the sample pixel and pixels that are homogenous with and connected to the sample pixel
• figure 3 shows the average reflectivity map of the area of the Etna volcano
• figure 4 shows for each pixel the number of statistically homogeneous pixels associated therewith, according to the process of the invention
• figure 5a shows an image before the anti-speckle filter is applied according to the process of the invention
• figure 5b shows the same image of the figure 5a after the application of the anti-speckle filter according to the process of the invention
• figure 6a shows an interferogram before the application of the filter for inter fero grams according to the process of the invention
• figure 6b shows the same image as figure 5a after the application of the filter
• the process according to the invention works on collections (or datasets) of N radar images A 1..AN (figure 1) relating to the same area, detected at different times and/or at different view angles, but such as to consent the common grid resampling of data. Downstream of the resampling, all the dataset images are made to be superimposable on one another on a common grid, such that homologous pixels in two general images of the set of data correspond to the same portion of terrain illuminated by the radar.
• the values on which the process operates are the amplitude values of the signal that is acquired by the SAR relating to a certain cell of resolution pixels on the ground, namely the N amplitude values recorded at a certain pixel in the N acquisitions available on the interest area.
• Operating only on the amplitude values enables a smaller computational calculation to be made and enables the process to be made independent of a whole series of effects, which are well known in SAR interferometry, which alter the phase value of the signal, but which maintain the module thereof substantially unaltered, such as, for example, the atmospheric effects or the trajectory variations of the satellite.
• the process can also be applied to the intensity values of the images, i.e. the squared amplitude values.
• the process comprises selecting a pixel as a sample pixel.
• a vector of N amplitude values is constructed relating to the area illuminated by the radar during the course of the acquisitions.
• Vectors of amplitude values with N dimensions are thus obtained.
• the vector of amplitude of the sample pixel is defined as the sample vector.
• the process according to the invention comprises (figure 2), for each sample pixel 1 (indicated by a black square), the definition of an area of interest 10, said estimation window, within which to seek the pixels that are statistically homogeneous to the sample pixel.
• the shape and dimension of the estimation window are parameters that may vary from one application to another but usually a rectangular window is selected that comprises a few hundred pixels.
• the vector of N amplitude values is calculated in the same manner as the sample vector.
• the estimation window identifies a set of pixels and a set of vectors of amplitude that have the same dimension N as the sample vector and which can be compared with the sample vector in the search for a statistically homogenous behaviour.
• the problem is traceable to the comparison between two vectors of random variables containing the same number N of samples, about which it is desired to ascertain whether they can be considered to be two embodiments of the same distribution function or embodiments of different distribution functions. It is necessary to conduct a test to compare each vector associated with the pixel of the estimation window 10 and the sample vector.
• test can be of the non-parametric type.
• pixels 2 and 3 (figure 2) will be marked as being homogenous with the sample pixel 1 (where the pixels 3 are indicated by grey circles and the pixels 2 are indicated by black circles) and the pixels 4 will be marked as not being homogenous with the sample pixel (indicated by white circles).
• the process according to the invention thus comprises the following steps: acquiring a plurality of N radar images Al..AN by means of a SAR on the same area with similar acquisition geometries and such as to consent the common grid resampling of data, downstream of the resampling, selecting a pixel and identifying the pixel as a sample pixel 1, calculating a vector of N amplitude values relating to the sample pixel in the N images available and identifying the vector as a sample vector, defining an estimation window 10 (figure 2) for the sample pixel, for identifying a set of pixels in the neighbourhood of the sample pixel, calculating the vectors of N amplitude values for each other pixel contained in the estimation window, similarly to what is done for the sample pixel; - comparing, via a statistical test, each vector of amplitude associated with the pixels belonging to the estimation window with the sample vector to ascertain which vectors of amplitude are generated by the distribution function of the sample vector, identifying as pixels (figure 2) that are homogenous with the sample
• the set of the pixels that are statistically homogeneous with the sample pixel can be used to conduct estimates of averages or other estimates.
• Each point of the image can be selected as a sample pixel and the process can thus be conducted on all the pixels of the area of interest.
• Some sample pixels may not have homogeneous pixels contained in the respective estimation windows. This latter case is typical of so-called punctiform pixels, with electromagnetic behaviour that is distinctively different from that of the surrounding terrain.
• the process for identifying statistically homogeneous pixels of SAR images according to the invention is implemented by application software installed in a memory of a processing device; the latter comprises a microprocessor that converses with the memory to run said application software.
• the estimation window is chosen by the user.
• the process according to the invention is used to reduce the speckle noise of a SAR image.
• a dataset consisting of 75 multitemporal radar data acquired by the ERS-I and ERS-2 satellites of the European Space Agency is used to create, for each pixel of the image, the set of pixels that are statistically homogenous with the process according to the invention.
• the estimation window used in this and in the subsequent examples is rectangular in shape and measures 13 x 25 pixels (respectively in the range and the azimuth directions): it can thus contain a maximum value of 325 homogenous pixels.
• the map of the average reflectivity of the area of interest is shown, i.e. the average of the amplitude values relating to the 75 multitemporal radar data acquired by the ERS-I and ERS-2 satellites on Mount Etna; the horizontal dimension is the azimuth coordinate whilst the vertical dimension is the range coordinate.
• figure 4 for each pixel there is shown the number of statistically homogenous elements associated therewith identified by the process, whilst figure 5b shows the benefits of the use of the process according to the invention as an anti-speckle filter compared with an anti- speckle filter of known type (figure 5a).
• the speckle noise corresponds to variations in the amplitude values of the signal that are also located on homogenous targets that are observed in the data obtained by the so-called consistent observation systems, such as the SAR systems.
• the speckle noise is reduced by means of a simple algorithm that is movable only on the amplitude values relating to the pixels that are statistically homogeneous with the current pixel.
• the result of the filtering conducted according to the process according to the invention is notable, such as to be able to compare the single acquisition (filtered on adaptive windows corresponding to the statistically homogeneous pixels) with the map of average reflectivity (figure 3), in which all 75 available images are used.
• This whilst the spatial resolution of the data is maintained unaltered, does not, however make a multitemporal analysis of the reflectivity values possible but can, on the other hand, show variations over time of the RCS values originated by various phenomena such as changes in ground humidity, variations in vegetation, presence of artificial targets, etc.
• the second application of the process according to the invention is the use thereof as a filtering tool of SAR inter fero grams.
• FIGS. 6a and 6b By using the same database of SAR images acquired by the ERS satellites of the preceding example, in figures 6a and 6b there are shown the interferometric fringes of the zone of Valle del Bove, obtained by a pair of ERS images. Each 'fringe' corresponds to a motion of the terrain along the view line of a little less than 3 cm.
• the third application of the process according to the invention relates to the estimation of consistency of two SAR images.
• Figures 7a and 7b are compared with what is obtainable with the adaptive process disclosed in the preceding paragraphs and the result of traditional consistency estimation, i.e. by means of a process of a movable average on a 13x25 samples window, i.e. of the same dimensions as the estimation window of the homogenous pixels used in the examples but without any selection of the homogenous pixels; one definition of consistency is disclosed in the article by Touzi, Lopes, Bruniquel, Vachon, "Coherence estimation for SAR imagery" IEEE, Trans. Geosc. Remote Sensing, vol. 37, No. 1, pages 135-149, January 1999.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Radar Systems Or Details Thereof (AREA)
• Image Processing (AREA)
• Testing Of Coins (AREA)
• Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
• Holo Graphy (AREA)

Priority Applications (7)

Applications Claiming Priority (2)

Publications (1)

Family

ID=41213312

Family Applications (1)

Country Status (9)

Cited By (4)

Families Citing this family (16)

Citations (1)

Family Cites Families (17)

• 2009
  • 2009-04-03 IT ITMI2009A000535A patent/IT1393687B1/it active
• 2010
  • 2010-03-26 DK DK10713323.3T patent/DK2415017T3/en active
  • 2010-03-26 WO PCT/EP2010/054016 patent/WO2010112426A1/en not_active Ceased
  • 2010-03-26 JP JP2012502602A patent/JP5587972B2/ja active Active
  • 2010-03-26 US US13/259,363 patent/US8587471B2/en active Active
  • 2010-03-26 EP EP10713323.3A patent/EP2415017B1/en active Active
  • 2010-03-26 AU AU2010230359A patent/AU2010230359B2/en active Active
  • 2010-03-26 ES ES10713323T patent/ES2701003T3/es active Active
  • 2010-03-26 CA CA2756843A patent/CA2756843A1/en not_active Abandoned

Patent Citations (1)

Non-Patent Citations (7)

Also Published As

Similar Documents

Legal Events