Image analysis device and image analysis method
The image analysis device 100 includes a pixel selection unit 11 which selects multiple pixels at a plurality of positions in one image among multiple images in which the same area is recorded, a dimension reduction unit 12 which compresses a complex vector as an evaluation value when an evaluation function is optimized into a low-dimensional space, an expanding unit 13 which returns a compression result by the dimension reduction unit 12 to an original pixel space, and calculates a spatial correlation phase estimate; and an optimization unit 14 which optimizes the evaluation function by bringing the evaluation value closer to the spatial correlation phase estimate and a pixel value at the position selected by the pixel selection unit 11.
1 . An image analysis device comprising:
a memory storing software instructions, and
one or more processors configured to execute the software instructions to;
obtain, using synthetic aperture radar (SAR) technology, multiple SAR images in which a same area is recorded;
select multiple first pixels at a plurality of positions in a first SAR image among the multiple SAR images, identify second pixels in one or more second SAR images corresponding to a same position as one of the first pixels in the first SAR image, and specifies a common position in the first SAR image and the second SAR image;
optimize an evaluation function by bringing an evaluation value closer to a spatial correlation phase estimate and a pixel value at the specified common position;
compress a complex vector as the evaluation value obtained by optimizing the evaluation function into a low-dimensional space; and
return a compression result to an original pixel space, and calculate the spatial correlation phase estimate,
wherein the evaluation function comprises adding first terms corresponding to observed pixel values and second terms corresponded to the spatial correlation phase estimate by weighting a product of a complex number indicating the pixel value at the specified common position and a conjugate complex number of the complex number.
2 . The image analysis device according to claim 1 ,
wherein the one or more processors are configured to execute the software instructions to restrict a value obtained by optimizing the evaluation function to within a predetermined value by nonlinear transformation.
3 . The image analysis device according to claim 1 , wherein the one or more processors are configured to execute the software instructions to;
multiply the complex vector by a matrix for low-dimensional compression, and
derive the spatial correlation phase estimate by multiplying the compression result by a matrix for expanding.
4 . The image analysis device according to claim 1 , wherein the one or more processors are further configured to execute the software instructions to:
smooth the compression result by a past compression result.
5 . The image analysis device according to claim 1 , further comprising:
absolute value storage configured to store an absolute value of an output of an expanding process.
6 . The image analysis device according to claim 1 , further comprising:
phase storage configured to store a phase of an output of an expanding process.
7 . The image analysis device according to claim 1 , further comprising:
absolute value storage configured to store an absolute value of an output of an optimizing process.
8 . The image analysis device according to claim 1 , further comprising:
phase storage configured to store a phase of an output of an optimizing process.
9 . An image analysis method, implemented by a processor, comprising:
obtaining, using synthetic aperture radar (SAR) technology, multiple SAR images in which a same area is recorded;
selecting multiple first pixels at a plurality of positions in a first SAR image among the multiple SAR images, identifying second pixels in one or more second SAR images corresponding to a same position as one of the first pixels in the first SAR image, and specifying a common position in the first SAR image and the second SAR image;
optimizing an evaluation function by bringing an evaluation value closer to a spatial correlation phase estimate and a pixel value at the specified common position;
compressing a complex vector as the evaluation value obtained by optimizing the evaluation function into a low-dimensional space; and
returning a compressing result to an original pixel space, and calculating a spatial correlation phase estimate,
wherein the evaluation function comprises adding first terms corresponding to observed pixel values and second terms corresponded to the spatial correlation phase estimate by weighting a product of a complex number indicating the pixel value at the specified common position and a conjugate complex number of the complex number.
10 . The image analysis method, implemented by the processor, according to claim 9 ,
further comprising restricting a value obtained by optimizing the evaluation function to within a predetermined value by nonlinear transformation.
11 . A non-transitory computer readable recording medium storing an image analysis program which, when executed by a processor, performs:
obtaining, using synthetic aperture radar (SAR) technology, multiple SAR images in which a same area is recorded;
selecting multiple first pixels at a plurality of positions in a first SAR image among the multiple SAR images, identifying second pixels in one or more second SAR images corresponding to a same position as one of the first pixels in the first SAR image, and specifying a common position in the first SAR image and the second SAR image;
optimizing an evaluation function by bringing an evaluation value closer to a spatial correlation phase estimate and a pixel value at the specified common position;
compressing a complex vector as the evaluation value obtained by optimizing the evaluation function into a low-dimensional space;
returning a compressing result to an original pixel space, and calculating a spatial correlation phase estimate; and
optimizing the evaluation function by bringing the evaluation value closer to the spatial correlation phase estimate and a pixel value at the selected position,
wherein the evaluation function comprises adding first terms corresponding to observed pixel values and second terms corresponded to the spatial correlation phase estimate by weighting a product of a complex number indicating the pixel value at the specified common position and a conjugate complex number of the complex number.
12 . The non-transitory computer readable recording medium according to claim 11 ,
wherein the image analysis program further performs
restricting a value obtained by optimizing the evaluation function to within a predetermined value by nonlinear transformation.