JPH0634222B2 - Synthetic aperture radar image reproduction processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Use of the Invention] The present invention relates to a Synthetic Aperture Radar (hereinafter referred to as "SAR") mounted on an artificial satellite or an aircraft.
The present invention relates to a digital processing system for reproducing a human-understandable image from image pickup data according to (1), and particularly to a SAR image reproduction processing method suitable for high-speed processing.

[Background of the Invention]

In the field of remote sensing using artificial satellites or aircraft, as a sensor for imaging the ground surface, SAR that can obtain a high-resolution image in the microwave band that penetrates clouds
Is attracting attention.

Fig. 1 shows the overall system of SAR. The SAR having the radar sensor 1 and the antenna 2 is mounted on an artificial satellite or the like and moves on the flight path 3 in the direction of the arrow 4 to image the ground surface. The image data from the SAR is received by the ground station 5 and processed by the data processor 6 to create a video film 7 and a data storage magnetic tape 8. In addition, 9 is a decomposition cell, 10 is the range direction on the ground surface of the data sampled by SAR, 11 is the same azimuth direction, 12 is an antenna beam, and 13 is a cutting ring.

The outline of the processing of the data collected by the SAR is described above. For details, see Proceedings of 13th Inte
rnational Symposium on Remote Sensing of Environme
nt, P337-360, April, Bennette and Cum in 1977.
"Digital SAR Image Format im. Airborne an by ming
d Satellite Results ".

In the received image of SAR, one point on the original image is distributed with the spread of the point image pattern h (x, y),
It cannot be used as it is. Here, x indicates the range direction and y indicates the azimuth direction. The information spread in the received image is first compressed in the range direction,
Next, it is compressed in the azimuth direction. This is shown in FIG. FIG. 2A schematically shows a received image when only two microwave reflection points are present on the ground surface.
If the compression processing is performed in the direction, the original ground surface pattern as in B can be obtained. The range compression processing is performed by the correlation processing with the point image pattern data for each line of the image data. However, if the correlation process is executed as it is, it will take an enormous amount of processing time.
It is called "FFT". ) Higher speed is achieved by using complex multiplication and fast inverse Fourier transform (hereinafter referred to as "IFFT"). In order to perform correlation processing using FFT, first, a point image pattern is generated by digital processing by a computer, and FFTs of both the point image pattern and one line of image data are calculated.
Since the correlation calculation is a mere multiplication in the frequency domain after the FFT is applied, the product of the FFT calculation results of the above two data is calculated and IFFTed to obtain the correlation processing result for one line.

Although there are many types of FFT calculation methods, they can be roughly classified into an in-place type and a non-in-place type. FIG. 3 shows a data flow graph of the in-place type FFT when the number of data is 8, and FIG. 4 shows a data flow graph of the non-in-place type FFT. In both figures, the left side represents input data and the right side represents output data.

In the conventional SAR image reproduction processing, one of the above types of FF
Although T has been used, the in-place type FFT of FIG. 3 has been used.
However, since the result obtained by repeating the butterfly operation is not aligned in the frequency domain, bit inversion processing for rearranging the data is necessary and extra calculation time is required. On the other hand, in the non-in-place type FFT of FIG. 4, the results aligned in the frequency domain can be obtained, but since the storage location of the data to be read and the storage location of the data to be written by the butterfly operation in the middle of the calculation are different In addition to the location, a work storage location is required, which requires a large storage capacity.

[Object of the Invention]

The object of the present invention is to solve the above-mentioned drawbacks and to reduce the calculation amount in the SAR image reproduction processing.
It is to provide a processing method for reproducing an image at high speed.

[Outline of Invention]

In the SAR image reproduction process, only the FFT is used for performing the correlation calculation at high speed, and the FFT result aligned in the frequency domain is not necessary. Therefore, the in-place type FFT is performed.
It is possible to omit the bit reversal processing in.

First, two data strings to be correlated in the conventional method are defined as f _i , g _i (i = 0, 1, ..., N−1). N is the number of data points. FFT results of f _i and g _i are respectively F
_{If k} and G _k (k = 0, 1, ... N−1), then F _k and G
the product of ^{_{k F k · G k * (}} k = 0,1, ..., N-1, * is a complex conjugate) correlation _{f i} and _{g i} is obtained by calculating the IFFT of.

Next, in the processing procedure according to the method of the present invention, assuming that the value obtained by bit-reversing k is r (k), the FFT without the bit-reversal process is performed.
The result is F _{r (k)} and G _{r (k)} . Therefore, F _{r (k)} , G
_r the product of _{(k) F r (k)} · G r * (k) calculated, by performing the IFFT is omitted bit inversion processing will be exactly the same manner correlated with the previous step is determined. That is, in the in-place type FFT that uses a small memory capacity, the bit inversion process can be omitted in both FFT and IFFT, so that the correlation can be calculated at high speed.

On the other hand, in the SAR image reproduction processing, range curve correction processing is necessary. The range curve correction (hereinafter abbreviated as RC correction) corrects geometric distortion caused by a temporal change in the distance between the SAR antenna and the observation point. RC correction is normally performed in the frequency domain after FFT in the azimuth direction by compression processing in the azimuth direction.
FIG. 5 shows the principle of RC correction. One piece of information on the surface of the earth is
It is spread on the curve 14, and the data is rearranged by RC correction so that it becomes a straight line 15. The compression processing in the azimuth direction is performed by multiplying the rearranged data by the FFT of the point image data and performing IFFT.

If the equation of the curve 14 in RC correction is y = C (x), the data at the position of y _k = C (k) is required as the kth data. In general, y _k is a real number, so it is necessary to interpolate neighboring data in order to obtain the data at the position of y _k .

In order to perform the RC correction by the method of the present invention, the data at the position y _k ′ = C ′ (k) = C (r (k)) is required as the kth data. That is, in the method of the present invention, since bit inversion processing is omitted and the data after FFT are not arranged in frequency order, an originally continuous curve y = C (x) becomes a discontinuous curve y '= C' (x). However, since the bit inversion r is known as a function, the RC correction can be performed in the method of the present invention in exactly the same manner as in the conventional method.

Next, a problem when buffering is performed in RC correction will be described. Data amount of SAR image is several thousand x number
It is as large as several tens of megabytes with 1000 pixels, and it is not impossible to store all this in memory, but memory utilization efficiency decreases.

Since the number of lines required to output one line after RC correction is 100 lines at most, the whole image data is stored in a large-capacity external memory, and only the part of the image data necessary for RC correction is stored in the memory. The required memory capacity can be reduced by performing the buffering.

The buffering method is a problem here. There are two buffering methods, a cyclic buffer method and a helix buffer method. The cyclic buffer system cyclically uses the buffer area. That is, the buffer area is a two-dimensional array A (M, N) (M is the number of pixels in one line,
If N is the number of lines that can be stored in the buffer), the image data I _{ij at} the j-th line and the i-th pixel position is A (i, mod
It is stored in (j-1, N) +1). Where mod (k,
l) represents the remainder when k is divided by l.

In the cyclic buffer system, for example, the Nth line and the Nth line
Since the +1 line of image data is not located at consecutive positions in the buffer area, special address calculation processing is required when the curve y = C (x) to be corrected crosses the Nth line. In particular, when the bit inversion process in the FFT is omitted in the method of the present invention, the correction curve y ′ = C ′ (x) becomes discontinuous, and therefore the correction position y _k ′ and the buffer boundary positional relationship determination process is required for each pixel. Therefore, the calculation amount increases.

On the other hand, in the helix buffer method, the buffer area is a one-dimensional array B (M * N + L), and the j-th line and the i-th
The image data I _{ij at the} pixel position is B (j + N * (i-1))
Stored in. Here, L is the number of lines of the entire image.
In this method, image data of adjacent lines, for example, I _ij , I _i , _{j + 1} are always at continuous positions in the buffer as long as they are in the buffer. Therefore, the positional relationship determination processing in the cyclic buffer method is not necessary, and the calculation amount can be reduced especially when the bit inversion is omitted.

An embodiment of the present invention will be described below with reference to FIGS. 6 and 7.

FIG. 6 shows a flow chart of the azimuth compression processing in the SAR image reproduction processing. FIG. 7 is a block diagram of an apparatus that executes the same process. The apparatus shown in FIG. 7 is an external storage device 23 in which data after range compression is stored, an arithmetic processing device 2
4, an external storage device 25 for storing the azimuth compression result, and a main storage device 26.

Here, the arithmetic processing unit 24 is a vector processor. Vector processors exercise their power when operations are vectorized. The vector operation is the same operation repeatedly performed on a plurality of elements, and is executed at high speed by dedicated hardware called a vector processor.

First, in step 16 of FIG. 6, the image data after range compression is read from the external storage device 23 and the main storage device 26 is read.
It is stored in the inside helicopter battery. Next, in step 17, the FFT in which the bit inversion processing is omitted is calculated for the data in the helix buffer. FFT
Can be performed for each line, but here, the FFT is simultaneously performed on the data of a plurality of lines to increase the number of repetitions of the same operation and make the best use of the performance of the vector processor.

Next, in step 18, the range curve correction is performed to rearrange the information of one point on the surface of the earth, which is expanded in a curved line, so as to be aligned on a straight line. At this time, since the image data in the helix buffer is not aligned in the frequency order, the curve to be corrected is discontinuous.

The corrected data is multiplied at step 20 by the azimuth reference function generated at step 19. The azimuth reference function generated here is also not aligned in the frequency order because the bit inversion is omitted in the FFT as in the image data. The result of multiplication is IFFTed by step 21 and S
Obtain an AR playback image. The result is output to the external storage device 25 at step 22.

As described above, according to this embodiment, since the bit inversion processing in FFT and IFFT can be omitted, there is an effect that the SAR image reproduction processing can be performed at high speed.

〔The invention's effect〕

According to the present invention, since bit inversion processing can be omitted in FFT and IFFT, there is an effect that the amount of calculation in the SAR image reproduction processing is reduced and the processing speed is increased.

[Brief description of drawings]

FIG. 1 is an overall view of the SAR system, FIG. 2 is a view showing the principle of SAR image reproduction, FIGS. 3 and 4 are data flow graphs, and FIG. 5 is a view showing the principle of range curve correction.
FIG. 6 is a flow chart of the azimuth compression process, and FIG. 7 is a block configuration diagram of an apparatus that executes the process of FIG.

Claims (2)

[Claims] 1. A range compression is performed on image data obtained by a synthetic aperture radar, the range-compressed image data is stored in a memory, and an operation equivalent to a high-speed Fourier transform without bit inversion processing is performed on the stored image data. , Inverse high-speed Fourier transform in which range correction is performed on the calculated image data, correlation calculation between the corrected data and a predetermined reference function is performed, and bit inversion processing is omitted based on the result of the correlation calculation. Is performed and the original image reproduced by the calculation is output, and the synthetic aperture radar image reproduction processing method. 2. The synthetic aperture radar image reproduction processing method according to claim 1, wherein the range-compressed imaging data is temporarily stored in a helical buffer.