JPH0799392B2 - Synthetic aperture radar image reproduction processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a synthetic aperture radar (hereinafter abbreviated as SAR) mounted on an artificial satellite or an aircraft.
The present invention relates to a digital processing system for reproducing a human-readable image from imaged data of the ground surface according to the present invention, and particularly to an image reproduction processing method suitable for reproducing a high quality image.

[Conventional technology]

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

Figure 7 shows the overall SAR system. Radar sensor
The SAR having the R and the antenna A _n is mounted on an artificial satellite or the like and moves along the flight path F _{P in the} direction of the arrow A to image the ground surface. The image data from the SAR is received by the ground station L _S and processed by the data processor D _P to create a video film IF and a data storage magnetic tape MT. Note that C is the decomposition cell, R _a is the range direction on the surface of the data collected by SAR, A _Z is the same azimuth direction, A _B is the antenna beam, and C _W is the cutting width. There is.

The outline of the processing of the data collected by SAR is described below.
For details, see, for example, JR Bennett.
e) et al. "Digital Synthetic Aperture Radar Image Generation, Airborne and Satellite Results (" Digital
SAR Image Formation, Airborne and Satellite Result
s "), Remote Sensing of the Environment 13th International Symposium Proceedings, P337-360, April 1977 (Proceedings of 13th I
nternational Symposium or Remote Slnsing of Enviro
nment, P337-360, April 1 1977).

1 in the original image (ground surface) in the received SAR image
The points are distributed with the spread of the point image pattern h (x, y), which cannot be understood by humans. 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 and then in the azimuth direction. The compression processing is performed by correlation processing with point image pattern data for each line of image data. However, if the correlation processing is executed as it is, it takes an enormous amount of processing time.
"T". ) Is used to increase the speed.

The point image pattern h (y) used for the above-described azimuth direction compression processing is a linear frequency modulation signal, Is expressed as Here, α is a constant, and K and β are called azimuth direction reproduction parameters. In the following, the term "reproduction parameter" simply refers to K or β. Of these, K is the frequency change rate, and is called the Doppler change rate because the wave number change is due to the Doppler shift.

In order to improve the image quality of the image data after the compression processing, it is important to accurately obtain the reproduction parameter. The Doppler change rate K is the satellite orbit data,
Although it is determined based on the attitude data, the earth rotation data, etc., an error occurs in the data used here, and an error from the true value also occurs in the Doppler change rate K, which appears as a blur pattern of the output image. FIG. 8 shows an example of this, where A ₁ is the original image, A ₂ is the point image pattern h (x, y) of two points on A ₁ ,
A ₃ schematically shows a blurring pattern caused by an error in the Doppler change rate when the point image pattern h (x, y) of A ₂ is compressed in the lens direction (x) and the azimuth direction (y). It is a thing.

In addition to this, multi-track processing is required for SAR data processing. The multi-track processing aims to reduce the speckle noise peculiar to a coherent system such as SAR processing to obtain a smooth image, and when SAR image azimuth compression is performed, the band is set in the frequency space immediately before the IFFT. The image is divided, compressed images are obtained for each of the divided bands, and the intensities of these images are added. As a result, since the frequency band is narrowed, the resolution of the output image is reduced, but the S / N ratio can be improved in proportion to (the number of divisions) ^−1/2 .

Conventionally, as a focusing method for removing an image blur pattern caused by an error in the Doppler change rate in an SAR output image, an image position shift between the two image data of the two image data generated by multi-track processing is measured, and the initial position is measured. A method for correcting the Doppler change rate is described in JP-A-58-186068.

[Problems to be solved by the invention]

Generally, the SAR imaging scene is about 75 km to 100 km square corresponding to the ground surface, and the number of pixels is several thousand pixels × several thousand pixels. In order to reproduce one picked-up scene by one computer processing, the above-mentioned data amount is enormous and it is disadvantageous in terms of utilization efficiency of the main storage device and other points. Playback processing is performed for each. Further, since the reproduction parameter changes gradually in the azimuth direction, it is necessary to process using different reproduction parameters for each divided portion.

However, the above-mentioned conventional technique does not take these points into consideration, and the image quality of the reproduced image is deteriorated because the processing is performed using a constant reproduction parameter in one shooting scene. Further, a method for improving the accuracy by combining the results of the reproduction parameter estimation performed for each divided portion, a method for accumulating the estimation results in a file and reusing them later have not been considered.

On the other hand, when processing is performed using different reproduction parameters for each divided portion, the problem that the complex phase changes discontinuously at the dividing boundary and the image quality deteriorates due to interpolation has not been recognized.

An object of the present invention is to solve the above-mentioned problems and to provide a high-quality SAR,
It is to provide a SAR image reproduction processing method that enables image reproduction.

[Means for Solving Problems] The above-mentioned object is to divide the imaged data in the azimuth direction, estimate the reproduction parameters for each of the divided partial imaged data, and use the estimated respective reproduction parameters. This can be achieved by reproducing a partial image for each of the above partial image pickup data, and correcting geometrical distortion between the obtained partial images to reconstruct continuous reproduced images.

Further, at the time of division, the reproduced images corresponding to the respective divided portions are overlapped by several pixels between the adjacent images, so that the deterioration of the image quality in the divided portions can be prevented.

[Action]

By performing the reproduction parameter estimation for each divided portion in the azimuth direction, an appropriate reproduction parameter value can be estimated over a wide range in the azimuth direction, so that it is possible to reduce the influence of an estimation error due to noise or the like.

In addition, by dividing so that adjacent reproduced images are overlapped by several pixels, interpolation at any position can be performed by using only pixels having a uniform phase, so that it is possible to prevent the occurrence of image quality deterioration due to position shift.

〔Example〕

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

FIGS. 1 (A) and 1 (B) are the overall processing flow of the satellite-mounted SAR image reproduction processing system. In the normal process flow,
First, in process 1, the SAR observation data 2 is input, compression processing is performed in the range direction, and the result is written to the work file 3. The process 1 is performed for the entire one imaged scene.

In process 4, the orbit / attitude data 5 of the satellite, which is acquired at the same time as the observation data 2, is input, and the change function of the reproduction parameter in the azimuth direction is calculated from these orbit / attitude data. Further, according to the calculated change function of the reproduction parameter, the value of the reproduction parameter (hereinafter referred to as the initial value of the reproduction parameter) for each division unit of the azimuth compression processing described later is calculated.

In process 6, temporary reproduction parameter estimation processing is performed for each division unit from the observation data after the compression processing in the range direction. This estimation processing is performed for each division unit when the division processing is performed in the azimuth compression processing described later. It is to be noted that what is called a tentative estimation process here is that the estimation value of the reproduction parameter from the observation data for each division unit includes an estimation error in the present embodiment, and is calculated previously in the next process 7. This is because the estimated value is corrected by taking into consideration the change function of the reproduction parameter, and the result is used as the final estimated reproduction parameter value for each division unit in the image reproduction process (azimuth direction compression process).

In process 7, the result of the temporary estimation of the reproduction parameter obtained in process 6 is subjected to fitting by the least square method with the azimuth direction change function of the reproduction parameter obtained in process 4. That is, the change function calculated in the process 4 is moved up or down in parallel, or the result is scaled up or down. The sum of squares of the error from the value of the reproduction parameter for each division unit is minimized, and the value of the reproduction parameter for each division unit according to the moved or enlarged / reduced change function is set as the corrected estimated value. At this time, if the estimation result of the adjacent image-capturing scene is stored in the reproduction parameter file 8, the fitting is performed using the estimation result. Further, the result of the fitting is stored in the reproduction parameter file 8 together with the information of the scene to be processed.

In process 9, the range-compressed data stored in the work file 3 is divided in the azimuth direction, and azimuth compression is performed for each divided portion using the reproduction parameter estimated in process 7. Next, in process 10, the complex image data obtained as a result of process 9 is sampled by interpolation to correct geometric distortion. In process 11, the result of process 10 is normed (realized) and the result is stored in the work file 12.

After performing processing 9 to processing 11 for all the divided parts,
In processing 13, the contents of the work file 12 are input, and multi-track addition processing and mosaicing, which is a stitching of divided portions, are performed.

In process 14, the result of process 13 is quantized into an 8-bit integer, and reproduced image data 15 is output.

The above is the normal processing flow, but on a regular basis, for example, once a week or once every 100 scenes,
Estimate the reproduction parameter in the range direction. In this mode of processing, as shown in FIG.
At 16, the observation data 2 is input, and range compression processing and range direction track division processing necessary for estimating reproduction parameters in the range direction are performed.

In process 17, azimuth compression is performed on each of the divided tracks, and in process 18, the reproduction parameter in the range direction is estimated using the result. In process 19, the estimation result is checked, and if the estimation result is within the accuracy sufficient for the range direction reproduction parameter given in advance as a system constant, it is determined to be normal, and if not, the SAR imaging system is determined to be abnormal. If it is determined to be abnormal, an alarm message for the operator is created in process 20 and an alarm is output to the console display 21 for the operator.

FIG. 2 shows a method of dividing the imaging data in the azimuth direction, which is performed in the processing 6 and the processing 9 of FIG. In FIG. 2, the horizontal axis represents the azimuth direction 22 and the vertical axis represents the range direction 23, which is an example divided into four parts. The number of divisions is determined by the size of the captured scene, which is the processing unit. N is the size of the division unit in the azimuth direction (pixel unit), and is selected to be a power of 2 because of the use of FFT. M is a size in which adjacent divided portions overlap each other, and if the size of the point image in the azimuth direction is L, M ≧ L / 4 is required in order to connect reproduced images without a gap. (When the number of multi-links = 4) Here, M = L / 4 + 3 is set so that reproduced images from adjacent divided portions overlap each other by three pixels.
In the geometric correction in the process 10 of FIG. 1, if interpolation using four neighboring points is performed, it is possible to perform seamless processing by the interpolation process using only the data that belong to the same divided portion and have the same phase. it can.

This is shown in FIG. For simplification, the vicinity of a division boundary of one line in the azimuth direction is shown. Pixels 31 reproduced from the left divided portion and pixels 32 reproduced from the right divided portion are shown. Since each segment is reproduced using different reproduction parameters, pixel 31 and pixel 32
And are out of phase. In the figure, the interpolation in the portion A can be performed using only the pixel 32, and the interpolation in the portion B can be performed using only the pixel 32. Therefore, interpolation is not performed using both pixels 31 and 32, and image quality deterioration does not occur. Needless to say, when four or more neighboring pixels are used in the interpolation, the value of M should be correspondingly increased.

FIG. 4 is a diagram showing the principle of the reproduction parameter estimation method in the processes 4, 6, and 7 of FIG. The horizontal axis 41 is the azimuth direction coordinate, and the vertical axis 42 is the reproduction parameter value. Straight line 43
Represents the change in the azimuth direction of the reproduction parameter calculated from the satellite orbit / attitude data. Value 44−a, 44−b, 44−
c and 44-d are the initial values of the reproduction parameters. Value 45−a,
45-b, 45-c, 45-d are provisional estimated values of the reproduction parameter obtained in the process 6 of FIG.
45-a to d are the best fit, value 46-
a, 46-b, 46-c, 46-d are the final estimated values of the reproduction parameters obtained as a result of the process 7 in FIG. Azimuth compression by the process 9 of FIG. 1 is performed using this final estimated value.

FIG. 5 is a principle diagram in the case of utilizing the estimation result of the reproduction parameters of adjacent scenes. The horizontal and vertical axes are the same as in FIG. In process 7 of FIG. 1, it is first checked whether or not the reproduction parameter estimation result of the adjacent scene of the processing target scene is stored in the reproduction parameter file 8 of FIG. If both adjacent scenes are stored, straighten it
51 and straight line 52. A straight line 51, a straight line 52, and a straight line 53 that best fits the estimation result of the reproduction parameter in the target scene as a whole are used as the estimation result of the target scene.
The same applies when only one value is stored in the adjacent scenes. Further, when processing a plurality of consecutive scenes at once, it is possible to perform the reproduction parameter estimation processing for each scene and perform high-accuracy estimation collectively by the same method as in FIG. Needless to say.

FIG. 6 shows a storage format in the reproduction parameter file 8 for storing the estimation result of the reproduction parameter obtained by the above method. The result estimated from FIG. 4 or FIG. 5 is expressed as a linear expression = at + b (t is a coordinate in the azimuth direction),
As the estimation result 65, a and b are stored together with the scene number 61, the latitude / longitude 62 of the center of the scene, the imaging date 63, the Path and Row number 64 indicating the position of the scene.

In the present embodiment, the change of the reproduction parameter in the azimuth direction can be expressed by a linear function, but the effect of the present invention is the same even if it is a higher-order function.

〔The invention's effect〕

According to the present invention, since the SAR image data is divided in the azimuth direction and the reproduction processing is performed using different reproduction parameters for each division, it is possible to use accurate reproduction parameters and improve the quality of reproduced images. To do. Further, as in the embodiment, if a change function in the azimuth direction of the reproduction parameter is obtained in advance from the position / velocity / orientation data of the SAR sensor and the geometric model of the earth, and the reproduction parameter estimation result is approximated by the change function, Further, the estimation accuracy of the reproduction parameter is improved, and as a result, the quality of the reproduced image is improved.

On the other hand, when the image data is divided in the azimuth direction, the reproduced images from the adjacent divided portions are overlapped with each other so that the image quality can be prevented from being deteriorated by the re-sampling at the time of geometric distortion correction.

[Brief description of drawings]

FIG. 1 is an overall processing flow of an embodiment of the present invention, FIG. 2 is a schematic diagram of division of imaging data in processing 6 and processing 9 in FIG. 1, and FIG. 3 is a division boundary in processing 10 in FIG. FIG. 4 is a diagram showing a state of re-sampling in the vicinity, FIGS. 4 and 5 are diagrams showing an example of an approximation process in the process 7 of FIG. 1, and FIG.
The figure shows an example of the storage format of the playback parameter file 8 in Fig. 1, Fig. 7 is a conceptual diagram of the entire SAR system, and Fig. 8 is the SAR.
It is a figure which shows the principle of an image reproduction process.

Claims (3)

[Claims] 1. In a computer system for reproducing an image from image data of a synthetic aperture radar, the image data is divided in the azimuth direction, a reproduction parameter is estimated for each of the divided partial image data, and each estimated image data is estimated. It is characterized in that a partial image is reproduced for each of the above-mentioned partial imaged data using a reproduction parameter, and geometric distortion between the obtained partial images is corrected to reconstruct a continuous reproduced image. Synthetic aperture radar image reproduction processing method. 2. The image pickup data is data obtained by subjecting a received signal of a synthetic aperture radar to compression in a range direction, and the reproduction parameter is an azimuth direction reproduction parameter. The synthetic aperture radar image reproduction processing method according to item 1. 3. The process of dividing the image pickup data in the azimuth direction is performed so that adjacent ones of partial images reproduced from the respective partial image pickup data partially overlap each other. The synthetic aperture radar image reproduction processing method according to claim 1.