JP2008089380A - Image reconstruction device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image reconstruction device with image processing time shortened by reducing the number of times of probe scanning at high resolution without extremely increasing storage capacity. <P>SOLUTION: This image reconstruction device comprises a wide-directivityprobe and an arithmetic unit with an image processing method incorporated thereinto, the processing method used for making three-dimensional image information by processing an output signal of the probe. The image processing method has a procedure for obtaining signals by Fourier-transforming time-series output signals obtained through thinned out sampling into time-series data corresponding to the thinned-out sampling, and signals by Fourier-transforming a reflection function corresponding to the thinned-out sampling by using directive function signals of time-series data corresponding to the thinned-out sampling, and for obtaining the distribution of the reflection function in real space by inversely Fourier-transforming the signals obtained through Fourier-transforming the reflection function. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image reconstruction apparatus that synthesizes a three-dimensional image having a desired resolution even when scanning is sampled at a coarser interval than the desired resolution.

In the image display using the ultrasonic flaw detection method / electromagnetic ultrasonic flaw detection method, the high-directivity single probe is mechanically scanned in the xy direction, or the multi-probe is electronically scanned, and the reflection time t of the reflected waveform signal as time series data is reflected. Is regarded as the depth z, and three-dimensional display is performed. In this case, the resolution in the xy direction is determined by the beam diameter, the sampling distance interval and wavelength in the xy direction, and the resolution in the z direction is determined by the sampling time interval and wavelength of the time series waveform. In the case of inspecting and observing defects that are sufficiently larger than the wavelength and probe diameter in a relatively small area, it is practically sufficient to prepare a computer memory that is sufficiently larger than the necessary memory.

However, in order to find a defect of 0.1 mm with a thick plate such as 1000 mm × 1000 mm × 30 mm, a unidirectional single probe requires more than 10,000 scanning lines. If it takes 1 second / line scanning time, it will take 10000 seconds. Further, in the case of a single probe, if each depth is inspected by using a convergent beam, the inspection time is further increased by several tens of times and dramatically increased.
In the case of a multi-probe, it is necessary to make a large probe in order to take advantage of the high speed of electronic scanning. However, the development of a large probe has problems in terms of cost and technology. Further, when scanning with a convergent beam inside the sample by a method called a phased array, the wave cannot be continuously emitted due to reverberation, and there is a limit to speeding up. In the case of a multi-probe, if the frequency is increased in order to increase the resolution, there is a problem that it becomes difficult to interfere with adjacent probes and to align the characteristics of each probe.

In the inspection of a large sample, the number of sampling points increases, so that it takes a long time to scan is a big obstacle. Specifically, this means that the entire inspection time becomes longer, and therefore, all products cannot be inspected within a practical time, and there has been a problem that a sampling inspection or a partial inspection has to be compromised. The large number of sampling points also requires a large-scale memory.

As an improvement measure, it is necessary to reduce the required memory amount, shorten the scanning time in the z direction, and shorten the scanning time in the xy direction.
As a method for shortening the z-direction scanning time of the multi-probe method, Nippon Kraut Kramer Co., Ltd. has proposed a method called volume focus. Rather than performing electronic scanning to focus on each element inside the sample, all multiprobes are excited only once and plane waves are transmitted. And the time series data of the reflected wave from a defect are acquired by it. Therefore, unlike the phased array, it is not affected by reverberation. By adding the amplitude for each position of the element in the time-series data of the reflected wave, the reflection intensity when the focus is on each position is virtually obtained. The higher the frequency and the larger the aperture, the higher the spatial resolution. This method is simple because the amplitude value of the measured waveform is used as it is. However, since the amount of calculation increases dramatically as the number of elements increases, a commercial machine with no high resolution has not been made. Since the distance from each probe to the defect is indistinguishable from the reflection from other elements having the same distance, there is a drawback that a ghost appears in the periphery.

There is another image reconstruction method using the inverse scattering method. It is assumed that the wave transmitted from one element is an SH cylindrical wave, and each transmits only once and scans in the xy plane. Wave time-series data is received by the same element for each transmission. What I want to find is the distribution of the reflection function in the sample. The reflection function is a Kirchhoff approximation describing the inside of the defect in terms of material mechanics or a Born approximation describing the inside of the defect. The spatial Fourier transform of the reflected wave is a scalar product of the spatial Fourier transform of the defect in the sample and the spatial Fourier transform of the probe directivity function, so the distribution of the defect in the sample can be obtained by measuring the reflected wave. be able to. In this case, since the directivity function of the probe is assumed to be a cylindrical wave that is easy to analyze, it can be calculated analytically. Therefore, even a small number of element divisions can be calculated with high accuracy, and the distribution of the reflection function can be obtained. The resolution in the xy direction is determined by the probe interval, and the resolution in the depth direction z is determined by the transmission wave frequency. The wave transmitted by a single element cannot always be approximated by a cylindrical wave, so when it cannot be approximated well, it is still controversial. This is why it is difficult to detect defects smaller than the probe interval.
In either case, as described above, there is a problem in that the resolution cannot be increased because the frequency cannot be increased with the multi-probe. To reduce the scanning time in the xy direction, a small-scale multi-probe is required. However, since the multi-probe becomes heavy, it is not suitable for high-speed mechanical scanning, which also causes a longer scanning time.
JP 2003-121502 A

An object of the present invention is to provide an image reconstruction device that reduces the number of probe scans at high resolution and shortens the time for image processing without extremely increasing the storage capacity.

In the present invention, in order to achieve the above object, the following device is devised to perform measurement within a practical time in the production line.
In order to realize the resolution of Nx, Ny, Nz (where N is the number of divisions of the sample) when the resolution in the x, y, z direction is an arbitrary resolution,
Using a wide directional probe (this directivity is numerically calculated and described with high accuracy according to the shape of the probe tip, not a cylindrical wave), and ((Nx × Ny) / (nx × ny)) locations in the xy direction , And Nt = nx × ny × nz pieces of time-series information of reflected waves are received at each point. Here, n is a number indicating how many samples are skipped. This time series information includes position information of defects in the x, y, and z directions. In this case as well, the spatial Fourier transform of the reflected wave is a scalar product of the spatial Fourier transform of the defect in the sample and the spatial Fourier transform of the probe directivity function. Can be obtained. By calculating this, it is possible to obtain resolutions in the x direction Nx, the y direction Ny, and the z direction Nz. The resolution in the xyz direction is determined by the product of the frequency and the sampling distance period.

The image reconstruction method using the inverse scattering method uses cylindrical wave approximation with relatively high directivity, and does not extract much information in the x, y, and z directions, and aims to detect defects far from the probe and larger than the probe. Unlike the above, in the image reconstruction apparatus of the present invention, the directivity function depending on the probe shape is calculated with high accuracy, and by using this, information in the xy direction can be extracted from the reflected waveform, and defects smaller than the probe can be extracted. The resolution can be increased so that it can be detected.
As described above, the main feature is that the number of sampling points / the number of scanning lines can be reduced and the measurement time can be shortened.
Specifically, the image reconstruction device is an image reconstruction device comprising a wide directivity probe and an arithmetic device incorporating an image processing method for processing the output signal of the probe to produce three-dimensional image information. The image processing method applies the Fourier transform corresponding to the thinning sampling to the time-series output signal obtained by the thinning sampling, and supports the thinning sampling on the signal obtained by the Fourier transformation and the directivity function of the probe. The Fourier transform data of the reflection function corresponding to the thinning sampling is obtained from the Fourier-transformed signal, and the Fourier transform data of the reflection function is inverse Fourier transformed to obtain the reflection function in real space. It is characterized by that.

  The present invention can reduce the number of probe scans at a high resolution and shorten the image processing time without extremely increasing the storage capacity. Further, it is not necessary to use a small-directed small probe to increase the resolution in the xy direction, and the number of sampling points / number of scanning lines can be reduced, so that the measurement time can be shortened and the required memory can be reduced.

The purpose of obtaining images with the same resolution as the number of scans in the xy direction with a smaller number of scans in the xy direction is included in the time-series information using the unidirectional probe to scan in the xy direction and reconstruct a 3D image. This was realized by using information in the xyz direction.

FIG. 1 is a conceptual diagram of one embodiment of the system of the present invention.
It consists of a scanning probe, transmitter / receiver and amplifier, xy scanner, FA computer for receiving / transmitting control, transmitting scanner control and AD conversion of received waveform, analysis computer for displaying a three-dimensional composite image, and parallel FFT computer. The difference from the conventional measurement system is that it has a wide directivity probe and a computer that performs parallel FFT and the like.

The xy scan of the probe can be performed at a larger sampling distance interval than the unidirectional probe as shown in FIG. 2, and the obtained time-series waveform is subjected to FFT conversion to synthesize an image with a resolution less than the sampling distance interval. can do.
Since the subject of the present invention is the use of a wide directional probe and its signal processing, description of the transmitter / receiver, the xy scanner, and the FA computer is omitted.

A sound wave transmitted at time t from a local coordinate x0 (vector display) = (x0, y0, 0) on the probe at a coordinate X (vector display) = (X, Y, 0) on the sample is represented by f (t, x0 (vector display)). Then, the wave v (t, x1 (vector display)) felt at local coordinates x1 (vector display) = (x1, y1, z1) in the sample is integrated over the entire probe regardless of x (vector display). The wave u (t, x2 (vector display), x (vector display)) felt at the local coordinates x2 (vector display) = (x2, y2, 0) on the probe as shown in (1) is the wave v (t , X1 (vector display)) is multiplied by the reflection function ρ (x (vector display) + x1 (vector display)) at that position, and integrated over x1, y1, and z1 over the entire interior of the sample. It becomes like this.
Since the probe output w (t, x (vector display)) is obtained by integrating the wave u (t, x2 (vector display), x (vector display)) with x2 and y2 over the entire probe, it is applied as shown in Equation (3). .
The directivity function h (t, x1 (vector display)) of the probe is applied as shown in Equation (4).

In the above formula, α is the attenuation coefficient, and c is the phase velocity of the wave.

Assuming that a one-dimensional convolution integral equation (5) of the integration interval L is defined, and X _{i and} x _j are m equal and lm equal points of the product X _i divided interval L, respectively, the discrete value form is It becomes like (6). When both sides are Fourier-transformed in the section L, the following equation (7) is obtained. Here, the Fourier transforms B _i (S _h ) and C _i (S _h ) of the functions b (x) and c (x) are defined as in equations (8) and (9).

The modified equation (7) is extended to a two-dimensional problem and applied to equation (3). Coordinates _X ix sampling points of the sample _surface, the _{Y iy} of xy direction _N x / _{n x} equal _points, and N y / _{n y} equal points. Further, when the coordinates X _jx , Y _{jy, and} Z _jz in the sample are Nx equality points, Ny equality points, and Nz equality points in the xyz direction, Expression (3) is expressed as Expression (10). When Expression (10) is Fourier transformed with respect to X and Y, Expression (11) is obtained. The time _tit is an n _{x ny} N _z equal dividing point. FIG. 4 shows the relationship between the equations (11), (12), (13), and (14). Therefore, the distribution ρ (x _ix , y _iy, Z _iz ) of the reflection function can be obtained from the probe output waveform w ( _tit, X _ix , Y _iy ).

FIG. 5 is a central section (xz section and yz section) of a sample having a rectangular parallelepiped defect. FIG. 6 shows time-series data when the probe is brought into close contact with the left end surface of the sample and scanned along the x-axis, and time-series data when scanned along the y-axis. The vertical axis is the scanning direction, and the horizontal axis is the time axis. It can be seen that the shape of the defect is greatly blurred due to the finite size of the probe and the coarse sampling interval. FIG. 7 shows the defect shape calculated by the method proposed in the present invention. You can see that it is well reproduced.

  It is desirable that electrodes of liquid crystal displays, ceramics XY stages for semiconductor production, etc. have no defects in large members of 1 m or more. To detect a defect of 0.1 mm, it is necessary to scan 10,000 lines or more, and if it is 1 second or more / line, it takes 10,000 seconds or more and cannot be used for inspection on a production line. By applying the present invention to a single probe or a multi-probe, the number of sampling points can be thinned out, so that the scanning time occupying the inspection time can be greatly shortened.

It is explanatory drawing of a three-dimensional image composition system. It is explanatory drawing in the case of sampling with a unidirectional probe and the case of sampling with a wide directional probe. It is explanatory drawing of the positional relationship of a sample and a probe, and how to take a coordinate. It is a relationship diagram between the probe output, the reflection function distribution due to the sample defect, and the flow of calculation. It is an xz cross section and a yz cross section of a rectangular parallelepiped sample having defects It is a time-series data display when a wide directional probe is scanned in the x direction and the y direction on the left end surface. FIG. 7 shows a distribution of reflection functions due to defects calculated from FIG. 6 using the proposed method of the present invention.

Claims (1)

An image reconstruction device comprising a wide directivity probe and an arithmetic device incorporating an image processing method for processing output signals of the probe to produce three-dimensional image information, wherein the image processing method is obtained by thinning sampling. The time-series output signal is subjected to Fourier transform corresponding to decimation sampling, and the signal obtained by the Fourier transform and the signal subjected to Fourier transform corresponding to decimation sampling to the directivity function of the probe are decimation. An image reconstruction apparatus comprising a procedure for obtaining Fourier-transformed data of a reflection function corresponding to sampling and obtaining a reflection function in real space by performing inverse Fourier transform on the Fourier-transformed data of the reflection function.