JPH10293848A - Parallel operation processor for image data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To check the faults and their positions with high accuracy and at a high speed by obtaining the continuous decision processing results of continuous scanning line groups distributed to plural decision processing boards and gathering those continuous processing results into a global memory of every decision processing board as the plane decision processing results. SOLUTION: Each of decision processing boards 201 to 206 performs the line decision processes of 312 times and also transmits and receives the data of the prescribed length to and from other boards in order to turn the discontinuous scanning line groups into the continuous ones. Finally, the area decision processing results of 1st to 1248-th lines, 1249-th to 2496-th lines, 2497-th to 3744-th lines, 3745-th to 4992-th lines, 4993-th to 6240-th lines and 6241-th to 7488-th lines are gathered on the boards 201, 202, 203, 204, 205 and 206 respectively. As a result, the discontinuous scanning line groups (four scanning lines) are turned into the continuous ones and accordingly a plane decision area having the planar decision processing result of a fixed area is formed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an observation method for inspecting the presence or absence of a defect at an observation point based on a plurality of luminance values obtained by imaging reflected light from the observation point on the surface of an object such as a hard disk. The present invention relates to a parallel arithmetic processing apparatus for image data for determining the presence or absence of a defect on the surface of an object by sequentially inputting measurement data for one scan line including points to individual processors in parallel.

[0002]

2. Description of the Related Art FIG. 10 is a diagram schematically showing a two-dimensional luminance distribution of reflected light as a light source distribution function when light is irradiated from a light source to an observation point on an object surface as a defect inspection target. In FIG. 10, reference numeral 10 denotes a light source, reference numeral 20 denotes an inspection medium such as a hard disk, reference numeral 21 denotes an observation point, and reference numerals 31 to 35 each denote a camera for imaging reflected light from the observation point 21. Here, the center camera 33 of the cameras 31 to 35 is arranged at a position for receiving the specularly reflected light from the observation point 21, and the cameras 32 and 34 on both sides and the outermost cameras 31 and 3.
5 are line-symmetrically arranged at positions separated by the same angle.

[0003] When a two-dimensional light source distribution is considered, luminance values b ₁ to b measured by the cameras 31 to 35 are considered.
₅ is represented by the light source distribution function f (θ _s ) in Equation 1.

[0004]

F (θ _s ) = Aexp {(2θ _n −θ _s ) ² / σ ² }

In the above equation 1, θ _s is the angle between the optical axis from the light source 10 and the optical axis of the regular reflected light. Next, each parameter in Equation 1 will be described with reference to FIG. In FIG. 11, f (θ _s ) is a light source distribution function serving as a reference when the observation point 21 is assumed to be a mirror surface, and is ± 1 [de] around the position of the central camera 33.
In this example, the luminance values b _{1 to} b ₅ are measured by disposing the cameras 34 and 32 at the position [g] and disposing the cameras 35 and 31 at the positions ± 3 [deg].

In Equation 1, A is a parameter indicating the reflectance of the observation point 21 and indicates the peak value (peak value of luminance) of the light source distribution function, and is influenced by the damage of the observation point 21 or the like. Further, θ _n is a parameter indicating the inclination of the observation point 21 and indicates the amount of deviation of the peak value from the reference light source distribution function f (θ _s ), and σ is the surface roughness of the observation point 21 such as roughness. This is a parameter indicating the degree of spread of the skirt from the reference light source distribution function f (θ _s ).

Therefore, by determining the parameters A, σ, and θ _n for each observation point using an undetermined coefficient determination method or the like, it is possible to inspect the observation point for defects such as scratches and roughness. The actual procedure is as follows. The inspection medium 20 such as a hard disk as shown in FIG. 12A is scanned along a scanning line SL in the radial direction, and for example, one line for a parameter A corresponding to the reflectance is scanned. Figure 1
If it is as shown in FIG. 2 (b), a range exceeding a predetermined threshold value T + or T- on the positive side or the negative side with respect to the reference value in the case where there is no defect is determined to have some defect. . Although not shown, the parameter σ changes mainly on the + side when there is a defect. θ
Regarding _n , since the correlation between the type of defect and the change on the + side or the − side is not uniformly determined, a case where there is a change on either the + side or the − side may be determined to be some kind of defect. .

[0008]

BRIEF Problems to be Solved] Here, to check the presence or absence of a defect on that scan line based on the luminance values b ₁ ~b ₅ measured by the five cameras 31 to 35 per scan line in the The measurement data of each camera 31-35 is DSP
(Digital signal processor) etc., it is necessary to perform processing in parallel, and by arranging the results of determining the presence or absence of defects for one scanning line continuously for many lines, the presence or absence of various defects having a certain area Can be detected. At present, when performing parallel processing of image data and alignment of image data for a large number of lines as described above, predetermined data can be efficiently transferred in a short time, and the load on each processor does not become excessive. Such a processing device has not been provided, and its realization has been desired. Therefore, the present invention has been made to solve these problems.

[0009]

According to a first aspect of the present invention, there is provided an image processing apparatus, comprising: a plurality of brightness values obtained by imaging reflected light from an observation point on an object surface; In order to check for the presence or absence of a defect, parallel processing of image data for determining whether or not there is a defect for each scan line by sequentially inputting measurement data for one scan line including the observation point to individual processors in parallel. In the apparatus, a plurality of determination processing boards each having a global memory and a plurality of processors accessible to the global memory are formed, and the global memories of the respective determination processing boards are connected in a ring shape. The measurement data of a plurality of continuous scanning lines (hereinafter referred to as “continuous scanning line group”) in the same number as the number is sequentially and supported to each processor in each determination processing board. Enter the click per successive scan lines groups of the defect existence determination processing performs respectively,
The determination processing results are stored in each global memory, and the determination processing results of the discontinuous continuous scanning lines stored in the respective global memories are transferred between the respective determination processing boards. , For storing the results of the determination process for a plurality of continuous scanning line groups.

According to a second aspect of the present invention, in the parallel processing device according to the first aspect, the specific processor of each determination processing board counts the number of executions of the process of determining whether or not there is a defect in the continuous scan line group, The operation mode of the own board is selected according to the count value.

[0011] The invention according to claim 3 is the invention according to claim 1 or 2.
In the parallel processing device described, each time each determination processing board finishes transferring the determination processing result of the continuous scanning line group, the determination processing board of the transfer source transfers the fixed pattern data, and the determination processing board of the transfer destination is The data transfer error is detected by comparing the received fixed pattern data with the reference data.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of this embodiment. In the figure, reference numerals 31 to 35 denote cameras for imaging reflected light from the observation point 21 of the inspection medium 20 as described above. Image data captured by these cameras 31 to 35 is input to the preprocessing boards 101 to 105. ing. The pre-processing boards 101 to 105 include AD converters ADC1 to ADC5, digital signal processors (DSPs) P1 to P5, and memories M1 to M5, respectively, and have outputs of six channels CP0 to CP5, respectively.

On the other hand, 201 to 206 are judgment processing boards. Each of the boards 201 to 206 has four digital signal processors and a global memory accessible by these processors. That is, the board 201 stores the processors Q11 to Q14 and the global memory GM1, and the board 202 stores the processors Q2
1 to Q24 and the global memory GM2
3 is a processor Q31 to Q34 and a global memory G
M3, the board 204 includes the processors Q41 to Q44 and the global memory GM4, the board 205 includes the processors Q51 to Q54 and the global memory GM5, and the board 206 includes the processors Q61 to Q64 and the global memory GM6.

Each of the judgment processing boards 201 to 206 has eight channels (communication ports) CN1.
To CN8, of which channel CN
The measurement data from the pre-processing boards 101 to 105 is sent to 1, CN2, CN4, CN5, and CN6 every four scan lines. The global memories GM1 to GM6 of the determination processing boards 201 to 206 are connected to each other by channels CN7 and CN8, whereby the global memories GM1 to GM6 are connected in a ring. Note that the channel CN3 is not used. Here, the processors P1 to P5, Q
For example, “TMS320C40” which is a DSP of Texas Instruments can be used for 11 to Q64. In FIG. 1, DI / DO is a digital input / output unit, MP is a main processor that controls the entire circuit, and MS is a storage device.

FIG. 2 shows one judgment processing board, for example, 2
1 shows a connection state between the processors Q11 to Q14 in FIG. 01, and the other determination processing boards 202 to 206 have the same connection configuration. FIG. 3 shows all the judgment processing boards 201 to 206 (global memories GM1 to GM
6) shows a state where they are connected in a ring shape.

Here, each parameter A,
An example of a processing procedure for obtaining σ, θ _n will be described with reference to FIG. First, based on the reference light source distribution function f (θ _s ) as shown in FIG. 10, luminance values b _{1 to} b ₅ corresponding to the imaging positions of the cameras 31 to 35 are obtained (S1).

[0017] Next, each of the parameters A, σ, per θ _n,
Performing a predetermined weighting to each luminance value b ₁ ~b ₅ (S2). This weighting is not always necessary in the present invention, but is a process required to accurately grasp the state of change according to the meaning of each parameter. FIG. 5 shows an example of the above-mentioned weighting. For example, the parameter A relating to the peak value of the light source distribution function places the highest priority on b _3, and the parameter σ indicating the degree of expansion of the tail of the light source distribution function is b ₁ , the most emphasis on b _5,
Further, b _{2 to} b ₄ are equally emphasized for the parameter θ _n indicating the shift of the peak value of the light source distribution function.

Next, first, while changing the parameter A as a target variable within a predetermined range (A _min ≦ A ≦ A _max ),
Approximate function for parameter A, ie, regression model f _{A
(B 1, b 2, b} 3, b 4, b 5) to create a (S3). The approximation function f _A (b ₁ , b ₂ , b ₃ , b ₄ , b ₅ ) is represented by, for example, Expression 2.

[0019]

[Number 2] _{f A (b 1, b 2} , b 3, b 4, b 5) = a 0 + a 11 b 1 + a 12 b 1 2 + a 13 b 1 3 + a 14 b 1 4 + a 15 b 1 5 + a _{21 b 2 + a 22 b 2} 2 + a 23 b 2 3 + a 24 b 2 4 + a 25 b 2 5 + a 31 b 3 + a 32 b 3 2 + a 33 b 3 3 + a 34 b 3 4 + a 35 b 3 5 + a 41 b ^{_{4 + a 42 b 4 2 +}} a 43 b 4 3 + a 44 b 4 4 + a 45 b 4 5 + a 51 b 5 + a 52 b 5 2 + a 53 b 5 3 + a 54 b 5 4 + a 55 b 5 5

Here, for each of the objective variables A within the above range, the values of the corresponding explanatory variables b _{1 to} b ₅ can be calculated based on the reference light source distribution function f (θ _s ).
26 regression coefficients a ₀ , a for each objective variable A
_11, a _12, ......, the approximation function f _A (b ₁ having a _54, a _55,
b ₂ , b ₃ , b ₄ , b ₅ ) can be created in large numbers. Next, these regression coefficients a ₀ , a ₁₁ , a ₁₂ ,..., A ₅₄ ,
_a55 is calculated by regression analysis (S4). In particular,
Each objective variable A = f _A (b ₁ , b ₂ , b ₃ , b ₄ ,
b ₅ ) and regression coefficients a ₀ , a ₁₁ , a
₁₂ ,..., A ₅₄ and a ₅₅ are obtained.

Hereinafter, similarly, the parameters σ and θ
_nWith the approximation function fσ (b₁, B_Two,
b_Three, B_Four, B_Five), Fθ_n(B₁, B_Two, B_Three, B_Four, B_Five)
And the regression coefficient h for each₀, H₁₁, H₁₂,…
…, H₅₄, H₅₅And p₀, P₁₁, P₁₂, ……, p₅₄, P
₅₅(S5 to S8). (The signs h,
p is for convenience. Here, the approximate function fσ (b₁, B_Two, B_Three, B_Four, B_Five)
And fθ_n(B₁, B_Two, B _Three, B_Four, B_Five) Is easy from Equation 1
Therefore, the description is omitted.

In the above embodiment, the brightness values b _{1 to} b ₅ of the cameras 31 to 35 arranged at five locations are used as explanatory variables. However, the number of brightness values (in other words, the number of cameras) may be arbitrary. .

As described above, the regression coefficients a ₀ , a ₁₁ , a
₁₂ ,..., A ₅₄ , a ₅₅ and h ₀ , h ₁₁ , h ₁₂ ,.
_54, h _55, and _{p 0, p 11, p 12} , ......, if p _54, p ₅₅ is calculated, the approximation function _{f A (b 1, b 2} , b 3, b 4,
_{b 5), fσ (b 1} , b 2, b 3, b 4, b 5), fθ
_{n (b 1, b 2,} b 3, b 4, b 5) the by substituting the luminance values b ₁ ~b ₅ arbitrary observation point, the parameter A, sigma, to obtain the theta _n at the observation point Can be.

For this reason, the pre-processing board 10 shown in FIG.
Processor P1~P5 in 1-105 measures the luminance values b ₁ ~b ₅ for each observation point on the scan line, and sends the measurement data to the determination processing board 201-206.
Processors Q11 to Q11 in the judgment processing boards 201 to 206
Q64 is pretreated boards luminance values b ₁ sent from 101 to 105 ~b ₅ parameters A for each scan line using,
σ, θ _n are determined, and the presence or absence of a defect on each scan line is determined by threshold processing, and data for a plurality of continuous scan lines is determined by the determination processing boards 201 to 201 as described later.
By transferring and arranging the defects between the pixels 206, the presence / absence and type of a defect having a certain area are also determined (surface determination processing).

FIG. 6 shows a surface determination area (scan line N) of each of the determination processing boards 201 to 206 in this embodiment.
O. FIG. 7 is a diagram illustrating the number of times of performing the line determination process.
Now, assuming that the total number of scanning lines is 7488 when inspecting the entire surface of the inspection medium 20 as shown in FIG.
Eighth, judgment processing board 202 has 1249 th to 249 th
Sixth, each board 201-206 is 1
A surface determination area to be shared by / 6 is determined in advance.

As shown in FIG. 1, the first determination processing board 201 has luminance values b of the first to fourth continuous scanning lines (referred to as a continuous scanning line group) from the preprocessing boards 101 to 105 first. ₁ ~b ₅ is sent, the interior of the processor Q11~Q14 are each one by in charge of scan lines each scan line per parameter a, sigma, calculation of the theta _n,
Threshold processing or the like is performed to determine the presence or absence of a defect on the first to fourth scan lines.

Similarly, the second judgment processing board 20
The 2, pretreatment luminance values b ₁ ~b ₅ of the first five to eighth -th continuous scan line group from the board 101 to 105 is transmitted, inside the processor Q21~Q24 is charge of each one by one scan line Then, the presence or absence of a defect on each of the fifth to eighth scan lines is determined. The same applies to each board 203 to
206 performs processing, and the sixth determination processing board 206 determines the presence or absence of a defect on each of the 21st to 24th scan lines. In the next operation, the first judgment processing board 201
, The 25th to 28th scanning lines, the second determination processing board 202 performs the 29th to 32nd scanning lines,.

That is, each of the judgment processing boards 201 to 20
No. 6 performs the determination processing of four scanning lines in parallel at a time, so that the determination processing of the first to twenty-fourth scanning lines is executed by one operation. Then, as shown in FIG. 6, in the second operation, the judgment processing board 201
Eighth, the determination processing board 202 is the 29th to 32nd, ...
.., The determination processing board 206 performs determination processing on the 45th to 48th scan lines, and the 25th to 4th
The processing for determining up to the eighth scan line is executed. That is, each of the determination processing boards 201 to 206 sequentially determines the presence / absence of a defect on four scanning lines cyclically, and this line determination processing is performed at 7488/2 lines.
4 = 312 times.

The determination results by the four processors in each of the determination processing boards 201 to 206 are determined as fixed-length data of 1000 bytes / line in each of the global memories GM1 to GM.
The data is output to M6 and stored in an area accessible by the processors Q14, Q24, Q34, Q44, Q54, and Q64 of the boards 201 to 206. The number of executions of the above-described line determination processing is determined by the processors Q14, Q24, Q34, Q44, Q54, Q6 of the boards 201 to 206.
4 and are cleared to zero by an initialization command from the main processor MP.

As is clear from FIG. 6, each board 201
In the global memories GM1 to GM6, 312 sets of determination processing results in which four continuous scan line groups are set as a set are stored in a discontinuous state. For example, the determination result of the first to fourth, 25th to 28th,..., 7465th to 7468 continuous scan line groups is stored in the global memory GM1 of the determination processing board 201. Have been.

In general, the defect on the surface of the inspection medium 20 has a certain area, and therefore, the determination processing when a surface is formed by integrating a large number of scanning lines is important. In other words, in order to make the determination processing result of the intermittent continuous scanning line group stored in each of the global memories GM1 to GM6 into the determination processing result for a predetermined area, the determination between the global memories GM1 to GM6, in other words, each board 2
It is necessary to perform data transfer between 01 and 206 to align the data, and to make the continuous scanning line group in a discontinuous state continuous.

The data transfer between the determination processing boards 201 to 206 for making the continuous scanning line groups in the discontinuous state continuous will be described below. The transfer data is obtained by assuming that the determination processing result for the four continuous scanning line groups is one unit.
This is a determination process result for up to 20 scan lines (a group of continuous scan lines of 1 to 5 units), and is distinguished according to the number of executions (1 to 312) in FIG. According to FIG. 9 described later), fixed-length data transfer is performed for each board. FIG. 7 shows a format of the transfer data, and shows a transmission buffer and a reception buffer of the determination processing board 206 as an example. The other boards 201 to 205 have the same configuration.

The data format of the judgment processing board 206 will be described below. Up to +4000 bytes from the head of the transmission buffer of the board 206, the determination processing result of the board 206 is stored. At this time, the processor Q64 corresponding to the fourth scan line
Is stored in the processors Q63, Q62, and Q61 in the descending order of the scanning lines. The reception buffer is a continuous 20,000-byte area starting at +4000 bytes of the transmission buffer, and this 20,000 bytes is used as a storage area for data received from the other boards 201 to 205. In normal transfer processing, after data is received in the reception buffer, its own data area (+0 to +40
00 bytes), and transfers the data together with the received data to the next board (board 201 in this example).

FIG. 8 shows each of the judgment processing boards 201 to 206.
6 is a time chart showing the operation of the first embodiment. Here, data (determination processing result) is transmitted from the board 201 and the board 2
02 is indicated by “COM1-2” in the transfer process for receiving data. For example the determination processing board 201 first performs a process of calculating the parameters A, sigma, theta _n indicating the surface condition described above. After that, after issuing a flag “COM IN” indicating that the data reception preparation is completed, the variance value calculation of each parameter and the presence / absence determination processing are performed, and during that time, the data transmitted from the board 206 is received.
When the processing is completed, the board 201 performs data alignment processing, and sets a flag “COM” indicating that transmission preparation is completed.
OUT ”, the data is transferred to the board 202. Of course, at this point, the flag “COM IN” by the board 202 already exists. The determination processing board 201 repeatedly executes the above operation at a constant cycle. Hereinafter, the other determination processing boards 202-2
06 repeats the same operation sequentially at predetermined time intervals (for example, 4 ms).

In the data transfer between the boards 201 to 206, the transfer source board transfers the fixed pattern data, and the transfer destination board compares the received fixed pattern data with the reference data to transfer the data. It is desirable to detect errors.

Next, FIG. 9 shows a phase according to the number of times of execution of the line determination processing in each of the boards 201 to 206, the operation mode (mode 0 to mode 17) of the transmitting side board, and the transfer data length (transferred). (The number of scanning lines). Here, the contents of each operation mode are as follows. In the following description, the number of scanning lines is referred to as “(number) lines”.

Mode 0: Reception 0 line, Transmission 0 line This mode is a special process of the judgment processing board 201 at the time of activation, and has neither reception data nor transmission data. It is only necessary to write the data subjected to the line determination processing into the aligned data storage area of the global memory GM1.

Mode 1: 20 lines for reception, 0 lines for transmission This mode is a normal aligned data storage mode. All the data received from the other board and its own data are written to the alignment data storage area of the global memory. As described above, 20 lines of data are stored in the receiving buffer in the reverse order of the scanning line number. The self data is the following four lines. For example, the judgment processing board 20
In 1, the line number +23 is the self data by the processor Q14, +22 is the self data by the same Q13, +21 is the self data by the same Q12, +20 is the self data by the same Q11, +19 is the head data of the reception buffer,. As the last data of the reception buffer.

Mode 2: 0 lines for reception, 4 lines for transmission This is a normal data transfer mode. However, there is no received data. Transfers four lines of its own data.

Mode 3: reception 4 lines, transmission 8 lines Mode 4: reception 8 lines, transmission 12 lines Mode 5: reception 12 lines, transmission 16 lines

Mode 6: reception 16 lines, transmission 20 lines This is a normal data transfer mode. Four lines of self-data are added to the received data and transferred.

Mode 7: 20 lines for reception, 4 lines for transmission This is the termination processing of the sorted data storage mode in the judgment processing board 201. 20 lines of received data are written in the alignment data storage area, and 4 lines of self-data are transmitted.

Mode 8: 4 lines for reception, 0 lines for transmission This is the start processing of the sorting data storage mode in the judgment processing board 202. Four lines of received data and four lines of self-data are written in the alignment data storage area.

Mode 9: reception 20 lines, transmission 8 lines This is the termination processing of the sorted data storage mode in the judgment processing board 202. Trailing 1 out of 20 lines of received data
Six lines (the first +0 line to the +15 line) are written in the alignment data storage area, and four lines of self-data are added to the remaining four lines of received data and transmitted.

Mode 10: reception 8 lines, transmission 0 lines This is the start processing of the sorting data storage mode in the judgment processing board 203. Write eight lines of received data and four lines of self-data to the alignment data storage area.

Mode 11: 20 lines for reception, 12 lines for transmission This is termination processing of the sorted data storage mode in the judgment processing board 203. Trailing 1 out of 20 lines of received data
Two lines (the first +0 line to the +11 line) are written in the alignment data storage area, and the remaining eight lines of received data are added with four lines of self-data and transmitted.

Mode 12: reception 12 lines, transmission 0 line This is the start processing of the alignment data storage mode in the judgment processing board 204. 12 lines of received data and 4 own data
Write the line to the alignment data storage area.

Mode 13: 20 lines for reception, 16 lines for transmission This is termination processing of the sorted data storage mode in the determination processing board 204. Trailing 8 out of 20 lines of received data
The line (the first +0 line to the +7 line) is written in the alignment data storage area, and the remaining 12 lines of received data are added with 4 lines of self-data and transmitted.

Mode 14: reception 16 lines, transmission 0 line This is the start processing of the sorted data storage mode in the judgment processing board 205. 16 lines of received data and self data 4
Write the line to the alignment data storage area.

Mode 15: 20 lines for reception, 20 lines for transmission This is termination processing of the sorted data storage mode in the determination processing board 205. Trailing 4 out of 20 lines of received data
The line (the first +0 line to the +3 line) is written in the aligned data storage area, and the remaining 16 lines of received data are added with 4 lines of self-data and transmitted.

Mode 16: reception 20 lines, transmission 0 line This is the start processing of the alignment data storage mode in the judgment processing board 206. 20 lines of received data and 4 self-data
Write the line to the alignment data storage area. The operation is the same as in mode 1.

Alignment Data Overlapping Process In phase 13 (mode 17), after the end of phase 12, 16 lines of data between the alignment data stored in each board are taken into account in consideration of the boundary area of the surface determination processing. To overlap. Mode 17: 16 lines for reception, 16 lines for transmission After transmitting the last 16 lines of the alignment data area of each board (only for activation), reception of 16 overlapping lines is performed.
Data transmission and reception are performed directly from the aligned data area to the aligned data area. The received data is added to the head of the aligned data area. After confirming the reception of the data, the data sorting process ends.

As described above, each judgment processing board 201
206 transmit and receive data of a predetermined length between the boards while performing the line determination process 312 times, so that the first to the 1248th, The board 202 is 1249 th to 2496 th and the board 203 is 2497 th to 37 th.
The 44th board, the board 204 collects the 3745th to 4992th boards, the board 205 collects the 4993th to 6240th boards, and the board 206 collects the determination processing results of the 6241st to 7488th areas. As a result, a plurality of continuous scanning line groups (four scanning lines) that were in a non-continuous state are made continuous to form a surface determination region having a planar determination result of a certain area. It is possible to detect the presence or absence and the position of a planar defect that cannot be detected with high accuracy.

In the above embodiment, the case where the present invention is applied to the defect inspection of the surface of a magnetic recording medium such as a hard disk has been described. However, the present invention is also applicable to the defect inspection of the surface of another object.

[0055]

As described above, according to the first aspect of the present invention, the determination processing results of the continuous scanning line group distributed to the plurality of determination processing boards are made continuous, and the surface determination is performed in the global memory in each determination processing board. In addition to collecting processing results and transferring data, as described in claim 2, the processor in each determination processing board counts the number of times of execution of the line determination processing and selects an operation mode according to the value. Therefore, the presence or absence of various defects having a certain area and their positions can be inspected with high accuracy and high speed. Further, as described in claim 3, each determination processing board compares the fixed pattern data with the reference data, so that a data transfer error can be reliably detected.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 2 is a diagram illustrating an internal configuration of a determination processing board according to the embodiment.

FIG. 3 is a diagram illustrating a connection state of a determination processing board in the embodiment.

FIG. 4 is a flowchart of a processing procedure for obtaining each parameter of a light source distribution function in the embodiment.

FIG. 5 is a diagram illustrating weighting of each parameter of a light source distribution function in the embodiment.

FIG. 6 is a diagram illustrating a surface determination area of each determination processing board and the number of executions of a line determination process according to the embodiment.

FIG. 7 is an explanatory diagram of a format of transfer data in the embodiment.

FIG. 8 is a time chart illustrating an operation of each determination processing board in the embodiment.

FIG. 9 is a diagram showing a phase, the number of executions, an operation mode, and a transfer data length in the embodiment.

FIG. 10 is an explanatory diagram of a light source distribution function.

FIG. 11 is an explanatory diagram of each parameter.

FIG. 12 is a diagram showing a scan line (FIG. 12A) on an inspection medium and parameter values (FIG. 12B) on the scan line.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 Light source 20 Inspection medium 21 Observation point 31-35 Camera 101-105 Pre-processing board 201-206 Judgment processing board ADC1-ADC5 AD converter P1-P5, Q11-Q64 Digital signal processor (DSP) M1-M5 Memory GM1-GM6 Global Memory DI / DO Digital input / output unit MP Main processor MS Storage device SL Scan line

Claims (3)

[Claims] An inspection method for inspecting the presence or absence of a defect at the observation point based on a plurality of luminance values obtained by imaging reflected light from the observation point on the surface of the object. An image data parallel processing device for inputting measurement data sequentially to individual processors in order to determine the presence or absence of a defect for each scanning line, comprising a global memory and a plurality of processors capable of accessing the global memory A plurality of determination processing boards are formed, global memories of the respective determination processing boards are connected in a ring shape, and a plurality of continuous scanning lines (hereinafter referred to as a continuous scanning line group) having the same number as the number of processors constituting the determination processing boards. Is sequentially and cyclically input to each processor in each judgment processing board to judge whether or not there is a defect in the continuous scanning line group. By performing each of the determination processing results, storing the determination processing results in each global memory, and transferring the determination processing results of the discontinuous continuous scanning line groups stored in the respective global memories between the respective determination processing boards, A parallel arithmetic processing device for image data, characterized in that a determination result of a plurality of continuous scanning line groups is stored in each global memory. 2. The parallel processing apparatus for image data according to claim 1, wherein the specific processor of each of the determination processing boards counts the number of executions of the determination processing of the presence or absence of a defect with respect to the continuous scanning line group, and the count value is provided. A parallel operation processing device for image data, wherein an operation mode of the own board is selected according to the operation mode. 3. The parallel processing apparatus for image data according to claim 1, wherein each time each of the judgment processing boards finishes transferring the judgment processing result of the continuous scanning line group, the judgment processing board of the transfer source is fixed. A parallel processing apparatus for image data, wherein pattern data is transferred and a transfer processing determination board compares the received fixed pattern data with reference data to detect a data transfer error.