KR101521725B1 - Labeling method, labeling apparatus and defect inspection apparatus - Google Patents

Abstract

(Problem) The labeling process is speeded up.
The present invention relates to an initial label setting process for setting labels that are different from each other for each run, and a process of determining whether or not there is a first connection run to connect to the process target run, An update step of performing an individual update process of updating the label of one of the run subject to be processed and the label of the first connection run to the other label when there is a first connected run; And a search step of searching for runs connected to each other among the runs of the plurality of sets, and assigning the same labels. In the update step, a plurality of runs are divided into a plurality of sets, A label rewritten by one update thread in addition to allowing rewriting of the label by the update thread of one of the plurality of update threads under exclusive control While prohibiting rewriting by another update thread executes in parallel a plurality of update thread.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a labeling method, a labeling apparatus, and a defect inspection apparatus. 2. Description of the Related Art LABELING METHOD, LABELING APPARATUS AND DEFECT INSPECTION APPARATUS

The present invention relates to a labeling technique for assigning labels to a plurality of runs created by rendering normal image data run length, and a defect inspection apparatus using the labeling technique.

BACKGROUND ART [0002] In the field of manufacturing technologies for semiconductor substrates and printed boards, in order to detect defects included in products and to analyze and evaluate them, an evaluation object is picked up through a microscope or the like, and an image including defects, . Then, based on the binary image data of the defect image, the number, area, and center of the defect portion are measured. At this time, in order to perform the measurement automatically, labeling (connection area extraction) processing is widely used. For example, in Japanese Patent Application Laid-Open No. 2008-186123, a two-pass labeling process is applied to a run generated by making the normal image data run lengthwise for each row, thereby speeding up the labeling process.

However, the invention disclosed in Japanese Patent Laid-Open Publication No. 2008-186123 (hereinafter referred to as "the conventional invention") is based on the assumption that the run data is sequentially inspected from the upper run data, . Further, in the above-mentioned conventional invention, a label table for recording whether or not the run data assigned different labels belong to the same connected component, and a run count table for associating the number of run data corresponding to the label separately There is a need. As a result, the number of elements of the label table and the run number table increases due to an increase in the number of runs and a complicated connection state. This is another factor that hinders the speeding up of labeling processing.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is an object of the present invention to provide a technique capable of speeding up the labeling process of binary image data and a technique for effectively performing defect inspection using the above technique.

A labeling method according to the present invention is a labeling method for labeling a plurality of runs created by rendering run-length binary image data, the labeling method comprising: an initial label setting step of setting labels different from each other for each run; Determines whether or not there is a first connected run to be connected to the target run, and if there is a first connected run, sets one of the target run and the first connected run to be different And a search step of searching for a run connected to each other among a plurality of runs based on the plurality of labels and assigning the same label after the update step; and in the update step, , A plurality of runs are divided into a plurality of sets, an individual update process executed in each set is made an update thread, and a plurality of The rewriting of the label by the update thread of one of the update threads is permitted and the plurality of update threads are executed in parallel while prohibiting the rewriting by the other update thread of the label rewritten by one update thread .

A labeling apparatus according to the present invention is a labeling apparatus for performing labeling processing on a plurality of runs created by rendering normal image data run length, the labeling apparatus comprising: a storage unit for storing labels of each run; A processor unit having a plurality of labels; and a processor unit for initially setting a different value to each of a plurality of labels, and for each run, determines whether or not there is a first connection run to connect to the processing object run, , And when there is a first connection run, performs a renewal process of updating one of the labels of the object run and the first connection run to the other label, Wherein the plurality of runs are divided into a plurality of sets according to the number of processor cores, and each of the processor cores The control unit executes an individual update process as an update thread for one or a plurality of divided runs, and the control unit permits rewriting of the label by the update thread of one of the plurality of update threads under the exclusive control, The individual update processing for each run is performed by executing a plurality of update threads in parallel while prohibiting rewriting by another update thread of the label to be rewritten.

According to another aspect of the present invention, there is provided a defect inspection apparatus comprising: an image obtaining unit that obtains an image to be inspected; an image extracting unit that extracts a defect image including a defective area by inspecting an image to be inspected; A run generating section for generating a plurality of runs by normalizing the binary image data to generate run data; and a runner section having the same configuration as the above-described labeling machine, And labeling means for giving the labeling means.

In the invention thus constituted, labels having different values from each other are initially set. Then, the individual update process is executed for each run, and the label update is executed depending on whether or not there is a connection run to be connected to the run subject to be processed. Thereafter, the runs connected to each other among the plurality of runs are searched based on the plurality of labels, and the same label is given thereto. In the present invention, the labeling process of the binary image data is performed by the label of the same number as the number of runs, and in the conventional invention, The label table and the run number table which are used are unnecessary, and the labeling process can be speeded up.

In addition, in the updating step of the present invention, a plurality of runs are divided into a plurality of sets, and individual update processing executed in each set is made an update thread, and these update threads are executed in parallel. That is, the individual update processing for each run is executed in parallel. Therefore, although the label update is exclusively controlled between update threads, the total time required for the labeling process can be significantly shortened as compared with the conventional invention in which labels are sequentially given from the upper run.

Here, in the update step, the rewriting of the label rewritten by one update thread is inhibited by another update thread, but rewriting by another update thread may be allowed for a label other than the above-mentioned label, that is, . As a result, the processing efficiency can be increased and the time required for the labeling processing can be further shortened.

When the label of the first connection run is updated in the individual updating process, it is determined whether or not there is a second connection run other than the process target run connected to the first connection run. If there is a second connection run , And one of the first connection run and the second connection run may be updated to the other label. In this manner, not only the first connection run but also the second connection run to be connected to the processing object run through the first connection run are updated by labeling, so that the second connection executed later than the individual update processing for the processing object run The time required for the individual update processing for the run can be shortened. As a result, the total processing time can be shortened.

A run having the same label as the label before the update before the label of the first connection run may be used as the second connection run and the second connection run can be easily and reliably found by using the label before the update It is appropriate.

Alternatively, the binary image data may be run length-wise for each row. In this case, a run connected to the downward run of the run to be processed can be determined as the first connected run.

The individual search process may be performed for each run in the search process among the two passes (a process of searching for runs connected to each other among a plurality of runs and assigning the same label). This " individual search process " refers to a process in which one run is a target to be processed, a connection run to be connected to the target run is searched based on a plurality of labels, and the labels of the target run and connection run are rewritten Processing. As described above, the individual search process for each run makes it possible to perform the search process rationally. Further, in the same manner as the individual update processing for each run described above, a plurality of runs are divided into a plurality of sets, individual search processing executed in each set is made a search thread, and one search search The plurality of search threads may be executed in parallel while the rewriting of the label by the thread is permitted and the rewriting by the other search thread of the label rewritten by one search thread is prohibited. As a result, the speed of the labeling process can be further improved by increasing the processing speed in the searching step.

According to the present invention, a part of the process for executing the labeling process can be executed in parallel, and the total time required for the labeling process of the normal image data can be shortened.

BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a view showing a schematic configuration of an inspection system using a defect inspection apparatus equipped with an embodiment of a labeling apparatus according to the present invention; FIG.
2 is a block diagram showing a schematic configuration of an image processing unit.
3 is a block diagram showing a schematic structure of a labeling section according to an embodiment of the labeling apparatus according to the present invention.
4 is a flowchart showing a labeling operation by the labeling unit.
5 is a diagram showing an example of the normal image data given to the labeling section.
FIG. 6 is a diagram showing initial values of run data of a run obtained by executing run length processing on the ordinary image data of FIG. 5 and run connection data indicating a connection relationship among the runs.
7 is a flowchart showing individual update processing executed in each processor core.
8 is a schematic diagram showing an example of individual update processing executed in each processor core.
9 is a flowchart showing individual search processing executed in each processor core.
10 is a schematic diagram showing an example of individual search processing executed in each processor core.

BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a view showing a schematic configuration of an inspection system using a defect inspection apparatus equipped with an embodiment of a labeling apparatus according to the present invention; FIG. This inspection system 1 is an inspection system for inspecting defects such as pinholes and foreign objects appearing on the appearance of a semiconductor substrate (hereinafter referred to as " substrate ") S to be inspected. The inspection system 1 has an imaging device 2 for imaging an area to be inspected on the substrate S and a control device 3 for performing defect inspection on the basis of image data from the imaging device 2. [

If a defect is found on the substrate S in the substrate detection apparatus M provided on the production line of the substrate S separately from the present system, the position coordinates of the defect are given to the inspection system 1. [ The substrate detecting apparatus M placed in the manufacturing line inspects the entire substrate S by a predetermined processing algorithm and acquires and outputs the position coordinates of the substrate surface if any area satisfying the requirements as defects is present on the substrate surface. Therefore, the imaging section of the substrate detection apparatus M is relatively low in resolution, and the processing algorithm is also fixed.

On the other hand, the inspection system 1 is connected to the substrate detection device M via an interface (not shown), and the region where the position coordinates are reported as defects from the substrate detection device M is referred to as an imaging device The image is imaged by the imaging device 2 and the control device 3 performs precise irradiation of the image to determine the presence or absence of a defect and the type thereof in detail and displays an image of the defect part on the display unit.

The imaging apparatus 2 includes an imaging section 21 for acquiring image data by imaging an area to be inspected on the substrate S, a stage 22 for holding the substrate S, And a stage driving unit 23 for moving the stage 22 relatively. The imaging unit 21 includes an illumination unit 211 for emitting illumination light, an optical system 212 for guiding the illumination light to the substrate S and for introducing light from the substrate S, And an image pickup device 213 for converting an image of the substrate S formed by the light-receiving element into an electric signal. The stage driving part 23 is constituted by a ball screw, a guide rail and a motor and the device control part 4 provided in the control device 3 controls the stage driving part 23 and the image pickup part 21, The area to be inspected is picked up.

The control device 3 has a device control section 4. The device control section 4 executes the control program that has been read in advance, thereby operating each section of the control device shown in Fig. 1 as follows. The control device 3 includes an image acquisition unit 5 and an image processing unit 6 in addition to the device control unit 4. [ The image acquisition section 5 acquires image data corresponding to the sensed image by converting the electrical signal output from the sensing section 21 into data. The image processing section 6 performs appropriate image processing on the image data acquired by the image acquisition section 5 to create a defect included in the image and an image of the defective part (hereinafter referred to as " defective image ") . The image processing unit 6 includes a labeling unit as an embodiment of the labeling apparatus according to the present invention, and is capable of executing a labeling process on a defective image at a high speed. The configuration and operation of the image processing section 6, particularly the labeling section, will be described in detail later.

The control device 3 includes a storage unit 7 for storing various data, an input accepting unit 8 such as a keyboard and a mouse for accepting an operation input from the user, A display unit 9 for displaying time information, and the like. Although not shown, it is also possible to have a reading device that reads information from a computer-readable recording medium such as an optical disk, a magnetic disk, or a magneto-optical disk, and transmits and receives signals with other configurations of the inspection system 1 And the communication unit is connected appropriately via the interface 1 / F or the like.

2 is a block diagram showing a schematic configuration of an image processing unit. The image processing section 6 includes a filtering section 61, a difference extracting section 62, a binarization processing section 63, a labeling section 64 and an isolation removal processing section 65. In the filtering section 61, a captured image is sent from the image capturing section 5, and a reference image is sent from the storage section 7. Of these two images, the picked-up image is an image of the substrate S picked up by the image pickup device 2 and corresponds to the inspection object image to be subjected to the defect detection inspection. The reference image is an image corresponding to an ideal substrate free from defects. In this embodiment, defect detection is performed from the inspection object image by comparing the inspection object image and the reference image, as described below. These defect images and reference images are stored in the storage unit 7 and referenced if necessary, but may be a form in which image data stored in an external storage medium is read as necessary.

The filtering unit 61 performs a filtering process for each of the inspection object image and the reference image to remove the difference between the image noise and the mild image not related to the defect and sends each image to the difference extracting unit 62 . The difference extracting unit 62 corresponds to an example of the "image extracting unit" of the present invention. The difference extracting unit 62 extracts a region where image contents differ from each other by obtaining the difference between the inspection target image and the reference image after the filtering process, And sends the image to the binarization processing unit 63. [ Then, the binarization processing section 63 binarizes the difference image by a suitable threshold value, generates the binary image data BFI, and sends it to the labeling section 64. [ The labeling unit 64 generates a plurality of runs by rendering the binary image data BFI run length, and assigns labels to the plurality of runs. In addition, the isolation removal processing unit 65 removes the residual isolation points from the labeled normal image. Thus, a defect position image to be obtained is generated, and the image processing section 6 is outputted.

3 is a block diagram showing a schematic structure of a labeling section corresponding to an embodiment of the labeling apparatus according to the present invention. The labeling unit 64 is provided with a GPU (Graphic Processing Unit) 642 having a plurality of processor cores 641. In this GPU 642, each processor core 641 functions as an individual update processing unit 643 and an individual search processing unit 644, and executes individual update processing and individual search processing for each run as one thread. In the present embodiment, the individual update process for each run corresponds to an example of the " update thread " of the present invention, and the individual search process for each run corresponds to an example of the " search thread " The individual update processing and individual search processing will be described later in detail.

The labeling unit 64 includes a labeling storage unit (not shown) for storing the binary image data BFI, data for specifying the run r [m] (hereinafter referred to as run data), and run connection data id [ 645). These data will also be described in detail later while showing specific examples.

The labeling section 64 is provided with an arithmetic processing section 646 composed of a CPU (Central Processing Unit), a memory, and the like. The arithmetic processing unit 646 includes a run generation unit 647 for generating a run r [m] by executing a logical operation on the binary image data BFI by a CPU in accordance with a predetermined program, A parallel processing control section 648 for executing threads executed in the processor core 641 in parallel while exclusion control and a data initial setting section 648 for initializing label numbers included in run r [m] and run connection data id [m] (649). In this way, the operation processing unit 646 functions as a " control unit "

Next, the labeling operation by the labeling unit 64 configured as above will be described with reference to Figs. 4 to 10. Fig. 4 is a flowchart showing a labeling operation by the labeling unit. When the binary image data BFI generated by the binarization processing section 63 is given to the labeling section 64, the arithmetic processing section 646 of the labeling section 64 controls each section of the labeling section 64 to execute labeling processing do. First, the binary image data BFI is temporarily registered in the labeling storage unit 645. [ Further, for the binary image data BFI, the run generation process is executed in the run generation unit 647 to generate a plurality of runs r [m] (step S1) and stored in the label storage unit 645 . In addition, many conventional techniques have been proposed for the run-length conversion process, and the general run length conversion process is used as it is in this embodiment as well. Therefore, the description of the run length conversion process is omitted.

5 is a diagram showing an example of the normal image data given to the labeling section. 6 is a diagram showing initial values of run data obtained by executing the run length process on the normal image data of FIG. 5 and run connection data indicating the connection relationship between the runs. The numerical values (0, 1, 2, ...) in the vertical and horizontal directions in FIGS. 5 and 6A indicate the coordinate positions in the row direction and the column direction of the pixels arranged in a matrix, the row index and the column index. In Fig. 5, the hatched pixel indicates that the pixel has pixel data of " 1 " while the pixel on which hatching is not performed indicates that the pixel has pixel data of " 0. " By adding run length processing for each row to the binary image data BFI having such pixel data, the run shown in Fig. 6 (a) is obtained. Specifically, since there is no pixel of "1" in the 0th row and the 4th row, no run is created, and 2 runs, 2 runs and 1 run are created in the 1st to 3rd rows . In this specification, the run number of each run created by the run-length process is denoted by "m", and 0, 1, 2, ... , And the run of the run number m is denoted by r [m].

The run data of runs r [0] to r [4] generated by run length processing of the binary image data BFI in Fig. 5 is stored in the run data storage area of the label storage section 645 in the run number order A state is shown in Fig. 6 (b). In the figure, "label number" indicates the label given to the run, "start-end pixel position" indicates a column index indicating the position of the start-end pixel of the run in the binary image data (BFI) End pixel position "indicates a column index indicating the position of the end pixel of the run in the binary image data BFI, and the" row index "indicates the row index of the run in the binary image data BFI. Among these, the " label number " is a value indicating the connection relationship between the run and the run other than the run. By executing the update process (step S3) and the search process (step S4) described later, (BFI), the same value is assigned to the runs connected to each other. Further, in the present embodiment, in order to efficiently execute the search process (step S4), the run connection table showing the connection relationship between the runs is provided. In this connection table, the run connection data id [m] is provided as data indicating the connection status of the run r [m], and the connection status of the run r [m] and other runs is indicated by the label number. Is stored in the run-connected data storage area of the labeling storage unit 645 in a tabular form as shown in Fig. 6 (c).

Returning to Fig. 4, the description of the labeling process is continued. In the present embodiment, when a run is created from the binary image data BFI, a different label number is set as a label indicating a connection relationship between the run and the run other than the run (step S2). In the present embodiment, the label number of run r [m] and the label number of run connection data id [m] are matched with run number m. That is, the run r [0], r [1], ... Quot; 0 ", " 1 ", " In addition to the initial setting of the run connection data id [0], id [1], ... Quot; 0 ", " 1 ", " .

In the next step S3, an update process (step S3) is executed. In this update processing, individual update processing is executed for each run. This " individual update process " refers to one run as a process target, and determines whether or not there is a connection run that connects vertically and obliquely in the downward direction of the run subject to be processed. If there is a connection run, In the present embodiment, the update process is performed for each of a plurality of (for example, n) runs. Here, although it is possible to perform these n individual update processes in serial, for example, run number order, in this case, a lot of time is spent in the update process.

Thus, in the present embodiment, the individual update processing for each run is made as one thread, and the threads are executed in parallel by each processor core 641 in the GPU 642. [ That is, as shown in Fig. 3, in the 0th processor core 641, the run r [0] is individually updated as the run target. In parallel thereto, the first processor core 641 executes the run r [ Is individually executed as a process target run. In addition, other runs r [2], r [3], ... , r [n] are also executed in parallel with the individual update processing. However, in each individual update process, the update process of the label number in the run connection data is performed as will be described later. Therefore, in this embodiment, the parallel process control unit 648 performs exclusion control among the processor cores 641. [

7 is a flowchart showing individual update processing executed in each processor core. 8 is a schematic diagram showing an example of individual update processing executed in each processor core. Here, the basic operation of the individual update process for the run r [p] of the run number p (p = 0, 1, 2, ..., or n) will first be described with reference to FIG. 7, An example of updating the label number will be described with reference to examples.

The individual update processing of the run [p] is executed by the p-processor core 641 in the following order. In step S11, the run r [p] is set as the target run, and the run connected to the downward run of this run target r [p] is extracted as the first connected run. If the first connection run has not been extracted (" NO " in step S12), the individual update processing in the thread ends.

On the other hand, if the first connection run is extracted ("YES" in step S12), the following steps S13 to S21 are executed to update the label number. Here, it is assumed that the run r [q] located on the downward run of the run r [p] is extracted as the first connected run.

In the next step S13, the run connection data id [p] and id [q] corresponding to the runs r [p] and r [q] are read from the label storage section 645. If the run connection data id [p] and id [q] have the same value, that is, the same label number is set and the connection relationship is already set (YES in step S14) The processing is terminated.

On the other hand, when the run connection data id [p] and id [q] which are determined as " NO " in step S14 are different (different label numbers are set), the run connection data is updated. The fact that the run connection data id [p] and id [q] are different from each other in this manner means that a connection relationship is not established between run [p] and run [q]. However, since the connection relationship is actually established, one label number is updated to the other label number in accordance with the magnitude relationship between the run connection data id [p] and id [q] (steps S15 to S19).

In the case where the run connection data id [p] is found to be a label number larger than the run connection data id [q] from the comparison result in the step S15 in comparison with the run connection data id [p] and id [q] Rewrites the connection data id [p] to the run connection data id [q], and updates the run connection data id [p] to a smaller label number (step S16). It is also possible to register the label number of the run r [p] on the side to be changed as the value old_dum, register the run connection data id [p] before the change as the value old_id, And registers the connection data id [p] as a value (new_id) (step S17).

Conversely, when it is found that the run connection data id [q] is a label number larger than the run connection data id [p], the run connection data id [q] is rewritten with the run connection data id [p] id [q] to a smaller label number (step S18). The label number of the run r [q] on the side to be changed in this way is registered as the value old_dum, the run connection data id [q] before the change is registered as the value old_id, The data id [q] is registered as a value (new_id) (step S19).

If the value (old_dum) matches the value (old_id) ("YES" in step S20), the individual update processing ends. On the other hand, when both are not coincident (" NO " in step S1O), the run on the side where the label number is updated (hereinafter referred to as the update run) (Hereinafter referred to as " second connection run "). Here, if such a second connected run exists, the process returns to step S13 after executing step S21. That is, in step S21, the run number p is rewritten to the value (old_id), and the run number q is rewritten to the value (new_id). Thereby, the renewed run is set as a new run r [p] and the second connected run is set as a new run r [q], and the process returns to step S13 and the difference of the label numbers (Step S14). When the label numbers are different, the larger label number is updated to the smaller label number (steps S15 to S19).

As described above, in the individual update processing, the label number is updated not only for the direct connection run directly connected to the run target, but also for the indirect run connected indirectly to the run target run through the direct connected run.

Next, the content of the update process (step S3) is described with reference to an example (Fig. 8) in the case where the update process is executed for the runs r [0] to r [4] shown in Figs. 6A and 6B I will explain in more detail. 8 shows the elapsed time. At the start time, the run connection data id [0] to id [4] is set to an initial value of 0 to 4, respectively. The run connection data id [o] to id [4] denoted by black indicates the label number after the update change. The items listed below the "0th thread" to the "fourth thread" ("connected run: existence", "prohibition of rewriting", etc.) indicate processing contents executed in each processor core 641.

In the five processor cores 641 of the n processor cores 641, the individual update processes of the runs r [0] to r [4] are executed in parallel. Among them, the individual update processes of the runs r [3] and r [4] are executed as the third thread and the fourth thread, respectively. However, since neither of the down and the first connection run exists, S3-3, and S3-4, and completes the individual updating process. On the other hand, in the individual update processing of the runs r [0] to r [2], it is confirmed that there is a connection run, and the run connection data corresponding thereto is read, compared, rewritten, and the like. Here, for example, in the second thread (the individual update processing of the run [2]), the run-in run r [2] and the run connection data id [2] of the connected run r [ The value of [4], that is, the label number, is read out. Here,

id [2] = 2,

id [4] = 4,

. As a result of comparison between them, since the run connection data id [4] is larger than the run connection data id [2], the label number of the run connection data id [4] is rewritten to the label number of the run connection data id [2] There is a need. When the preparation for rewriting in the second thread is first provided, the parallel processing control unit 648 temporarily stores the run-in data in the state in which the data rewriting in the individual update processing of the O-th thread and the first thread is temporarily inhibited by the parallel processing control unit 648. [ id [4] is rewritten, and the label number of the run connection data id [4] is updated from "4" to "2". The value (old_dum) which is the label number of the updated run r [4] and the value (old_id) which is the run connection data id [4] before the change are both equal to "4" Step S3-2) ends.

Next, as shown in the figure, if the first thread is ready to be rewritten, the parallel processing control unit 648 judges whether or not the rewriting of data in the individual update processing of the first thread is temporarily prohibited by the parallel processing control unit 648 Rewriting of the connection data id [2], id [3] is executed. More specifically, in the first thread (individual update processing of run [1]), run-in run r [1] and run r [2] and r [3] are connected in descending order, It is recognized as a connection run. Then, the run connection data id [1] of the object run r [1] and the run connection data id [2] and id [3] of the connect runs r [2] and r [3] are read out. they,

id [1] = 1,

id [2] = 2,

id [3] = 3,

Respectively. Since the run connection data id [2] and id [3] are both larger than the run connection data id [1] as a result of comparison between them, the run connection data id [2] Rewrites to the label number of data id [1]. For the updated run r [2], the value (old_dum) as the label number and the value (old_id) as the run connection data id [2] before the change are both "2". For the updated run r [3], the value (old_dum) which is the label number and the value (old_id) which is the run connection data id [3] before the change are both "3". Thus, the first thread (step S3-1) is terminated because the value (old_dum) and the value (old_id) have the same value for both the runs r [2] and r [3].

, The run-in run r [0] and the run r [2] are connected in the last run of the 0th thread (the individual update processing of the run [0] And rewrites the label number of the run connection data id [2] to the label number of the run connection data id [0] in the same manner as the second thread. The updated run connection data id [2] has already been rewritten once and becomes "1" from the initial value (initial label number of the run r [2] = 2), and the value which is the label number of the run r [ (old_dum) and the value (old_id) of the run connection data id [2] before the change are "2" and "1", respectively. As a result, the run connection data id [0] and id [1] are read, compared, and rewritten as executed in step S21 of FIG. 7, 1] to the label number of the run connection data id [0]. Thus, the value (old_dum) which is the label number of the updated run r [1] and the value (old_id) which is the run connection data id [1] before the change are both equal to "1" (Step S3-0). Thus, the run r [1] is indirectly connected through the run r [2] although it is not directly connected to the run r [0] to be processed, and the run connection data is also updated there.

When all the threads are thus terminated, the update processing (step S3) is completed and the update processing is executed for the runs r [0] to r [4] shown in Figs. 6A and 6B, [0] to id [4] are updated from (0, 1, 2, 3, 4) to (0, 0, 0, 1, 2). Also, as can be seen from the above description, since the processes are performed concurrently by one thread and one run, the label numbers of the run connection data id [0] to id [4] after the update processing take different values depending on which run is processed first .

Next, returning to Fig. 4, the description of the labeling process is continued. When the update process (step S3) is completed, the search process (step S4) is executed. This search process is also performed by each processor core 641 in the GPU 642 while controlling the rewriting timing by the parallel processing control unit 648, In parallel. That is, as shown in Fig. 3, in the 0th processor core 641, the individual search process is executed as the run r [0] as the process target run, and concurrently, the first processor core 641 executes the run r [ The individual search process is executed as the process target run. In addition, other runs r [2], r [3], ... , r [n] are also executed in parallel with the individual search processing. In addition, the "individual search process" refers to a process of connecting a run to connect to the process target run as one run, as a run connection data id [0], id [1] [n]), and rewrites the label number of the object run and the connected run in the same manner. Hereinafter, individual search processing will be described in detail with reference to FIG. 9 and FIG.

9 is a flowchart showing individual search processing executed in each processor core. 10 is a schematic diagram showing an example of individual search processing executed in each processor core. Here, as in the case of the individual update process, the basic operation of the individual search process for the run r [p] is first described with reference to Fig. 9, and then the rewriting of the label number Will be described.

The individual search processing of the run [p] is executed by the p-processor core 641 in the following order. In step S31, the run-link data id [p] corresponding to the run-in target r [p] is read from the run-connection table and is acquired. Then, it is judged whether or not the connection data id [p] is equal to the initial value of the label number of the process target run r [p] (step S32). Here, if it is determined that they are the same, the individual search processing is completed as it is. On the other hand, when it is judged that they are different ("NO" in the step S32), the steps S33 to S38 are executed to rewrite the label number.

After registering the run connection data id [p] as a value (new_dum) in step S33, the run connection data id [new_dum] is read from the run connection table (step S34). Then, it is judged whether or not the run connection data id [new_dum] and the value (new_dum) are the same (step S35) and rewrites the value new_dum to the run connection data id [new_dum] Step S36) and returns to step S34 to repeat the reading of the new run connection data id [new_dum].

Then, the label number of the run r [p] of the run number p is set to the value (new_dum) at the point of time when the run connection data id [new_dum] and the value (new_dum) (Step S37). Also, the run connection data id [p] is rewritten to the same value (new_dum) (step S38), and the individual search process is terminated. This rewriting is executed in a state in which the rewriting timing is controlled by the parallel processing control unit 648 in the same manner as in the case of the individual update processing.

Next, the contents of the search process (step S4) will be described in more detail with reference to an example (FIG. 10) in the case where the run connection table is updated from the initial state by the update process as shown in FIG. 8 and then the search process is executed do. Here, individual search processing of the run r [4] as the target run will be described with reference to Fig. The vertical axis in FIG. 10 indicates elapse of time, and the initial values of the label numbers of the runs r [0] to r [4] are set to 0 to 4, respectively. The label number in black indicates the label number to be searched. The items listed under "fourth thread" (acquisition of id [4] → id [4] = 2, etc.) indicate processing contents executed in fourth processor core 641.

The fourth processor core 641 reads the run connection data id [4] corresponding to this process run r [4] from the run connection table, using the run r [4] = 2) is compared with the initial value (= 4) of the label number of run r [4], and it is confirmed that the two are inconsistent. Then, the read object is moved to the run connection data id [new_dum] of the label number (new_dum) indicated by the run connection data id [4], that is, the run connection data id [2] 2] and compares this value (= 0) with the initial value (= 2) of the label number of run r [2]. Then, if it is confirmed that the two are inconsistent, the read object is moved again to the run connection data id [new_dum] of the label number (new_dum) indicated by the run connection data id [2], that is, the run connection data id [ The run connection data id [0] is read from the run connection table and this value (= 0) is compared with the initial value (= 0) of the label number of run r [0]. Then, it is confirmed that they are identical, the label number of the run r [4] to be processed is rewritten to the value, and this thread is ended. Also, the same individual search process is executed for the other threads, and as a result, the runs connected to each other are searched and given the same label. For example, the same label (= 0) is given to the runs r [0] to r [4] (see FIG. 6) obtained by run length processing on the binary image data shown in FIG.

As described above, in the present embodiment, the two-pass process of the update process (step S3) and the search process (step S4) is performed after the label numbers having different values are initially set for each run r [m]. In the renewal process (step S3), the individual update process (steps S3-0, S3-1, ...) is executed for each run r [m] The number is renewed. In the search process (step S4), the individual search process (steps S4-0, S4-1, ...) is executed for each run r [m] And the labels of the object run and the connection run are rewritten in the same manner. As described above, labeling processing of the binary image data (BFI) can be performed by the number of labels equal to the number of runs m, and labeling processing can be speeded up.

In addition, since the update processing (step S3) and the search processing (step S4) are performed in parallel on the individual update processing and the individual search processing for each run with one run of one thread, the total time required for the labeling processing Can be greatly shortened.

Further, in the individual update processing (steps S3-0, S3-1, ...), when the label number of the first connection run to be directly connected to the processing target run is updated, not only the first connection run, When there is a second connected run other than the target run connected to the run, the second connected run is also updated with the label. Therefore, it is possible to shorten the time required for the individual update process for the second connection run to be executed later than the individual update process for the process target run. As a result, the total processing time can be shortened.

The present invention is not limited to the above-described embodiments, and various modifications other than those described above can be made without departing from the spirit of the present invention. For example, in the above-described embodiment, the individual update processing is executed in parallel by one run-1 thread, but the division mode of the plurality of runs r [m] is not limited to this. That is, the division mode of the plurality of runs r [m] may be appropriately changed according to the number of the processor cores 641, the plurality of runs r [m] may be divided into a plurality of sets according to the number of the processor cores 641, Each of the processor cores may be configured to execute the individual update processing as one update thread for one or a plurality of divided runs. This is exactly the same in the individual search process.

In the above embodiment, the GPU 642 having a plurality of processor cores 641 is used to execute the individual updating process and the individual updating process. However, instead of the GPU, a plurality of CPUs may be installed, and one May be configured to execute the thread of FIG. In addition, an arithmetic processing unit 646 and a labeling storage unit 645 may be provided in the GPU 642.

In the above embodiment, the parallel processing control unit 648 located at the upper position of the 0th to nth processor cores 641 performs access management of the labeling storage unit 645, and in both the update processing and the search processing , And data rewriting by other threads is temporarily prohibited in accordance with data rewriting by one thread. That is, when rewriting the run connection data id [m] (where 0? M? N) with one thread, for example, temporarily rewriting the run connection data id [m] Not only the contention of the update is avoided but also the rewriting of other run-in connection data is temporarily prohibited. Such access management is not necessarily optimum, and there is room for improvement. For example, in order to achieve a new improvement in processing speed, it is preferable that access control of the labeling storage unit 645 by the parallel processing control unit 648 be executed as follows. That is, when rewriting the run connection data id [m] with one thread, it is possible to temporarily prohibit rewriting of the run connection data id [m] by another thread to avoid conflict of label update data, it is desirable to allow access to the run connection data other than [m] to permit data rewriting.

In the above embodiment, a run r [m] is obtained by making the normal image data BFI run length-by-row to obtain a run r [m], and a run connected in the downward direction and in the oblique direction However, the embodiment of the run length and the determination of the connection run are not limited to this. For example, the run length may be run lengthwise and the labeling method according to the present invention may be applied to the run created by the run length processing Can be applied. In the above-described embodiment, the connection run is determined by the 8-sided search for the downward run as described above, but the connection run may be determined by another search method, for example, 4-sided search. In the run-length process of the above embodiment, the run is generated at once in the entire image. However, the image may be divided into an appropriate size region and a run may be generated in each divided range. In this case, however, it is necessary to perform run connection determination processing at the boundary between the divided regions.

In the above embodiment, the label number (0, 1, 2, ...) of the serial number is used as the "label" of the present invention. Values other than numbers and numbers different from this are used as "labels" May be used. In addition, the label is renewed by rewriting to a smaller number, but the aspect of the label update is not limited to this.

Further, in the above-described embodiment, the run image data (BFI) is run-lengthized in the apparatus, and then the run process is performed on the run generated by the run-length process. But the present invention may be applied to an apparatus or method for labeling a run created by an external apparatus, for example.

The present invention can be appropriately applied to a labeling technique for assigning labels to a plurality of runs created by rendering normal image data run length.

1: Inspection system
5:
6:
21:
62: Difference extracting unit (image extracting unit)
63: Binary processing unit
64: labeling section
641: processor core
642: GPU
643: Individual update processing unit
644: Individual search processor
645: labeling storage unit
646: Operation processing unit (control unit)
647:
648: Parallel processing controller
649: Data initial setting section

Claims (11)

A labeling method for carrying out a labeling process on a plurality of runs created by rendering normal image data as run lengths,
An initial label setting step of setting a different label for each run,
The method comprising the steps of: determining whether or not there is a first connection run to connect to the processing object run, and if the first connection run exists, An update step of performing an individual update process of updating one of the labels on the connection run to the other label,
And a search step of searching for runs connected to each other among the plurality of runs based on the plurality of labels and assigning the same labels after the updating step,
Wherein the updating step divides the plurality of runs into a plurality of sets and sets the individual update processing executed in each set as an update thread and rewrites the label by one of the plurality of update threads under exclusive control Wherein the plurality of update threads are executed in parallel while prohibiting rewriting by another update thread of the label rewritten by the one update thread. The method according to claim 1,
Wherein the updating step permits rewriting of the label other than the label rewritten by the one update thread by the another update thread. The method according to claim 1 or 2,
Wherein in the individual update processing, the run of the target run and the one of the first connected run that has updated the label are set as the update run, and the target run and the second connection And when the second connection run exists, the label of one of the update run and the second connection run is updated to the other label. The method according to claim 1 or 2,
Wherein the plurality of runs are obtained by making the normal image data run lengthwise for each row,
And in the individual update processing, the run connected to the downward run of the run to be processed is determined as the first connected run. The method according to claim 1 or 2,
In the searching step, the run is determined as a to-be-processed run, and a connection run to be connected to the to-be-processed run is searched based on the plurality of labels and the labels of the to-be- And performs an individual search process to rewrite the search result. The method of claim 5,
Wherein the searching step divides the plurality of runs into a plurality of sets and makes the individual search processing executed in each set a search thread, and rewriting the label by the search thread of one of the plurality of search threads under exclusive control Wherein the plurality of search threads are run in parallel while prohibiting rewriting by another search thread of the label rewritten by the one search thread. The method of claim 6,
Wherein the search step permits rewriting by the other search thread of a label other than the label rewritten by the one search thread. A labeling apparatus for carrying out a labeling process for a plurality of runs created by rendering normal image data as run lengths,
A storage unit for storing a label of each run;
A processor unit having a plurality of processor cores for performing arithmetic processing;
Wherein the control unit judges whether or not there is a first connection run to be connected to the object to be processed together with the run as the object to be processed for each run after initially setting different values to the plurality of labels, When there is a connection run, performs an individual update process of updating one of the run of the object to be processed and the first connection run to the other label, and connects the other run among the plurality of runs based on the plurality of labels And a control section for searching for the run and giving the same label,
Wherein the plurality of runs are divided into a plurality of sets according to the number of the processor cores,
Each processor core executing the individual update processing for one or more divided runs as an update thread in each of the plurality of sets,
Wherein the control unit permits rewriting of the label by one of the plurality of update threads under exclusive control and prohibits rewriting of the label by another update thread of the label rewritten by the one update thread, And performs update processing for each run by executing the update threads in parallel. The method of claim 8,
The control unit searches for a connection run that connects the run to be processed with the run as a to-be-processed run based on the plurality of labels for each run and identifies the labels of the to-be- And performs a rewriting individual search process. The method of claim 9,
Each processor core executing the individual search process for one or more divided runs as a search thread in each of the plurality of sets,
Wherein the control unit permits rewriting of the label by the search thread of one of the plurality of search threads under exclusive control and prohibits rewriting by the other search thread of the label rewritten by the one search thread, And performs individual search processing for each run by executing search threads in parallel. An image acquisition unit for acquiring an image to be inspected;
An image extracting unit for inspecting the inspection object image to extract a defective image including a defective portion;
A binarization processor for binarizing the defect image to generate binary image data,
A run generation section for generating a plurality of runs by rendering the above-mentioned normal image data run length,
A defect inspection apparatus having the same structure as the labeling apparatus according to any one of claims 8 to 10, further comprising labeling means for giving identical labels to runs connected to each other among the plurality of runs.

Images