TWI434230B - Image defect instpection system and method - Google Patents

Description

Image flaw detection system and method

The present invention relates to a flaw detection system and method, and more particularly to an image flaw detection system and method.

In recent years, in order to maintain a good yield in the process, the cost of testing has a considerable proportion. However, most of the related equipment used in the detection of cockroaches is still provided by foreign manufacturers, and the investment cost is quite expensive. Such an important, technical and high-value testing equipment is completely controlled by foreign manufacturers. Not only does the domestic related industry have to spend huge sums of money to purchase and use, but in practical applications, it needs to adjust equipment and adapt to the process. The above adjustments often result in immediate and perfect technical services due to geographical relationships. As a result, the production capacity and yield are reduced, and the product delivery schedule is seriously delayed and the manufacturing cost is increased.

In addition, the detection of large-area images contains many programs with high computational complexity and low correlation of data. Most of the prior art processes the detection of large-area images by the central processing unit (CPU) of the computer. However, at present, the number of cores of the CPU is small, which limits the number of programs that the CPU processes at the same time, and increases the execution time required for the detection program.

Therefore, an aspect of the present invention provides an image detection system for processing a plurality of regions of a to-be-tested image in parallel by a graphic processing unit (GPU) having a plurality of cores. Piece. The image detection system includes a graphics processing component and a central processing component. The graphics processing component contains several graphics processing cores. The central processing component is electrically coupled to the graphics processing component. The central processing component includes an image acquisition module, an image segmentation module, a graphics processing component driver module, and a clustering module. The image acquisition module obtains an image to be tested. The image segmentation module divides the image to be tested into a plurality of blocks to be tested. The graphics processing component driving module drives the graphics processing core to process the block to be tested in parallel, thereby enabling the graphics processing core to obtain a plurality of defects from the block to be tested. The 瑕疵 clustering module groups the points into at least one group.

Another aspect of the present invention provides an image detection method for processing a plurality of blocks of a to-be-tested image in parallel by a graphics processing component having a plurality of cores in a computer. The image detection method includes the following steps: obtaining an image to be tested to a computer. The computer includes a central processing component and a graphics processing component, and the graphics processing component includes a plurality of graphics processing cores. The image to be tested is divided into several blocks to be tested. The block to be tested is processed in parallel by the graphics processing core to obtain a plurality of defects from the block to be tested. The points are clustered into at least one group by a central processing element.

It will be apparent from the above-described embodiments of the present invention that the application of the present invention has the following advantages. Since the graphics processing component has more cores than the central processing component, the operation of acquiring multiple defects from the image to be tested is processed by the graphics processing component, so that more processing operations can be performed simultaneously on each core. When the embodiment of the present invention is applied to the panel 瑕疵 detection, the larger the size of the panel to be tested, the more the number of blocks that can be divided by the image to be tested obtained by taking the panel to be tested, and thus the graphics processing component Multiple core parallel processing can greatly reduce the execution time of panel defects. In addition, the higher the amount of data of the image to be tested detected by the present invention, the more the number of blocks that can be divided into the image to be tested, so that the parallel processing of the plurality of cores of the graphics processing component can also greatly reduce the detection of defects. Execution time. In addition, image defect detection can be performed by the central processing unit of the computer and the graphics processing component of the display card, and the required hardware cost is relatively low. In addition, it is possible to determine the cause of the occurrence of defects on the image by grouping a plurality of defects into groups.

Please refer to FIG. 1 , which is a functional block diagram of an image detection system according to an embodiment of the invention. The image detection system processes a plurality of blocks of a to-be-tested image in parallel by a graphics processing component having more cores.

The image detection system 100 includes a graphics processing component 110 and a central processing component 120. Graphics processing component 110 includes a plurality of graphics processing cores 112, 113, ..., 11n. The central processing component 120 is electrically coupled to the graphics processing component 110. The central processing component 120 includes an image acquisition module 121, an image segmentation module 122, a graphics processing component driver module 123, and a clustering module 124. The image acquisition module 121 obtains an image to be tested.

The image flaw detection system 100 can be applied to panel flaw detection. Accordingly, the image detection system 100 can include an imaging element 140 that is electrically coupled to the central processing component 120. In this way, a panel 200 to be tested can be taken by the imaging component to obtain an image to be tested. However, in other embodiments, the image detection system 100 may include a plurality of imaging elements for capturing the panel 200 to be tested, and is not limited to the embodiment. In this way, the image acquisition module 121 can obtain the image to be tested of the panel to be tested from the imaging component 140. However, in other embodiments, the imaging element may be caused to capture other objects to be tested to obtain an image to be tested, and is not limited to the embodiment.

In addition, the image detection system 100 can include a data transmission interface 130 electrically coupled to the central processing component 120. The data transmission interface 130 can be a bus, a network card or other data transmission interface. In this way, the image acquisition module 121 can obtain the image to be tested from other devices through the data transmission interface 130.

The image segmentation module 122 divides the image to be tested into a plurality of blocks to be tested. The graphics processing component driving module 123 drives the graphics processing cores 112, 113, ..., 11n to process the blocks to be tested in parallel, thereby obtaining the graphics processing cores 112, 113, ..., 11n from the respective blocks to be tested. A few defects. The graphics processing component driver module 123 can drive the graphics processing cores 112, 113, ..., 11n by computing the architecture of a Compute Unified Device Architecture (CUDA) or other graphics processing component 110 that drives multiple cores. . In addition, the graphics processing cores 112, 113, ..., 11n are self-contained by Laplacian, Smoothing, Adaptive thresholding, or other image processing methods. The test block has several defects.

The 瑕疵 clustering module 124 clusters the defects into at least one group. The 瑕疵 clustering module 124 can first form a plurality of 瑕疵 line segments by running length encoding (RLE). Next, the 瑕疵 clustering module 124 forms a plurality of objects in the 瑕疵 line segment by means of a blob table. Then, the 瑕疵 clustering module 124 groups the objects into at least one group according to the spatial relationship of the objects. In this way, the group on the image to be tested can be obtained.

The graphics processing component 110 can include a memory 111 that is electrically coupled to one of the graphics processing cores 112, 113, ..., 11n. The image segmentation module 122 can divide the image to be tested into a plurality of storage blocks of the memory 111 after dividing the image to be tested into a plurality of blocks to be tested. In this way, each graphics processing core 112, 113, ..., 11n can obtain each of the to-be-tested blocks of the image to be tested from the storage blocks of the memory 111 for processing.

In addition, the image segmentation module 122 can directly store the image to be tested to the memory 111, and calculate the block storage address of each block to be tested according to the image storage address of the image to be tested. Each of the graphics processing cores 112, 113, ..., 11n obtains each of the blocks to be tested from the memory 111 for processing according to the block storage address. In this way, the image segmentation module 122 can simulate dividing the image to be tested into multiple blocks to be tested without actually dividing the image to be tested.

Please refer to FIG. 2, which is a flowchart of an image detection method according to another embodiment of the present invention. The image detection method parallelly processes a plurality of blocks of a to-be-tested image by using a graphics processing component having more cores in the computer.

The image detection method 300 includes the following steps: In step 310, an image to be tested is obtained to a computer. The computer includes a central processing component and a graphics processing component, and the graphics processing component includes a plurality of graphics processing cores.

In step 320, the image to be tested is divided into a plurality of blocks to be tested. Next, each of the blocks to be tested is processed in parallel by each graphics processing core (step 330) to obtain a plurality of defects from the block to be tested (step 340). The graphics processing core may process the block to be tested in parallel by using a Laplacian operation, a smoothing operation, an adaptive binarization operation, or other image processing methods (step 330) to obtain a defect from the block to be tested ( Step 340).

In step 350, the points are grouped into at least one group by a central processing element. The central processing component may first be encoded by running length, and the adjacent defects are formed into a plurality of 瑕疵 line segments. Next, the central processing component forms a number of objects by connecting the connected segments by the object table. Then, the central processing component groups the objects into at least one group according to the spatial relationship of the objects. In this way, the defects can be clustered into at least one group by the central processing component (step 350).

There are multiple embodiments of step 310. In an embodiment, before the step 310, a panel to be tested may be taken to obtain an image to be tested of the panel to be tested (step 310). In this way, the panel detection can be performed by the image detection method 300. However, in other embodiments, before the step 310, other objects to be tested may be taken to obtain an image to be tested of the object to be tested (step 310), and is not limited to the embodiment.

In another embodiment, the computer can include a data transmission interface, and the image to be tested is obtained through the data transmission interface (step 310). The data transmission interface can be a bus, a network card or other data transmission interface. In this way, the image to be tested can be obtained from other devices through the data transmission interface.

In another embodiment, a to-be-tested image file can be obtained, and raw data of one of the to-be-tested image files is taken from the image to be tested as the image to be tested. The format of the image to be tested may be a bitmap (bmp), a joint photographic experts group (jpeg) or other image format.

Figure 3 is a diagram in which the image to be tested is divided into a plurality of blocks in step 2 (step 320), and an embodiment of each block to be tested (step 330) is processed in parallel by each graphics processing core. The graphics processing component further includes a memory. Dividing the image to be tested into the block to be tested (step 320), and processing the block to be tested in parallel by the graphics processing core (step 330) includes: in step 410, storing the image to be tested into the memory of the graphics processing component .

In step 420, the image to be tested is stored in an image storage address of the memory.

In step 430, the block storage address of the plurality of blocks to be tested is calculated according to the image storage address.

In step 440, each graphics processing core obtains the to-be-tested block from the block storage address of the memory to process each of the to-be-tested blocks in parallel. In this way, the block storage address of each block to be tested of the image to be tested can be calculated, and the image to be tested can be simulated. However, in other embodiments, the image to be tested may be actually divided into a plurality of blocks to be tested, and is not limited to the embodiment.

It will be apparent from the above-described embodiments of the present invention that the application of the present invention has the following advantages. Since the graphics processing component has more cores than the central processing component, the operation of acquiring multiple defects from the image to be tested is processed by the graphics processing component, so that more processing operations can be performed simultaneously on each core. When the embodiment of the present invention is applied to the panel 瑕疵 detection, the larger the size of the panel to be tested, the more the number of blocks that can be divided by the image to be tested obtained by taking the panel to be tested, and thus the graphics processing component Multiple core parallel processing can greatly reduce the execution time of panel defects. In addition, the higher the amount of data of the image to be tested detected by the present invention, the more the number of blocks that can be divided into the image to be tested, so that the parallel processing of the plurality of cores of the graphics processing component can also greatly reduce the detection of defects. Execution time. In addition, image defect detection can be performed by the central processing unit of the computer and the graphics processing component of the display card, and the required hardware cost is relatively low. In addition, it is possible to determine the cause of the occurrence of defects on the image by grouping a plurality of defects into groups.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and the present invention can be modified and modified without departing from the spirit and scope of the present invention. The scope is subject to the definition of the scope of the patent application attached.

100. . . Image detection system

110. . . Graphics processing component

111. . . Memory

112, 113, ..., 11n. . . Graphics processing core

120. . . Central processing component

121. . . Image acquisition module

122. . . Image segmentation module

123. . . Graphics processing component driver module

124. . .瑕疵 clustering module

130. . . Data transmission interface

140. . . Shooting component

200. . . Panel to be tested

300. . . Image detection method

310～350. . . step

410～440. . . step

The above and other objects, features, advantages and embodiments of the present invention will become more apparent and understood.

FIG. 1 is a functional block diagram of an image detection system according to an embodiment of the invention.

2 is a flow chart of an image detection method according to another embodiment of the present invention.

Figure 3 is a diagram in which the image to be tested is divided into a plurality of blocks in step 2 (step 320), and an embodiment of each block to be tested (step 330) is processed in parallel by each graphics processing core.

100. . . Image detection system

110. . . Graphics processing component

111. . . Memory

112, 113, ..., 11n. . . Graphics processing core

120. . . Central processing component

121. . . Image acquisition module

122. . . Image segmentation module

123. . . Graphics processing component driver module

124. . .瑕疵 clustering module

130. . . Data transmission interface

140. . . Shooting component

200. . . Panel to be tested

Claims (12)

An image detection system includes: a graphics processing component, comprising: a plurality of graphics processing cores; and a central processing component electrically coupled to the graphics processing component, comprising: an image acquisition module to obtain an image to be tested; The image segmentation module divides the image to be tested into a plurality of blocks to be tested; a graphics processing component driving module drives the graphics processing cores to process the blocks to be tested in parallel, thereby enabling the graphics processing cores A plurality of defects are obtained from the blocks to be tested, and a clustering module is clustered into at least one group. The image detection device of claim 1, wherein the graphics processing component further comprises: a memory, electrically connected to the graphics processing core, wherein the image segmentation module stores the to-be-tested blocks to the memory The graphics processing component driving module controls the graphics processing cores to respectively obtain the to-be-tested blocks from the memory to process the to-be-tested blocks in parallel. The image detection system of claim 1, further comprising: at least one imaging component electrically connected to the central processing component to capture an object to be tested to obtain the image to be tested. The image detection system of claim 1, further comprising: a data transmission interface electrically connected to the central processing component, wherein the image acquisition module obtains the image to be tested through the data transmission interface. The image detection system of claim 1, wherein the graphics processing component driver module drives the graphics processing cores by computing a unified device architecture. An image detection method includes: obtaining a to-be-tested image to a computer, wherein the computer includes a central processing component and a graphics processing component, the graphics processing component includes a plurality of graphics processing cores; and the image to be tested is divided into plural The blocks to be tested are processed in parallel by the graphics processing cores to obtain a plurality of defects from the blocks to be tested; and the group of nodes are obtained by the central processing component Gather into at least one group. The image detection device of claim 6, wherein the graphics processing component further comprises a memory, the image to be tested is divided into the to-be-tested blocks, and the graphics processing cores are processed in parallel by the graphics processing cores. The step of measuring the block includes: storing the image to be tested in the memory of the graphics processing component; obtaining the image to be tested and storing the image storage address in the memory; calculating the image according to the image storage address And storing a plurality of blocks of the block to be tested; and causing the graphics processing cores to store the addresses from the blocks, and acquiring the blocks to be tested to process the blocks to be tested in parallel. The image detection method of claim 6, wherein the grouping of the defects into the group by the central processing component comprises: forming a neighboring one of the plurality of defects into a plurality of line segments; The connected ones of the plurality of segments form a plurality of objects; and according to the spatial relationship of the objects, the objects are merged into the group. The image detection method of claim 6, wherein the parallel processing of the to-be-tested blocks by the graphics processing core comprises: performing Laplace on the to-be-tested blocks by using the graphics processing cores Operations, smoothing operations, and adaptive binarization operations to obtain the defects from the blocks to be tested. The image detection method of claim 6, wherein the obtaining the image to be tested comprises: obtaining a to-be-tested image file; capturing the original image data of the image to be tested from the image to be tested; and The image data is used as the image to be tested. The image detection method of claim 6, further comprising: capturing an object to be tested to obtain the image to be tested. The image detection method of claim 6, wherein the computer further comprises a data transmission interface, and the image to be tested is obtained through the data transmission interface.