Method for analyzing defect

Abstract

A method for analyzing defects is suitable for performing the following steps using an electronic device: A detection layout is divided into multiple vertical stripes along a horizontal direction, and a first defect point count is calculated for each vertical stripe. Furthermore, the detection layout is divided into multiple horizontal stripes along a vertical direction, and a second defect point count is calculated for each horizontal stripe. The detection layout is determined to have a defect cluster if at least one of the following predetermined conditions is met: the first defect point count of at least one vertical stripe is greater than a first preset value; and the second defect point count of at least one horizontal stripe is greater than a second preset value.

Description

[Technical Field]

[0001] This invention relates to a panel inspection method, and more particularly to a method for analyzing defects. [Previous Technology]

[0002] Most thin-film transistor liquid crystal display (TFT-LCD) manufacturers use multiple layers of defect inspection to identify defective panels in real time. The inspection methods mostly involve using inspection machines to automatically scan and photograph for comparison, thereby identifying problematic panels. Therefore, these defect inspection processes generate a large amount of defect information and defect image files.

[0003] After defects are detected by the machine, yield technicians must manually determine whether there are defect clusters in these defect information and defect image files. Since yield technicians must review the inspection results of more than 150 production batches every day, and need to check and judge defect clusters for each production batch and each panel, they must manually report any abnormalities with defect clusters. Manual judgment can lead to different judgment results due to inconsistent standards, and may also result in missed detections due to factors such as fatigue, which in turn affects factory capacity, yield, and cost. [Summary of the Invention]

[0004] The present invention provides a method for analyzing defects, which can improve the problem of inefficiency in the past.

[0005] The method for analyzing defects according to the present invention is suitable for performing the following steps using an electronic device, including: dividing a detection layout into multiple straight strip blocks along a horizontal direction and calculating a first number of defect points in each straight strip block; dividing the detection layout into multiple horizontal strip blocks along a vertical direction and calculating a second number of defect points in each horizontal strip block; and determining that the detection layout has a defect cluster if at least one of the following predetermined conditions is met, the predetermined conditions including: the first number of defect points in at least one straight strip block is greater than a first preset value; and the second number of defect points in at least one horizontal strip block is greater than a second preset value.

[0006] In one embodiment of the present invention, the above-mentioned method for analyzing defects further includes: calculating the first defect percentage of each vertical block in the detection layout using the first defect point count and the total defect point count of the detection layout; and calculating the second defect percentage of each horizontal block in the detection layout using the second defect point count and the total defect point count of the detection layout. The specified conditions further include: the first defect percentage of at least one vertical block is greater than the first single-piece percentage; and the second defect percentage of at least one horizontal block is greater than the second single-piece percentage.

[0007] In one embodiment of the present invention, the above-mentioned detection layout is obtained after one of a plurality of test items received from the defect detection machine is defect-detected, and the test items belong to the same production batch.

[0008] In one embodiment of the present invention, the above-mentioned method for analyzing defects further includes: calculating the total batch defect of all defect points obtained after defect detection in a production batch; calculating the vertical area proportion of each straight strip block in the production batch using the first number of defect points and the total batch defect; and calculating the horizontal area proportion of each horizontal strip block using the second number of defect points and the total batch defect. The specified conditions further include: the vertical area proportion of at least one straight strip block is greater than the proportion of the first batch; and the horizontal area proportion of at least one horizontal strip block is greater than the proportion of the second batch.

[0009] In one embodiment of the present invention, after calculating the total batch defects of all defect points obtained after defect detection of the production batch, the method further includes: if the total batch defects are greater than a first baseline number, calculating the per-piece defect ratio of the detection layout based on the total batch defects and the total number of defect points in the detection layout. The specified conditions further include: the total number of defect points is greater than or equal to a first preset value, and the per-piece defect ratio is greater than or equal to a second preset value.

[0010] In one embodiment of the present invention, the above-mentioned method for analyzing defects further includes: when the total number of batch defects is greater than a second benchmark number, the specified condition further includes: the vertical area ratio of at least one straight stripe block is greater than the first batch ratio, wherein the number of straight stripe blocks is 3; when the total number of batch defects is greater than a third benchmark number, the specified condition further includes: the horizontal area ratio of at least one horizontal stripe block is greater than the second batch ratio, wherein the number of horizontal stripe blocks is 3.

[0011] In one embodiment of the present invention, after determining that the detection layout has defective clusters, the method further includes: recording the corresponding cluster labels for the specified conditions that are met.

[0012] In one embodiment of the present invention, after determining that the detection layout has a defect cluster, the method further includes: generating an anomaly notification for the production batch with the defect cluster.

[0013] Based on the above, this disclosure groups the defect coordinates in the horizontal and vertical directions, calculates the number and proportion of each group, and the corresponding threshold value, thereby performing linear detection in the horizontal and vertical directions. Accordingly, the determination of defect clusters is automated to improve the inefficiency of the past. [Simplified Explanation of the Diagram]

[0035] FIG1 is a block diagram of an electronic device for analyzing defects according to an embodiment of the present invention.

[0036] Figure 2 is a flowchart of a method for analyzing defects according to an embodiment of the present invention.

[0037] Figures 3A and 3B are schematic diagrams of equal division according to an embodiment of the present invention.

[0038] Figures 4A and 4B are schematic diagrams of equal division according to another embodiment of the present invention.

Implementation Method

[0014] FIG1 is a block diagram of an electronic device for analyzing defects according to an embodiment of the present invention. Referring to FIG1, the electronic device 100 receives machine inspection data D to analyze defects. The electronic device 100 includes a processor 110 and a storage unit 120. The processor 110 is coupled to the storage unit 120. The machine inspection data D is generated by a defect inspection machine and includes defect information and defect coordinate data obtained after defect inspection of multiple test items belonging to the same production batch.

[0015] The processor 110 is, for example, a central processing unit (CPU), a physical processing unit (PPU), a programmable microprocessor, an embedded control chip, a digital signal processor (DSP), an application specific integrated circuit (ASIC), or other similar devices.

[0016] Storage unit 120 is, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), flash memory, hard disk, or other similar device or combination thereof. Storage unit 120 includes one or more code segments, which, after being installed, are executed by processor 110 to perform the following method for analyzing defects.

[0017] Figure 2 is a flowchart of a defect analysis method according to an embodiment of the present invention. Referring to Figures 1 and 2, after the electronic device 100 receives the machine inspection data D, the processor 110 can analyze the machine inspection data D to determine whether defect clusters appear on each panel. The machine inspection data D includes defect information and defect coordinate data corresponding to each batch and each panel. The processor 110 can first extract the inspection layout corresponding to each panel one by one from multiple defect coordinate data of the same production batch.

[0018] In step S205, the detection layout is divided into multiple straight strip blocks along the horizontal direction, and the number of first defect points in each straight strip block is calculated. Furthermore, in step S210, the detection layout is divided into multiple horizontal strip blocks along the vertical direction, and the number of second defect points in each horizontal strip block is calculated. The execution order of steps S205 and S210 is not limited here; for example, step S205 may be executed first, or step S210 may be executed first, or steps S205 and S210 may be executed simultaneously.

[0019] In one embodiment, the detection layout can be divided into multiple straight strip regions and multiple horizontal strip blocks at intervals of a specified length in both the horizontal and vertical directions, as shown in Figures 3A and 3B. Figures 3A and 3B are schematic diagrams of the equal division according to one embodiment of the present invention. In another embodiment, the detection layout is also divided into 3 straight strip regions and 3 horizontal strip blocks at intervals in both the horizontal and vertical directions, as shown in Figures 4A and 4B. Figures 4A and 4B are schematic diagrams of the equal division according to another embodiment of the present invention.

[0020] After the division, the number of first and second defect points is calculated for each of the divided vertical and horizontal regions. Then, in step S215, if specified conditions are met, it is determined that the detection layout has a defect cluster. The specified conditions include at least one of the following: the number of first defect points in at least one vertical block is greater than a first preset value; the number of second defect points in at least one horizontal block is greater than a second preset value.

[0021] Referring to Figures 3A and 3B, it is assumed that the detection layout is divided into n vertical strip blocks and n horizontal strip blocks along the horizontal and vertical directions at intervals of length (e.g., 1 mm). The number of first defect points xi (i=1~n) included in each vertical strip block and the number of second defect points yi (i=1~n) included in each horizontal strip block are calculated. Then, the number of first defect points xi is compared with a first preset value T1, and the number of second defect points yi is compared with a second preset value T2. Furthermore, the following conditions are set: xi>T1 (condition 1); yi>T2 (condition 2). In one embodiment, the first preset value and the second preset value can be set to 25 points.

[0022] Furthermore, the first defect ratio_xi (=xi / total defect number) of each straight block in the detection layout can be calculated using the first defect number xi and the total defect number of the detection layout; the second defect ratio_yi (=yi / total defect number) of each horizontal block in the detection layout can be calculated using the second defect number and the total defect number of the detection layout. Moreover, the specified conditions further include: the first defect ratio_xi of at least one straight block is greater than the first single-piece ratio Tr1; and the second defect ratio_yi of at least one horizontal block is greater than the second single-piece ratio Tr2. That is, ratio_xi>Tr1 (condition 3); ratio_yi>Tr2 ​​(condition 4).

[0023] That is, in the defect analysis disclosed herein, it can be further determined whether the proportion of the first defect point in this vertical strip block to the total number of defect points in a single sheet (the proportion of the first defect) is greater than the proportion of the first defect. Alternatively, it can be further determined whether the proportion of the second defect point in this horizontal strip block to the total number of defect points in a single sheet (the proportion of the second defect) is greater than the proportion of the second defect.

[0024] In one embodiment, the first preset value and the second preset value can be set to 30, and the first defect ratio and the second defect ratio can be set to 30%. Further, straight stripe blocks with xi greater than 30 and ratio_xi greater than 30% can be identified; or, horizontal stripe blocks with yi greater than 30 and ratio_yi greater than 30% can be identified. Alternatively, in other embodiments, it can be further set as follows: when the total number of defect points is greater than 42, horizontal stripe blocks with yi 20 and ratio_yi 30% can be identified.

[0025] Conditions 1 to 4 are used to identify defect clusters in a single panel, and can further be used to identify defect clusters within the same production batch. To identify defect clusters in a production batch, the image can be divided into three vertical bars (left, middle, and right) horizontally and into three horizontal bars (top, middle, and bottom) vertically for analysis, as shown in Figures 4A and 4B.

[0026] First, calculate the total batch defects of all panels in the same production batch after defect detection. If the total batch defects exceed the first baseline number, calculate the per-panel defect ratio of the detection layout (total defect number / total batch defects) based on the total batch defects and the total number of defect points in the detection layout. The specified conditions further include: the total number of defect points is greater than or equal to the first set value, and the per-panel defect ratio is greater than or equal to the second set value (condition 5).

[0027] For example, the first baseline quantity is 36, the first setting value is 27, and the second setting value is 23%. When the total number of defects in the batch is greater than 36, find the image corresponding to the panel with a total number of defect points greater than 27 and a single panel defect rate greater than or equal to 23%.

[0028] In addition, using the first number of defect points xi (i=1~3) and the total batch defects, the vertical area proportion Sratio_xi (=xi / total batch defects) of each straight strip block in the production batch is calculated; and using the second number of defect points yi (i=1~3) and the total batch defects, the horizontal area proportion Sratio_yi (=yi / total batch defects) of each horizontal strip block is calculated. The specified conditions further include: the vertical area proportion of at least one straight strip block is greater than the first batch proportion Tlotr1; and the horizontal area proportion of at least one horizontal strip block is greater than the second batch proportion Tlotr2. That is, Sratio_xi>Tlotr1 (condition 6); Sratio_yi>Tlotr2 (condition 7). In addition, it can be further set as follows: when the total batch defects are greater than the second benchmark quantity, condition 6 is judged; when the total batch defects are greater than the third benchmark quantity, condition 7 is judged.

[0029] For example, in condition 6, the second baseline quantity is set to 22, and the first batch percentage (Tlotr1) is 65%. If the total batch defects exceed 22, identify vertical stripe blocks with a vertical area percentage greater than 65%. In condition 7, the third baseline quantity is set to 21, and the second batch percentage (Tlotr2) is 80%. If the total batch defects exceed 21, identify horizontal stripe blocks with a horizontal area percentage greater than 80%.

[0030] In another embodiment, it can also be set as follows: first use conditions 1 and 2 to determine the defect clusters of a single panel, and after determining that it is a defect cluster, then analyze the area of ​​the defect clusters of the production batch in detail.

[0031] After determining that the detection layout has defective clusters, the corresponding cluster labels are further recorded for the specified conditions that are met. For example, a recording label is set, which includes 5 recording bits, namely (B1, B2, B3, B4, B5). Recording bits B1 to B4 are used to record whether conditions 1 to 4 are met, where "0" represents not met and "1" represents met. In addition, recording bit B5 is used to record whether conditions 5 to 7 are met, where "0" represents that none of conditions 5 to 7 are met, "1" represents that one of conditions 5 to 7 is met, "2" represents that two of conditions 5 to 7 are met, and "3" represents that all of conditions 5 to 7 are met.

[0032] Furthermore, for production batches with defect clusters, an anomaly notification is generated. For example, an anomaly notification is generated when one of the record bits B1 to B5 is not 0.

[0033] In one application example, the Python programming language is used to perform cluster detection and image recognition detection on the machine detection data D. The cluster detection includes the methods mentioned in the above embodiments, and can be further combined with the DBSCAN (Density-based spatial clustering of applications with noise) algorithm. The image recognition detection uses the OpenCV (Open Source Computer Vision Library) suite. Afterwards, the detection results are stored on the data access server, and after obtaining the weighted risk value, they are output to the interface via wired or wireless transmission for the user to make decisions.

[0034] In summary, this disclosure can perform horizontal and vertical linear detection on a single chip, and further, it can detect the proportion of regional clusters within the same production batch. Accordingly, the judgment of defect clusters is automated, improving the previously inefficient process. In addition, it can proactively report anomalies, thereby improving the batch processing efficiency for precise focusing anomalies and shortening time-consuming multi-platform operations.

Claims

1. A method for analyzing defects, suitable for using an electronic device to perform the following steps, including: Along a horizontal direction, a detection layout is divided into multiple straight strip blocks, and a first defect point count is calculated for each of the straight strip blocks; along a vertical direction, the detection layout is divided into multiple horizontal strip blocks, and a second defect point count is calculated for each of the horizontal strip blocks, wherein the division of the horizontal strip blocks and the division of the straight strip blocks are performed independently; and if at least one of the following conditions is met, it is determined that the detection layout has a defect cluster in at least one of the horizontal and vertical directions, the specified conditions including: the first defect point count of at least one of the straight strip blocks is greater than a first preset value; and the second defect point count of at least one of the horizontal strip blocks is greater than a second preset value.

2. The method for analyzing defects as described in claim 1 further includes: Using the first defect point count and the total defect point count of the detection layout, calculate the first defect percentage of each of the vertical strip blocks in the detection layout; and using the second defect point count and the total defect point count of the detection layout, calculate the second defect percentage of each of the horizontal strip blocks in the detection layout, wherein the specified conditions further include: the first defect percentage of at least one of the vertical strip blocks is greater than a first single-piece percentage; and the second defect percentage of at least one of the horizontal strip blocks is greater than a second single-piece percentage.

3. The method for analyzing defects as described in claim 1, wherein the detection layout is obtained by defect detection of one of a plurality of test items received from a defect detection machine, the test items belonging to the same production batch.

4. The method for analyzing defects as described in claim 3 further includes: Calculate the total batch defects of a production batch, including all defect points obtained after defect detection. Using the first number of defect points and the total number of defects in the batch, calculate the percentage of a vertical region for each of the vertical stripe blocks in the production batch; and using the second number of defect points and the total number of defects in the batch, calculate the percentage of a horizontal region for each of the horizontal stripe blocks, wherein the specified conditions further include: the percentage of the vertical region of at least one of the vertical stripe blocks is greater than the percentage of a first batch; and the percentage of the horizontal region of at least one of the horizontal stripe blocks is greater than the percentage of a second batch.

5. The method for analyzing defects as described in claim 4, further comprising, after calculating the total batch defects of the production batch obtained after defect detection, including: If the total number of defects in the batch is greater than a first baseline number, the defect percentage of a single piece in the detection layout is calculated based on the total number of defects in the batch and the total number of defect points in the detection layout; wherein, the specified conditions further include: the total number of defect points is greater than or equal to a first set value, and the defect percentage of a single piece is greater than or equal to a second set value.

6. The method for analyzing defects as described in claim 4, wherein, When the total number of defects in the batch exceeds a second baseline quantity, the specified conditions further include: the vertical area proportion of at least one of the vertical stripe blocks is greater than the proportion of the first batch, wherein the number of the vertical stripe blocks is 3; and when the total number of defects in the batch exceeds a third baseline quantity, the specified conditions further include: the horizontal area proportion of at least one of the horizontal stripe blocks is greater than the proportion of the second batch, wherein the number of the horizontal stripe blocks is 3.

7. The method for analyzing defects as described in claim 1, further comprising, after determining that the detection layout has a cluster of defects: For each of the specified conditions that are met, the corresponding cluster label is recorded.

8. The method for analyzing defects as described in claim 1, further comprising, after determining that the detection layout has a cluster of defects: An anomaly notification is generated for a production batch with defective clusters.

Images