US 11,948,292 B2
Systems and methods for detecting flaws on panels using images of the panels
Andre S. Yoon, Seoul (KR); Sangwoo Shim, Sokcho-si (KR); Yongsub Lim, Gunpo-si (KR); Ki Hyun Kim, Yongin-si (KR); Byungchan Kim, Seoul (KR); JeongWoo Choi, Seoul (KR); and Jongsun Shinn, Seoul (KR)
Assigned to MakinaRocks Co., Ltd., Seoul (KR)
Filed by MakinaRocks Co., Ltd., Seoul (KR)
Filed on Jul. 1, 2020, as Appl. No. 16/918,994.
Claims priority of provisional application 62/869,919, filed on Jul. 2, 2019.
Claims priority of application No. 10-2020-0041480 (KR), filed on Apr. 6, 2020.
Prior Publication US 2021/0004946 A1, Jan. 7, 2021
Int. Cl. G06T 7/00 (2017.01); G06T 1/20 (2006.01)
CPC G06T 7/001 (2013.01) [G06T 1/20 (2013.01); G06T 2207/30108 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A non-transitory computer readable medium storing a computer program, wherein when the computer program is executed by one or more processors of a computing device, the computer program performs operations to provide methods for detecting flaws, and the operations comprise:
extracting a flaw patch from a first flaw image including a flaw, wherein the first flaw image comprises one or more flaw regions and the operation of extracting the flaw patch comprises at least one of:
extracting the flaw patch so that at least one of the flaw regions is included in a center of the flaw patch;
extracting the flaw patch so that the at least one of the flaw regions is included in a boundary of the flaw patch; or
extracting the flaw patch so that a predetermined number or more of flaw regions placed side by side continuously are included in the flaw patch;
preprocessing at least one of the first flaw image or a non-flaw image not including a first flaw;
extracting a non-flaw patch from at least one of the preprocessed first flaw image or non-flaw image;
training a neural network model for classifying patches to flaw or non-flaw with a training data set comprising the flaw patch and the non-flaw patch; and
determining, using the trained neural network model, a flaw threshold number indicating a minimum number of flaw patches included in an image for the image to be classified as a flaw image.