US 12,293,481 B2
Method and system for supporting mold design of vehicle body panel
Wonkang Heo, Busan (KR)
Assigned to Hyundai Motor Company, Seoul (KR); and Kia Corporation, Seoul (KR)
Filed by Hyundai Motor Company, Seoul (KR); and Kia Corporation, Seoul (KR)
Filed on Nov. 9, 2022, as Appl. No. 17/983,761.
Claims priority of application No. 10-2021-0154227 (KR), filed on Nov. 10, 2021.
Prior Publication US 2023/0145302 A1, May 11, 2023
Int. Cl. G06T 19/20 (2011.01); G06F 30/15 (2020.01); G06T 7/50 (2017.01)
CPC G06T 19/20 (2013.01) [G06F 30/15 (2020.01); G06T 7/50 (2017.01); G06T 2200/04 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20224 (2013.01); G06T 2219/2004 (2013.01); G06T 2219/2021 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method for supporting a mold design for a vehicle body panel, comprising:
generating a target depth map (a depth map), which is a 2D image, from target design data of a target vehicle body panel, which is a 3D image;
generating a prediction model of an artificial neural network structure;
obtaining a correction value data for a mold design for designing mold data from the target design data by inputting the target depth map to the prediction model of the artificial neural network structure; and
visualizing and/or displaying correction value data for the mold design;
wherein the generating of the prediction model comprises:
respectively converting and generating a learning depth map as a 2D image from learning design data and learning mold data as a 3D image for each of a plurality of vehicle types;
generating a difference map between the learning depth map of the learning design data and the learning depth map of the learning mold data for each of the plurality of vehicle types;
dividing the learning depth map and the difference map corresponding to each of the learning design data and the learning mold data into a plurality of learning patches, respectively;
normalizing the plurality of learning patches;
generating a plurality of learning data from a plurality of normalized learning patches through data augmentation; and
learning the artificial neural network to generate the prediction model by using the plurality of learning data.