US 11,886,541 B2
Systems and methods for generating synthetic images of a training database
Walter Laplante, Bloomfield Hills, MI (US); Francis Maslar, Grosse Ile, MI (US); Harry Kekedjian, Tecumseh (CA); Nagharajhan KPK, Salem (IN); and Siva Sankar R., Chennai (IN)
Assigned to Ford Motor Company, Dearborn, MI (US)
Filed by Ford Motor Company, Dearborn, MI (US)
Filed on Nov. 17, 2021, as Appl. No. 17/528,866.
Prior Publication US 2023/0153385 A1, May 18, 2023
Int. Cl. G06T 17/20 (2006.01); G06F 18/214 (2023.01); G06T 15/04 (2011.01); G06V 20/10 (2022.01); G06F 18/2431 (2023.01)
CPC G06F 18/214 (2023.01) [G06F 18/2431 (2023.01); G06T 15/04 (2013.01); G06T 17/20 (2013.01); G06V 20/10 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method for training an image analysis system comprising:
generating one or more nominal images of at least a portion of a digital twin of an environment, wherein the digital twin is a virtual representation of the environment, and wherein the one or more nominal images are based on a field of view of an image sensor of the digital twin and one or more nominal characteristics of one or more components of the digital twin;
defining one or more anomalous characteristics of the one or more components;
generating one or more anomalous images of the portion of the digital twin of the environment based on the field of view and the one or more anomalous characteristics;
performing a tessellation routine and a texture mapping routine on the one or more nominal images and the one or more anomalous images to generate a plurality of synthetic images; and
labeling, for each synthetic image from among the plurality of synthetic images, the synthetic image as one of an anomalous type, a nominal type, or a combination thereof.