US 11,988,630 B2
Method to use artificial intelligence to enhance visual inspection of oxygen sensors
Craig Magera, Simpsonville, SC (US)
Assigned to Robert Bosch GmbH, Stuttgart (DE)
Filed by Robert Bosch GmbH, Stuttgart (DE)
Filed on Oct. 25, 2021, as Appl. No. 17/509,982.
Prior Publication US 2023/0130751 A1, Apr. 27, 2023
Int. Cl. G01N 27/417 (2006.01); G01N 23/046 (2018.01); G01N 23/18 (2018.01); G01N 27/416 (2006.01); G06N 20/00 (2019.01); G06T 11/00 (2006.01)
CPC G01N 27/4175 (2013.01) [G01N 23/046 (2013.01); G01N 23/18 (2013.01); G01N 27/4163 (2013.01); G06N 20/00 (2019.01); G06T 11/003 (2013.01); G01N 2223/04 (2013.01); G01N 2223/401 (2013.01); G01N 2223/419 (2013.01); G01N 2223/426 (2013.01); G01N 2223/645 (2013.01); G01N 2223/646 (2013.01); G06T 2207/10081 (2013.01); G06T 2207/10116 (2013.01); G06T 2207/20081 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A system configured to detect defects in a first oxygen sensor, the system comprising:
an X-ray imaging device configured to capture a production X-ray image of the first oxygen sensor;
an electronic processor configured to use a trained oxygen sensor defect detection model to identify a defect of the first oxygen sensor by
producing a pseudo X-ray image by simulating a projection of a fan beam through computed tomography (CT) data of a second oxygen sensor;
measuring, via the trained oxygen sensor defect detection model, a fan-beam distortion in the production X-ray image;
selecting, via the trained oxygen sensor defect detection model, the pseudo X-ray image based on the fan-beam distortion;
performing a comparison, via the trained oxygen sensor defect detection model, of the production X-ray image to the pseudo X-ray image; and,
classifying, based on the comparison, the production X-ray image as representing an improperly assembled oxygen sensor.