US 11,807,181 B2
Vision-based airbag enablement
Itai Druker, Rosh Haayin (IL); Gila Kamhi, Zichron Yaakov (IL); Michael Baltaxe, Ra'anana (IL); Kobi Nistel, Zoran (IL); Ruben Mergui, Ramat-Gan (IL); Dorel M. Sala, Troy, MI (US); and Lisa A. Fallon, Oxford, MI (US)
Assigned to GM GLOBAL TECHNOLOGY OPERATIONS LLC, Detroit, MI (US)
Filed by GM GLOBAL TECHNOLOGY OPERATIONS LLC, Detroit, MI (US)
Filed on Oct. 27, 2020, as Appl. No. 17/081,428.
Prior Publication US 2022/0126772 A1, Apr. 28, 2022
Int. Cl. B60R 22/00 (2006.01); E05F 15/00 (2015.01); G05D 1/00 (2006.01); G05D 3/00 (2006.01); G06F 7/00 (2006.01); G06F 17/00 (2019.01); B60R 21/015 (2006.01); B60R 21/16 (2006.01); G06T 7/194 (2017.01); G06T 7/70 (2017.01)
CPC B60R 21/01538 (2014.10) [B60R 21/01516 (2014.10); B60R 21/16 (2013.01); G06T 7/194 (2017.01); G06T 7/70 (2017.01); B60R 2021/161 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30268 (2013.01)] 18 Claims
OG exemplary drawing
 
10. An apparatus, comprising:
a camera providing a two-dimensional image of a seating area of a vehicle;
a seat weight sensor providing a sensed weight on a seat in the seating area;
a controller comprising:
a segmentation model performing background segmentation of the two-dimensional image to provide a segmented two-dimensional image of a seated passenger in the seating area of the vehicle appearing in the two-dimensional image;
a pose model determining a pose of the seated passenger in the seating area of the vehicle based upon the segmented two-dimensional image;
a height and weight attribute model determining a seated height and a weight of the seated passenger in the seating area of the vehicle based upon the segmented two-dimensional image and the sensed weight of the seated passenger; and
a controller enabling deployment of an airbag based upon the determined seated height and weight of the seated passenger;
wherein the pose model comprises a neural network trained on a training database of segmented two-dimensional training images, wherein the segmented two-dimensional training images are produced based upon pairs of simultaneously captured two-dimensional and three-dimensional training images.