US 11,869,253 B2
Vehicle environment modeling with a camera
Gideon Stein, Jerusalem (IL); Itay Blumenthal, Tel Aviv (IL); Jeffrey Moskowitz, Tel Aviv (IL); Nadav Shaag, Jerusalem (IL); and Natalie Carlebach, Ra'anana (IL)
Assigned to Mobileye Vision Technologies Ltd., Jerusalem (IL)
Filed by Mobileye Vision Technologies Ltd., Jerusalem (IL)
Filed on Jan. 11, 2023, as Appl. No. 18/095,626.
Application 18/095,626 is a continuation of application No. 17/841,937, filed on Jun. 16, 2022, granted, now 11,568,653.
Application 17/841,937 is a continuation of application No. PCT/US2021/020272, filed on Mar. 1, 2021.
Prior Publication US 2023/0162513 A1, May 25, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06T 7/00 (2017.01); G06V 20/56 (2022.01); G06T 7/50 (2017.01); B60W 40/06 (2012.01)
CPC G06V 20/588 (2022.01) [B60W 40/06 (2013.01); G06T 7/50 (2017.01); B60W 2420/42 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 25 Claims
OG exemplary drawing
 
1. A device for modeling an environment, the device comprising:
a hardware sensor interface to obtain a sequence of unrectified images representative of a road environment, the sequence of unrectified images including an initial unrectified image, a previous unrectified image, and a previous-previous unrectified image; and
processing circuitry to perform operations including:
access the initial unrectified image, a warped previous unrectified image, a warped previous-previous unrectified image, wherein the warped previous unrectified image is based on a first best-fit homography between the previous unrectified image and the initial unrectified image, wherein the warped previous-previous unrectified image is based on a second best-fit homography between the previous-previous unrectified image and the initial unrectified image, and wherein the first and second best-fit homographies are based on a pinhole homography and rectification parameters for a lens used to capture the sequence of unrectified images;
provide the initial unrectified image, the warped previous unrectified image, and the warped previous-previous unrectified image to a trained network to produce a three-dimensional structure of a scene; and
apply the first or second best-fit homography to the three-dimensional structure of the scene to create a model of the road environment.