Validating vehicle image sensor timestamps

A computer-implemented method for validating image sensor timestamps of a vehicle 600A comprising two or more image sensors, comprising: obtaining, from a first image sensor (e.g. camera) 610A of the vehicle, a plurality of time-stamped first images covering a period while the vehicle is in motion,...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Karim Tarek Mahmoud Elsayed Ahmed Shaban
Format: Patent
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:A computer-implemented method for validating image sensor timestamps of a vehicle 600A comprising two or more image sensors, comprising: obtaining, from a first image sensor (e.g. camera) 610A of the vehicle, a plurality of time-stamped first images covering a period while the vehicle is in motion, (e.g. T1, T2); obtaining, from a second image sensor 620 of the vehicle, a second time-stamped image from within the period; processing the first plurality of images, by a computer vision model, to identify a plurality of features and corresponding positions of an object 660 in an environment and to estimate the position and orientation of the first image sensor for each of the plurality of first images at their associated timestamps; obtaining a calibration value of a position and orientation of the second image sensor relative to the first image sensor; extrapolating a position track 650 and orientation track in the period from the estimated positions and orientations of the first image sensor at the timestamps; estimating, by the computer vision model, the position and orientation of the second image sensor at the second image timestamp using: the extrapolated position track, the extrapolated orientation track, and the calibration value; generating, by the computer vision model, a simulated image at the estimated position and orientation of the second image sensor at the second image timestamp using the identified plurality of features; and calculating an error value by comparing the simulated image and the second image.