Determining a False-Negative Rate for Object Detection Based on Vehicle Sensor Output Data

autonomous-driving
A method for estimating object detection false-negative rates in a motor vehicle by matching sensor output data against a database of environmental states, each annotated with the corresponding detection failure probability.
Authors

Luca Parolini

Sebastian Schneider

Published year

2021

Inventors
Luca Parolini, Sebastian Schneider
Patent no.
Assignee
Bayerische Motoren Werke AG

Abstract

A method for determining a false-negative rate of detecting an object of a predetermined object class based on output data of at least one sensor of a motor vehicle and/or an environment model computed from the sensor data. The method assigns the output data to a dataset from a database of stored records, where each record corresponds to a predetermined state of the vehicle’s surroundings and carries information about the false-negative detection rate for objects of the predetermined class. The false-negative rate is then determined from the information of the matched record.

Filings

Jurisdiction Kind Application / Publication no. Date
DE Application DE102021110835.7A 2021-04-28
DE Publication (A1) DE102021110835A1 2022-11-03

Related