US 2021/0263157 A1 · AUTOMATED LABELING SYSTEM FOR AUTONOMOUS DRIVING VEHICLE LIDAR DATA

Abstract

A system and method for using high-end perception sensors such as high-end LIDARs to automatically label sensor data of low-end LIDARs of autonomous driving vehicles is disclosed. A perception system operating with a high-end LIDAR may process sensed data from the high-end LIDAR to detect objects and generate metadata of objects surrounding the vehicle. The confidence level of correctly identifying the objects using the high-end LIDAR may be further enhanced by fusing the data from the high-end LIDAR with data from other sensors such as cameras and radars. The method may use the detected objects and metadata of the detected objects processed from the data captured by the high-end LIDAR and other sensors as ground truth to label data of a same scene captured by a low-end LIDAR mounted on the vehicle. A neural network may use the labeled sensor data from the low-end LIDAR during offline supervised training.

Claims20

Dependent claims

A computer-implemented method for labeling sensor data captured by an autonomous driving vehicles (ADV), the method comprising:

receiving data captured by a reference light detection and range (LIDAR) sensor of the ADV;
detecting a plurality of objects in a driving environment of the ADV from the data received by the LIDAR sensor to generate information for the plurality of objects;
collecting target LIDAR data captured a target LIDAR sensor of the ADV;
and
labeling the target LIDAR data using the information for the plurality of objects as ground truth to generate a plurality of labeled objects for the target LIDAR data.

2 more elements

Dependencies are inferred from the claim text.

Automated labeling system for autonomous driving vehicle lidar data

Life of this patent

Automated labeling system for autonomous driving vehicle lidar data

Life of this patent

Abstract

Claims20