EP-4738258-A1 - CALIBRATION METHOD AND APPARATUS FOR SENSOR

EP4738258A1EP 4738258 A1EP4738258 A1EP 4738258A1EP-4738258-A1

Abstract

A method and an apparatus for calibrating a sensor are provided, and relate to the field of sensor technologies. The sensor is one of at least one sensor disposed on a vehicle. The method may include: identifying at least one traffic participant in an environment in which the vehicle is located; obtaining a first data set associated with the at least one traffic participant, where the first data set includes standardized data of the at least one traffic participant; determining a driving status of the vehicle; and if the vehicle does not shake, performing intrinsic parameter calibration on the sensor based on the first data set; or if the vehicle shakes, performing joint calibration on an extrinsic parameter of the at least one sensor of the vehicle based on the first data set. This method is used to implement universal adaptive calibration on a plurality of sensors, so as to improve fusion precision of sensing information of the sensors.

Inventors

LUO, Wentao
LI, Xiaosong
WANG, Yaoyuan
ZHANG, ZIYANG
FU, Sheng
ZHANG, JINGYI

Assignees

Huawei Technologies Co., Ltd.

Dates

Publication Date: 20260506
Application Date: 20240819

Claims (14)

A method for calibrating a sensor, wherein the sensor is one of at least one sensor disposed on a vehicle, and the method comprises: identifying at least one traffic participant in an environment in which the vehicle is located; obtaining a first data set associated with the at least one traffic participant, wherein the first data set comprises standardized data of the at least one traffic participant; determining a driving status of the vehicle; and if the vehicle does not shake, performing intrinsic parameter calibration on the sensor based on the first data set; or if the vehicle shakes, performing joint calibration on an extrinsic parameter of the at least one sensor of the vehicle based on the first data set.
The method according to claim 1, wherein the sensor comprises a camera, and if the vehicle does not shake, the performing intrinsic parameter calibration on the sensor based on the first data set comprises: if the vehicle does not shake within first duration, performing, by using the first data set as a calibration target, intrinsic parameter calibration on the sensor based on a plurality of data frames that are collected within the first duration and that are relative to a same traffic participant.
The method according to claim 1 or 2, wherein the sensor comprises an inertial measurement unit IMU, and if the vehicle does not shake, the performing intrinsic parameter calibration on the sensor based on the first data set comprises: if the vehicle does not shake within second duration, performing, by using the first data set as the calibration target, intrinsic parameter calibration on the sensor based on pose sensing data collected within the second duration.
The method according to any one of claims 1 to 3, wherein if the vehicle shakes, the performing joint calibration on the extrinsic parameter of the at least one sensor of the vehicle based on the first data set comprises: in a process in which the vehicle shakes, performing, by using the first data set as the calibration target, joint calibration on the extrinsic parameter of the at least one sensor of the vehicle based on sensing data of a traffic participant collected in the shake process.
The method according to any one of claims 1 to 4, wherein the determining the driving status of the vehicle comprises: obtaining at least one driving parameter via the at least one sensor, wherein the at least one sensor comprises at least one of the following types of sensors: gyroscope, accelerometer, or magnetometer; and determining the driving status of the vehicle based on the at least one driving parameter.
The method according to any one of claims 1 to 5, wherein the obtaining the first data set associated with the at least one traffic participant comprises: receiving the first data set from a cloud server.
The method according to claim 6, wherein the first data set is obtained based on a target tracking model, and the target tracking model comprises a deep network model.
The method according to any one of claims 1 to 7, wherein the method further comprises: assigning a weight to a confidence level of the at least one sensor based on prior information, to obtain a weight factor associated with the at least one sensor; obtaining a pose estimation result of the vehicle based on the weight factor associated with the at least one sensor and sensing data of the at least one sensor; and performing error compensation on a vehicle-mounted component of the vehicle based on the pose estimation result and a three-dimensional rigid body model.
The method according to claim 8, wherein the vehicle-mounted component comprises an optical sensing device of the vehicle, and the optical sensing device comprises at least one of the following: augmented reality head-up display AR-HUD component, optical projection component, and intelligent ranging component.
An apparatus for calibrating a sensor, wherein the sensor is one of at least one sensor disposed on a vehicle, and the apparatus comprises: an identification unit, configured to identify at least one traffic participant in an environment in which the vehicle is located; an obtaining unit, configured to obtain a first data set associated with the at least one traffic participant, wherein the first data set comprises standardized data of the at least one traffic participant; a determining unit, configured to determine a driving status of the vehicle; an intrinsic parameter calibration unit, configured to: if the vehicle does not shake, perform intrinsic parameter calibration on the sensor based on the first data set; and an extrinsic parameter calibration unit, configured to: if the vehicle shakes, perform joint calibration on an extrinsic parameter of the at least one sensor of the vehicle based on the first data set.
An electronic device, comprising a processor, wherein the processor is coupled to a memory; and the processor is configured to execute a computer program or instructions stored in the memory, to enable the electronic device to perform the method according to any one of claims 1 to 9.
A vehicle, comprising a unit configured to implement the method according to any one of claims 1 to 9.
A readable storage medium, comprising a program or instructions, wherein when the program is executed or the instruction are executed, the method according to any one of claims 1 to 9 is performed.
A computer program product, wherein when the computer program product is run on a computer, the computer is enabled to perform the method according to any one of claims 1 to 9.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to Chinese Patent Application No. 202311076082.3, filed with the China National Intellectual Property Administration on August 24, 2023 and entitled "METHOD AND APPARATUS FOR CALIBRATING SENSOR", which is incorporated herein by reference in its entirety. TECHNICAL FIELD This application relates to the field of sensor technologies, and in particular, to a method and an apparatus for calibrating a sensor. BACKGROUND With rapid development of intelligent vehicle technologies, an intelligent driving function of a vehicle is increasingly improved, and increasingly more sensors are installed on the vehicle. Measurement precision of the vehicle-mounted sensors is critical to implementation of the intelligent driving function and driving safety. Most sensors need to be calibrated after installation to ensure measurement precision of the sensors. Sensor calibration includes intrinsic parameter calibration and extrinsic parameter calibration of the sensor. An intrinsic parameter is a group of parameters that determine an internal mapping relationship of the sensor, and an extrinsic parameter is a parameter that determines a conversion relationship between a coordinate system of the sensor and a specified coordinate system. The intrinsic parameter calibration is used to obtain the internal mapping relationship of the sensor, and the extrinsic parameter calibration is used to convert two or more sensors into a unified spatial coordinate system. Currently, a frequently used method for calibrating a sensor either needs to use a three-dimensional radar sensor as a medium to implement calibration, which lacks a universal online method for calibrating a sensor, or needs a large quantity of data sets to train a deep learning model used for calibration, which has high training difficulty. This is not applicable to real-time calibration on a plurality of types of vehicle-mounted sensors in a vehicle scenario. Therefore, how to design a universal method for calibrating a vehicle-mounted sensor is still an important problem that needs to be urgently resolved. SUMMARY Embodiments of this application provide a method and an apparatus for calibrating a sensor, to implement universal adaptive calibration on a plurality of sensors, so as to improve fusion precision of sensing information of the sensors. According to a first aspect, an embodiment of this application provides a method for calibrating a sensor. The method may be implemented by a calibration apparatus. The calibration apparatus may be an independent device, or may be a chip or a component in the device, or may be software, and the calibration apparatus may be deployed on a vehicle side, a cloud, a roadside device, a remote server, a local server, or the like. A product form and a deployment manner of the calibration apparatus are not limited in this application. The sensor may be one of at least one sensor disposed on a vehicle. The method may include: identifying at least one traffic participant in an environment in which the vehicle is located; obtaining a first data set associated with the at least one traffic participant, where the first data set includes standardized data of the at least one traffic participant; determining a driving status of the vehicle; and if the vehicle does not shake, performing intrinsic parameter calibration on the sensor based on the first data set; or if the vehicle shakes, performing joint calibration on an extrinsic parameter of the at least one sensor of the vehicle based on the first data set. According to the foregoing method, the calibration apparatus can use a scenario element (standardized traffic participant) as a reference to replace a calibration target, to perform intrinsic parameter calibration and/or extrinsic parameter calibration on a vehicle-mounted sensor. The method for calibrating the vehicle-mounted sensor can be decoupled from a medium like a three-dimensional radar sensor, and a large quantity of data sets are not needed to train a deep learning model used for calibration. This can greatly reduce calibration time and labor costs, so that the calibration method is universal, an error of joint calibration on a plurality of sensors is reduced, and fusion precision of sensor information is improved. With reference to the first aspect, in a possible implementation, the sensor includes a camera, and if the vehicle does not shake, the performing intrinsic parameter calibration on the sensor based on the first data set includes: if the vehicle does not shake within first duration, performing, by using the first data set as a calibration target, intrinsic parameter calibration on the sensor based on a plurality of data frames that are collected within the first duration and that are relative to a same traffic participant. With reference to the first aspect, in a possible implementation, the sensor includes an inertial measurement unit IMU, and if the veh