Extrinsic Calibration Method for Multiple 3D LiDAR Sensors in Autonomous Driving Systems

Extrinsic Parameter Calibration for Multiple 3D LiDAR Sensors Using Plane Extraction and Variance Minimization

This technology is an optimization-based calibration method designed to calibrate extrinsic parameters—the relative positions between multiple 3D LiDAR sensors. It extracts planar information from measured point clouds, calculates initial values through similarity analysis between reference and corresponding planes, and minimizes the variance of measured points.

Existing methods using artificial markers require additional environmental setup and costs. Point cloud registration-based techniques often fail when there is a significant difference in the field of view between sensors, while methods relying on trajectory data are prone to estimation errors.

This technology sequentially performs data collection, plane extraction, corresponding plane detection, and the calculation of initial and final extrinsic parameters, calibrating based on planes that satisfy mathematical conditions such as planarity and normal direction variance. It can be applied to autonomous vehicles and multi-sensor robots, providing an economical solution for immediate on-site sensor alignment without the need for specialized calibration equipment.

‍

‍

Key Features:

• Data collection phase where point clouds are gathered by each 3D LiDAR sensor
• Plane extraction phase where multiple target planes are identified from the point clouds
• Detection phase where one target plane is set as the reference plane to identify corresponding planes from other sensors
• Calculation phase for extrinsic parameters through similarity analysis between corresponding planes and minimization of measured point variance

‍

This invention was developed with support from the Ministry of Science and ICT and the Ministry of Agriculture, Food and Rural Affairs for the training and research of personnel in unmanned agricultural production automation.