Extrinsic calibration between multiple long baseline 4D mmWave radars for V2X

• 主要作者: Yang, Zihan
• 其他作者: Wang Dan Wei
• 格式: Thesis-Master by Coursework
• 語言: English
• 出版: Nanyang Technological University 2024
• 主題: Engineering; Extrinsic calibration; Long baseline; Feature detection
• 在線閱讀: https://hdl.handle.net/10356/173878

4D mmWave Radar(x, y,z, velocity), as a new sensor, possesses significant potential. Compared to LiDAR, it boasts advantages such as longer distance ranging, lower cost, and greater accuracy in adverse weather conditions. However, as a key aspect of multi-radar fusion, there is a scarcity of research on the extrinsic calibration of long baseline multi-radar systems. The main reasons can be summarized in three aspects: 1) As a new type of sensor, 4D radar has entered the market relatively recently. 2) Existing studies mostly focus on short baseline scenarios such as unmanned vehicles, with little attention given to long baseline and large viewpoint difference scenarios. 3) The point cloud from 4D radar is sparse and noisy, thus it is challenging to locate the target and extract the feature. To solve these problems, LB-R2R-Calib is proposed. The main contributions of this dissertation are: 1) A new target is introduced, which is suitable for long baseline and large viewpoint difference scenarios. 2) Based on some important characteristics of 4D radar point cloud, an algorithm for rapidly locating targets within the point cloud were proposed. Experiments with two 4D radars were conducted in real environments with four configurations. Both quantitative and qualitative analysis are implemented to prove that the method is accurate and robust. The projection error is only 0.13m at a distance of 50 meters.