基于3D毫米波雷达与相机的改进联合标定方法OA北大核心CSTPCD
Improved Method of Joint Calibration Based on 3D Millimeter Wave Radar and Camera
针对传统的相机和毫米波雷达联合标定方法只能在某一固定平面标定存在数据采集繁琐的问题,基于传统方法提出一种适用于三维空间内的快速标定改进方法.首先进行雷达与图像预处理;其次通过时间同步策略在短时间内完成大量数据的采集与匹配;然后在空间标定上使用随机采样一致性算法(RANSAC)减少离群点的干扰,完成初步的粗标定;最后使用列文伯格-马夸尔特(LM)算法对粗标定结果进一步迭代完成精标定.实验表明:改进的标定方法使整体均方根误差收敛约10 pixels,与使用传统联合标定算法相比,标定精度显著提升;整体标定过程时间约12 min完成.
Traditional camera and millimeter-wave radar joint calibration methods have the problem of cumbersome data collection as they can only calibrate on a fixed plane.A fast calibration improvement method applicable to three-dimensional space is proposed based on the traditional methods.Firstly,radar and image preprocessing are performed.Secondly,a time synchronization strategy is used to collect and match a large amount of data within a short period.Then,the random sample consensus algorithm(RANSAC)is used in spatial calibration to reduce the interference of outliers and achieve preliminary coarse calibration.Finally,the Levenberg-Marquardt(LM)algorithm is used to refine the coarse calibration results for precise calibration iteratively.Experimental results show that the improved calibration method converges to an overall root mean square error of about 10 pixels,significantly improving the calibration accuracy compared to traditional joint calibration algorithms,and the calibration process takes about 12 minutes to complete.
邵剑浩;张远辉;陈科峰;刘康
中国计量大学机电工程学院,浙江杭州 310018
机器视觉时空标定3D毫米波雷达随机采样一致性联合标定
machine visionspatiotemporal calibration3D millimeter wave radarRANSACjoint calibration
《计量学报》 2024 (005)
631-638 / 8
浙江省自然科学基金(LY19F010007)
评论