基于TOF相机的喷涂工件在线三维重建

喷涂生产线轨迹规划和喷涂机器人自编程技术大都以工件的在线三维测量为基础。近年来TOF相机作为一种高性价比的3D成像设备,被应用于工件测量。针对TOF相机成像视场有限、单次成像只能获取局部轮廓深度信息等问题,提出一种基于工件在位旋转和图形处理器(GPU)加速的TOF点云视频流三维重建算法。该方法在有向距离函数(SDF)点云融合基础上,采用空间散列表存储管理海量点云数据,同时引入快速视觉里程(FOVIS)算法用于姿态估计,以提高喷涂工件在位三维重建算法的效率和鲁棒性。喷涂生产线仿真平台实验表明,在线重建过程中平均帧数可达58 f/s,失败率≤2%,显存占用率25%,为随后的三维测量和喷涂轨迹规划提供完整的点云数据。

TOF camera based 3D object modeling for spraying production line

Spray production line trajectory planning and spray robot self programming technology are based on the workpiece on line measurement. As a cost effective three dimensional imaging device, TOF camera has been applied to workpiece measurement in recent years. Aiming at the problem that existing TOF camera in terms of limited field of imaging view and single image can only obtain local contour depth information. A method of 3D object modeling based on GPU accelerated computing and TOF point cloud streaming is proposed. The main algorithm is signed distance function (SDF) point cloud fusion, then spatial hashing storage is used to manage massive point cloud data, meanwhile, fast odometry from vision (FOVIS) system for pose estimation is introduced to improve the efficiency and robustness of in situ 3D object modeling algorithms of workpices. The experimental results on simulation platform of spraying production line show that the average number of frames in the modeling process can reach 58 frames per second, failure rate less than 2%, graphic memory usage rate about 25%, provides complete point cloud data for subsequent 3D measurement and spray traajectory planning.