基于三维重建的动物体尺获取原型系统

Prototype System of Shape Measurements of Animal Based on 3D Reconstruction

为提高现有动物体尺获取技术的效率和自动化程度,提出双深度摄像头动物实时三维重建系统,进而进行动物体尺获取。基于随机采样一致性算法的圆球标定方法对摄像头外参数进行自动标定,再利用外参数将同步获取的点云数据进行配准达到实时重建,最后采用优化拾取机制后的交互式测量方法得到体尺。选取Xtion PRO作为点云数据采集设备并以猪标本作为重建对象,利用高精度激光扫描仪的重建数据与该系统重建结果进行了对比试验,结果表明圆球标定算法能够全自动快速地获取摄像头外参数,用该参数配准后的数据平均误差在7.50 mm以内,该系统以15帧/s的速度重建猪体全身,获取误差在4%以内的体尺信息,达到农业上动物体尺测量的一般要求,该系统可用于动物体尺测量。

In order to improve the accuracy, efficiency and automatic degree of existing acquisition techniques of animal’s conformation, based on two depth cameras, a real time system which can reconstruct the animal for body measurement, an automatic ball target based method for the extrinsic calibration of reconstruction system which is based on RANSAC were put forward. The extrinsic parameters can be used for registering point clouds acquired synchronously from depth camera in real time. And then, interactive measuring method which is optimized in picking mode for body measurement was used to get shape traits of animal. Xtion PRO was chosen as data acquisition equipment, and taking the pig specimen as an example, point clouds data acquired from high precision laser scanner was applied to evaluation of the system. The results show that the ball-target-based method can get the extrinsic parameters of depth camera automatically; the average error of registered data is less than 7.50mm; the speed of reconstructing the whole pig can reach in excess of 15f/s; body measurement error is less than 4%. So the system can be applied to body measurements in agricultural field.