摘要
In this paper,an innovative 3D motion parameters estimation method from stereo image sequences based on infrared(IR) reflective markers is presented.It was assumed that two high speed CCD cameras had been calibrated previously.The method consists of the following steps:1) the coordinate of several markers and depth map for each stereo pair was determined from the sequences of stereo images by relations of markers' coordinate the correspondence between markers was established,2) the 3D motion parameters of the target was computed based upon the matched markers' coordinate,and 3) translated 3D motion parameters estimation into the problem of least square according to the movement model of the object to be measured.Without using line,curve or corner correspondence,this method can calculate the depth of these markers feature easily and quickly in contrast to traditional approaches.The two CCD cameras work on 200 f/s,and each processing cost time is about 3 ms.It was found that,by using several markers and a large number of stereo images,this method can improve the computational speed,robustness and numerical accuracy of the motion parameters in comparison with traditional methods.The virtual simulation experiment was conducted using synthesized stereo image sequences based on 6-DOF motion platform and the experimental results proved the validity of our approach and showed that the translation and rotation precision is up to 0.1 mm and 0.1°.
In this paper, an innovative 3D motion parameters estimation method from stereo image sequences based on infrared（IR） reflective markers is presented. It was assumed that two high speed CCD cameras had been calibrated previously. The method consists of the following steps ： 1 ） the coordinate of several markers and depth map for each stereo pair was determined from the sequences of stereo images by relations of markers＇coordinate the correspondence between markers was established, 2） the 3D motion parameters of the target was computed based upon the matched markers＇coordinate, and 3） translated 3D motion parameters estimation into the problem of least square according to the movement model of the object to be measured. Without using line, curve or corner correspondence, this method can calculate the depth of these markers feature easily and quickly in contrast to traditional approaches. The two CCD cameras work on 200 f/s, and each processing cost time is about 3 ms. It was found that, by using several markers and a large number of stereo images, this method can improve the computational speed, robustness and numerical accuracy of the motion parameters in comparison with traditional methods. The virtual simulation experiment was conducted using synthesized stereo image sequences based on 6-DOF motion platform and the experimental results proved the validity of our approach and showed that the translation and rotation precision is up to 0. 1 mm and 0. 1 o.
基金
Sponsored by the National Natural Science Foundation of China(Grant No. 50275040)
