光学定位系统立体匹配中伪点的自动去除方法

Automatic removal for pseudo markers in optical positioning system

背景:光学定位系统捕捉到的空间标记点的图像存在特征点相似、场景纹理信息少等情况,使得立体匹配存在多义性,即产生匹配伪点。目的:分析匹配过程,提出一种有效的自动去除立体匹配中伪点的方法。方法:根据同一空间点发射的光线位于同一极平面的原理,将匹配特征点的搜索范围限制在一特定区域内,从而减小了搜索区域。然后通过伪点空间坐标的突变性剔除匹配中产生的伪点,从而得到正确的匹配标记点。结果与结论:实验中对装有3个标记点的模拟手术器械工具进行跟踪。对其连续拍摄40组图像,并选取δ=0.0025作为初始匹配阈值进行测试。重建的运动轨迹表明该方法可有效的孤立伪点,同时将所得同一时刻的真实点间的距离与真实距离比较,平均误差为0.2844mm,相对误差为0.48%。实验结果验证了算法的可行性和有效性。

BACKGROUND：The spatial markers can be captured by optical positioning system to form images,in which the feature points are similar and the texture information is difficult to be obtained.There is an ambiguity in stereo matching,which produce pseudo markers.OBJECTIVE：To analyze the process of stereo matching and propose an effective and automatic method for removing the pseudo markers.METHODS：Because all rays emitted from the same marker lie in the same epipolar plane,the searching for matching feature points could be determined in specific region to reduce the searching area.Pseudo markers existing in stereo matching were removed according to the abrupt change of spatial coordinates of pseudo markers.So the correct markers were obtained.RESULTS AND CONCLUSION：A simulation operation tool with three markers was tracked in the experiments.40 groups of images were obtained and δ=0.002 5 was as the initial matching threshold for testing the algorithm.The reconstruction of the movement showed that this method could effectively isolate the pseudo markers.Compared with the real distances,the error of the computed distances of real markers at the same time was 0.284 4 mm,the relative error was 0.48%,which demonstrated the effectiveness and feasibility of the proposed algorithm.