基于PnP算法的单目视觉位移测量方法

Displacement Measurement Method of Monocular Vision based on PnP Algorithm

为实现结构位移的快速非接触式测量,建立结合棋盘格靶标和PnP算法的单目视觉位移测量方法。选取L-M、EPNP、IPPE和SQPNP等4种求解算法开展对比研究,以评价不同算法在位移测量中的适用性。通过Python编程进行仿真计算探究特征点数量、特征点定位误差、靶标变形、位移水平对算法求解精度的影响,为后续试验提供依据和指导。搭建三向位移视觉测量样机系统开展竖向及深度位移测量试验,验证PnP算法在位移测量中的可行性。结果表明:L-M、IPPE和SQPNP算法的求解效果优于EPNP算法,位移测量误差<0.5 mm。该研究提供了一种低成本、高精度的工程结构位移监测方法,可为相关工程应用提供参考。

With the purpose of realizing fast non-contact measurement of structural displacement,it was established that a displacement measurement method of monocular vision combining the checkerboard target and the PnP algorithm.To evaluate applicability of different PnP algorithms to displacement measurement,four algorithms including L-M,EPNP,IPPE and SQPNP were selected for comparative research.Simulations were carried out through Python programming to explore the influence of different factors on solution accuracy of different PnP algorithms including the number,the positioning accuracy of feature points,target deformation,and different displacement levels.The simulations’results provided evidence and guidance for the subsequent tests.A visual measurement prototype with three-dimensional displacement was built and planar and depth displacement measurement tests were carried out on it to testified the feasibility of PnP algorithms in the displacement measurement.The results showed that L-M,IPPE and SQPNP algorithms acquired a better result than that of EPNP algorithm and the former’s displacement measurement errors were all less than 0.5 mm.The research provided a method with low cost and high precision for displacement monitoring of engineering structures,and had the value of reference for related engineering applications.