【摘要】 传统视觉测量方法难以有效感知透明玻璃表面，而触觉检测受限于采样效率和测量分辨率，难以实现高精度面型重建。为克服这些局限性，提出了一种基于相位测量偏折术与主动触觉探测的视触融合测量方法。视觉测量提供相对表面相位梯度信息，触觉传感提供绝对位置信息，从而有效消除梯度-高度二义性，并抑制由底面玻璃偏折引起的幽灵条纹干扰。实验搭建了一套基于深度相机Realsense D435和UR10机械臂的测量系统，并在平面与近似平面玻璃样品上进行验证。实验结果表明：该方法可实现平面与近似平面玻璃的感知与重建，相比深度相机结合显像剂的方法，提出的方法在凹面玻璃表面重建中RMS降低了80.59%，在凸面玻璃表面重建中RMS降低了96.33%。提出的方法充分利用机器人自身的视觉-触觉传感能力实现了透明玻璃的三维测量与定位，为机器人在透明玻璃的感知与重建的应用中提供了一种解决方案。更多还原

【Abstract】 Traditional vision-based measurement techniques perform poorly on transparent glass surfaces, while tactile sensing is constrained by low sampling efficiency and limited resolution, impeding high-precision surface reconstruction. To address these limitations, a vision-tactile fusion measurement method is presented, integrating phase-measuring deflectometry with active tactile sensing. The vision subsystem supplies relative surface phase-gradient information, whereas the tactile subsystem provides absolute positional measurements; this combination resolves the gradient-height ambiguity and suppresses ghost-fringe artefacts produced by light deflection at the lower glass surface. An experimental setup comprising a Realsense D435 depth camera and a UR10 robotic arm was implemented and evaluated on planar and nearplanar glass specimens. Results demonstrate that the proposed method enables accurate perception and reconstruction of both planar and near-planar transparent glass surfaces. Relative to a depth-camera approach augmented by a developer agent, the proposed method reduced RMS error by(80. 59%) for concave surface reconstruction and by(96. 33%) for convex surface reconstruction. The approach exploits the robot’s inherent vision-tactile capabilities to achieve three-dimensional measurement and localization of transparent glass, offering a practical solution for robotic perception and reconstruction of transparent surfaces.更多还原