Speckle-Assisted Binocular 3D Vision Sensing Method for Intelligent Cutting

Intelligent 3D vision sensors are fundamental for 3D point cloud acquisition of workpieces in cutting processes. However, existing robotic 3D vision sensors exhibit inherent limitations such as significant measurement errors and constrained fields of view. This paper proposes a binocular stereo vision technique based on active speckle projection, which integrates spatial propagation, view propagation, and an expansion algorithm. The measurement accuracy of the vision sensor is enhanced through a multi-modal cost function and cross-view consistency verification. A hierarchical block-based optimization framework is introduced to improve computational efficiency. The proposed speckle-assisted binocular vision fusion scheme demonstrates both effectiveness and feasibility at both hardware and algorithmic levels.