A Model-Based Analysis-Design Approach for Virtual Binocular Vision System With Application to Vision-Based Tactile Sensors

Abstract

The virtual binocular vision system (VBVS) that consists of a single camera and multiple mirrors has been applied to the stereo vision. Several VBVS configurations have been proposed, showing the advantages including the compact structure, simple calibration, and flexible adjustment. However, due to the high complexity of the optical system, few researchers have studied the comprehensive analyzing and designing of the VBVS. To fill this gap, we propose a universal model and the design criteria, and then construct a model-based analysis-design approach of the VBVS with mirrors. First, based on a novel understanding of the reflective optical system, the imaging cone model (ICM) is introduced to describe the geometry properties of an arbitrary VBVS completely, and the VBVS equations are constructed to calculate the imaging cones. Second, the equivalent descriptions of the design criteria are proposed as the expected field of view (FOV) and performances, deriving the analysis method of the VBVS. Then, the analysis method is combined with traversal calculation to establish a design process and its algorithm, which is applied to a novel vision-based tactile sensor with greatly improved design efficiency. Finally, we conduct the measurement of 3D tactile information to demonstrate the performances of the designed sensors.