Sensitivity evaluation of embedded code detection in imperceptible structured light sensing

Abstract

We address the use of pre-trained primitive-shape detectors for identifying embedded codes in imperceptible structured light (ISL) sensing. The accuracy of the whole sensing system is determined by the performance of such detectors. In training-based methods, generalization of the training results is often an issue, and it is especially so when the work scenario could have substantial variation between the training stage and the operation stage. This paper presents sensitivity evaluation results of embedded code detection in ISL sensing, together with the associated statistical analysis. They show that the scheme of embedding imperceptible codes into normal video projection can be maintained effective despite possible variations on sensing distance, projection-surface orientation, projection-surface shape, projection-surface texture and hardware configuration. The finding indicates the feasibility of integrating the ISL method into robotic systems for operation over a wide domain of circumstances.