3D Shape Recognition Based on 1D Signal Processing for Real-Time Applications

Abstract

Nowadays, many researches are being carried out in the field of 3D object recognition. All developed 3D recognition approaches are based on 2D data processing (images) or 3D data (3D models), obtained through 3D reconstruction processes. The main disadvantages of these approaches, especially for real-time applications, are the processing and calculation complexity, which directly affects the recognition time response. In this paper, we propose a new approach for 3D shape recognition adapted to real time applications that is effective for many industrial applications such as: inspection and objects storing, which must be fast enough to satisfy the speed requirements of their application environment. It consists of 3D recognition based on 1D signals, resulting from projecting uncoded structured light lines on the surface of the object. Thereafter for each signal, a group of descriptors are calculated in order to recognize the 3D shape of the object using a classification system. Our approach has been evaluated on a dataset formed by descriptors of real 3D objects calculated from various view-points. The experimental results show that our approach attains an important classification accuracy (99.8%) and requires less than 1ms run-time. It offers an incomplex and clear methodology, and therefore a reduced computational time, since the processing is carried out directly on 1D signals without passing through the 3D model, which makes our system suitable for real time applications.