3D Shape Recognition Based on Uncoded Structured Light Using ANN Classifier

Abstract

In this article we present a suite of our research work on object shape recognition based on uncoded structured light that is used to acquire 3D information in the form of lines distorted by the object’s relief, from these lines we extract the 1D signals corresponding to the object. These signals are used to extract the features of the 3D object shape. In this article we propose a new approach to determine 3D shape descriptors using 1D signals. And to improve the performance of the recognition system based on 1D signal processing, we thought about implementing more information on the object shape by adding to the characteristic vector other descriptors calculated in frequency domain called frequency-based descriptors. Once the shape descriptors are calculated, we proceed to the classification of descriptor vectors in order to recognize the different shapes of 3D objects. The results of the proposed approach allow 3D object recognition with an accuracy of 99.6% using the ANN classifier on a database formed by 10 objects. We present a comparison between the results obtained by applying our approach to different databases made up of 6, 7 and 10 objects and treat these results according to two classifiers KNN (K Nearest Neighbors) and ANN (Artificial Neural Network).