Object shape and size recognition from tactile images

Artificial touch sensing system for various Human Computer Interaction (HCI) applications is required to be capable of recognizing various parameters viz. object shape, size, texture and surface. However, only identifying object-shapes is not sufficient for object recognition. It is necessary to distinguish the object shapes according to their dimensions or sizes. Thus in the present work object shapes as well as their sizes are recognized by processing and analysis of tactile images obtained by grasping different objects. In this study, statistical features are extracted from a number of acquired tactile images for classification in their respective object shape and size classes. Both inter-subject and intra-subject classifications are performed using four different classifiers (k-nearest neighbor (kNN), Naïve Bayes classifier, Linear Discriminant Analysis (LDA) and Ensemble) in one-versus-one (OVO) basis, which resulted in high classification accuracy independent of the type of classifier. The mean classification accuracies for inter-subject and intra-subject shape and size recognition are found to be 93%, 87% and 94% and 88% respectively.

[1]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[2]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[3]  Rafael C. González,et al.  Digital image processing using MATLAB , 2006 .

[4]  Thomas J. Watson,et al.  An empirical study of the naive Bayes classifier , 2001 .

[5]  Jinhai Cai,et al.  Hidden Markov Models with Spectral Features for 2D Shape Recognition , 2001, IEEE Trans. Pattern Anal. Mach. Intell..

[6]  Thomas G. Dietterich.  An Experimental Comparison of Three Methods for Constructing Ensembles of Decision Trees: Bagging, Boosting, and Randomization , 2000, Machine Learning.

[7]  Pedro M. Domingos,et al.  On the Optimality of the Simple Bayesian Classifier under Zero-One Loss , 1997, Machine Learning.

[8]  R. Schapire.  The Strength of Weak Learnability , 1990, Machine Learning.

[9]  M. Friedman.  The Use of Ranks to Avoid the Assumption of Normality Implicit in the Analysis of Variance , 1937 .

[10]  Beatrice Lazzerini,et al.  A fuzzy approach to 2D-shape recognition , 2001, IEEE Trans. Fuzzy Syst..

[11]  Thomas Parisini,et al.  Shape from touch by a neural net , 1992, Proceedings 1992 IEEE International Conference on Robotics and Automation.

[12]  Amit Konar,et al.  Object-shape recognition from tactile images using a feed-forward neural network , 2012, The 2012 International Joint Conference on Neural Networks (IJCNN).

[13]  Terry Windeatt,et al.  Accuracy/Diversity and Ensemble MLP Classifier Design , 2006, IEEE Transactions on Neural Networks.

[14]  Shaogang Gong,et al.  Audio- and Video-based Biometric Person Authentication , 1997, Lecture Notes in Computer Science.

[15]  Ravinder Dahiya,et al.  Tactile sensing for robotic applications , 2008 .

[16]  Yoshiki Uchikawa,et al.  A method of 3D object reconstruction by fusing vision with touch using internal models with global and local deformations , 1993, [1993] Proceedings IEEE International Conference on Robotics and Automation.

[17]  Amit Konar,et al.  Object shape recognition from tactile images using regional descriptors , 2012, 2012 Fourth World Congress on Nature and Biologically Inspired Computing (NaBIC).

[18]  R. Polikar,et al.  Ensemble based systems in decision making , 2006, IEEE Circuits and Systems Magazine.

[19]  Thomas G. Dietterich.  What is machine learning? , 2020, Archives of Disease in Childhood.

[20]  D. N. Tibarewala,et al.  Object-shape classification and reconstruction from tactile images using image gradient , 2012, 2012 Third International Conference on Emerging Applications of Information Technology.