Tactile approach to Material Classification - Evaluated with Human Performance

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Knowledge of the physical properties of objects is a requirement to enable effective robotic grasping. Identifying the material from which the object is made, is one such physical property. Characteristics of the material can be retrieved using different sensors; vision-based, tactile based or sound based. Physical contact with materials using tactile sensors can enable the retrieval of detailed information about the material, i.e. compressibility, surface texture and thermal properties. This paper describes a system to classify a wide range of materials based on their thermal properties and surface texture. This system will work towards a combined system using both tactile sensing and vision based sensing. Following acquisition of data from a sophisticated tactile sensor, the system uses principal component analysis (PCA) to extract features from the data which are used to train a two stage Artificial Neural Network (ANN) to classify materials, first into groups and then as individual materials. The system is compared with human performance and the results demonstrate that the proposed system can almost performed as effectively as humans.
LanguageEnglish
Title of host publicationUnknown Host Publication
Pages175-180
Number of pages6
Publication statusPublished - 27 Aug 2014
EventIrish Machine Vision and Image Processing 2014 -
Duration: 27 Aug 2014 → …

Conference

ConferenceIrish Machine Vision and Image Processing 2014
Period27/08/14 → …

Fingerprint

Sensors
Thermodynamic properties
Physical properties
Textures
Compressibility
Principal component analysis
Robotics
Acoustic waves
Neural networks

Cite this

@inproceedings{6ae1be5e96b94dc8b8a1cd2251833f58,
title = "Tactile approach to Material Classification - Evaluated with Human Performance",
abstract = "Knowledge of the physical properties of objects is a requirement to enable effective robotic grasping. Identifying the material from which the object is made, is one such physical property. Characteristics of the material can be retrieved using different sensors; vision-based, tactile based or sound based. Physical contact with materials using tactile sensors can enable the retrieval of detailed information about the material, i.e. compressibility, surface texture and thermal properties. This paper describes a system to classify a wide range of materials based on their thermal properties and surface texture. This system will work towards a combined system using both tactile sensing and vision based sensing. Following acquisition of data from a sophisticated tactile sensor, the system uses principal component analysis (PCA) to extract features from the data which are used to train a two stage Artificial Neural Network (ANN) to classify materials, first into groups and then as individual materials. The system is compared with human performance and the results demonstrate that the proposed system can almost performed as effectively as humans.",
author = "Emmett Kerr and TM McGinnity and SA Coleman",
year = "2014",
month = "8",
day = "27",
language = "English",
pages = "175--180",
booktitle = "Unknown Host Publication",

}

Kerr, E, McGinnity, TM & Coleman, SA 2014, Tactile approach to Material Classification - Evaluated with Human Performance. in Unknown Host Publication. pp. 175-180, Irish Machine Vision and Image Processing 2014, 27/08/14.

Tactile approach to Material Classification - Evaluated with Human Performance. / Kerr, Emmett; McGinnity, TM; Coleman, SA.

Unknown Host Publication. 2014. p. 175-180.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Tactile approach to Material Classification - Evaluated with Human Performance

AU - Kerr, Emmett

AU - McGinnity, TM

AU - Coleman, SA

PY - 2014/8/27

Y1 - 2014/8/27

N2 - Knowledge of the physical properties of objects is a requirement to enable effective robotic grasping. Identifying the material from which the object is made, is one such physical property. Characteristics of the material can be retrieved using different sensors; vision-based, tactile based or sound based. Physical contact with materials using tactile sensors can enable the retrieval of detailed information about the material, i.e. compressibility, surface texture and thermal properties. This paper describes a system to classify a wide range of materials based on their thermal properties and surface texture. This system will work towards a combined system using both tactile sensing and vision based sensing. Following acquisition of data from a sophisticated tactile sensor, the system uses principal component analysis (PCA) to extract features from the data which are used to train a two stage Artificial Neural Network (ANN) to classify materials, first into groups and then as individual materials. The system is compared with human performance and the results demonstrate that the proposed system can almost performed as effectively as humans.

AB - Knowledge of the physical properties of objects is a requirement to enable effective robotic grasping. Identifying the material from which the object is made, is one such physical property. Characteristics of the material can be retrieved using different sensors; vision-based, tactile based or sound based. Physical contact with materials using tactile sensors can enable the retrieval of detailed information about the material, i.e. compressibility, surface texture and thermal properties. This paper describes a system to classify a wide range of materials based on their thermal properties and surface texture. This system will work towards a combined system using both tactile sensing and vision based sensing. Following acquisition of data from a sophisticated tactile sensor, the system uses principal component analysis (PCA) to extract features from the data which are used to train a two stage Artificial Neural Network (ANN) to classify materials, first into groups and then as individual materials. The system is compared with human performance and the results demonstrate that the proposed system can almost performed as effectively as humans.

M3 - Conference contribution

SP - 175

EP - 180

BT - Unknown Host Publication

ER -