Occluded objects recognition using multiscale features and Hopfield neural networks

Jiann Shu Lee, Chin-Hsing Chen, Yung-Nien Sun, Guan Shu Tseng

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

Abstract

A new method to recognize partially visible two-dimensional objects by means of multiscale features and Hopfield neural network is proposed. The Hopfield network is employed to perform global feature matching. Since the network only guarantee to converge to a local optimal state, the matching results heavily depend on the initial network state determined by the extracted features. To acquire more satisfactory initial matching results, a new feature vector, consisting of the multiscale evolution of the extremal position and magnitude of the wavelet transformed contour orientation, is developed. These features contain the contour primitives information in a multiscale manner, hence good initial states can be obtained. The good initiation enables the method to recognize objects of even heavily occluded, that can not be achieved by using the Nasrabadi-Li’s method. In addition, to make the matching results more insensitive to the threshold value selection of the network, we replace the step-like thresholding function by a ramp-like one. Experimental results have shown that our method is effective even for noisy occluded objects.

Original languageEnglish
Title of host publicationImage Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings
EditorsGianni Vernazza, Leila DeFloriani, Carlo Braccini
PublisherSpringer Verlag
Pages171-176
Number of pages6
ISBN (Print)3540602984, 9783540602989
Publication statusPublished - 1995 Jan 1
Event8th International Conference on Image Analysis and Processing, ICIAP 1995 - San Remo, Italy
Duration: 1995 Sep 131995 Sep 15

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume974
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other8th International Conference on Image Analysis and Processing, ICIAP 1995
CountryItaly
CitySan Remo
Period95-09-1395-09-15

Fingerprint

Hopfield neural networks
Hopfield Neural Network
Object recognition
Object Recognition
Hopfield Network
Feature Matching
Thresholding
Threshold Value
Feature Vector
Wavelets
Converge
Experimental Results
Object

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Lee, J. S., Chen, C-H., Sun, Y-N., & Tseng, G. S. (1995). Occluded objects recognition using multiscale features and Hopfield neural networks. In G. Vernazza, L. DeFloriani, & C. Braccini (Eds.), Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings (pp. 171-176). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 974). Springer Verlag.
Lee, Jiann Shu ; Chen, Chin-Hsing ; Sun, Yung-Nien ; Tseng, Guan Shu. / Occluded objects recognition using multiscale features and Hopfield neural networks. Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings. editor / Gianni Vernazza ; Leila DeFloriani ; Carlo Braccini. Springer Verlag, 1995. pp. 171-176 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{bd6f1aef49f74362af440eefe5d9cead,
title = "Occluded objects recognition using multiscale features and Hopfield neural networks",
abstract = "A new method to recognize partially visible two-dimensional objects by means of multiscale features and Hopfield neural network is proposed. The Hopfield network is employed to perform global feature matching. Since the network only guarantee to converge to a local optimal state, the matching results heavily depend on the initial network state determined by the extracted features. To acquire more satisfactory initial matching results, a new feature vector, consisting of the multiscale evolution of the extremal position and magnitude of the wavelet transformed contour orientation, is developed. These features contain the contour primitives information in a multiscale manner, hence good initial states can be obtained. The good initiation enables the method to recognize objects of even heavily occluded, that can not be achieved by using the Nasrabadi-Li’s method. In addition, to make the matching results more insensitive to the threshold value selection of the network, we replace the step-like thresholding function by a ramp-like one. Experimental results have shown that our method is effective even for noisy occluded objects.",
author = "Lee, {Jiann Shu} and Chin-Hsing Chen and Yung-Nien Sun and Tseng, {Guan Shu}",
year = "1995",
month = "1",
day = "1",
language = "English",
isbn = "3540602984",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "171--176",
editor = "Gianni Vernazza and Leila DeFloriani and Carlo Braccini",
booktitle = "Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings",
address = "Germany",

}

Lee, JS, Chen, C-H, Sun, Y-N & Tseng, GS 1995, Occluded objects recognition using multiscale features and Hopfield neural networks. in G Vernazza, L DeFloriani & C Braccini (eds), Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 974, Springer Verlag, pp. 171-176, 8th International Conference on Image Analysis and Processing, ICIAP 1995, San Remo, Italy, 95-09-13.

Occluded objects recognition using multiscale features and Hopfield neural networks. / Lee, Jiann Shu; Chen, Chin-Hsing; Sun, Yung-Nien; Tseng, Guan Shu.

Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings. ed. / Gianni Vernazza; Leila DeFloriani; Carlo Braccini. Springer Verlag, 1995. p. 171-176 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 974).

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

TY - GEN

T1 - Occluded objects recognition using multiscale features and Hopfield neural networks

AU - Lee, Jiann Shu

AU - Chen, Chin-Hsing

AU - Sun, Yung-Nien

AU - Tseng, Guan Shu

PY - 1995/1/1

Y1 - 1995/1/1

N2 - A new method to recognize partially visible two-dimensional objects by means of multiscale features and Hopfield neural network is proposed. The Hopfield network is employed to perform global feature matching. Since the network only guarantee to converge to a local optimal state, the matching results heavily depend on the initial network state determined by the extracted features. To acquire more satisfactory initial matching results, a new feature vector, consisting of the multiscale evolution of the extremal position and magnitude of the wavelet transformed contour orientation, is developed. These features contain the contour primitives information in a multiscale manner, hence good initial states can be obtained. The good initiation enables the method to recognize objects of even heavily occluded, that can not be achieved by using the Nasrabadi-Li’s method. In addition, to make the matching results more insensitive to the threshold value selection of the network, we replace the step-like thresholding function by a ramp-like one. Experimental results have shown that our method is effective even for noisy occluded objects.

AB - A new method to recognize partially visible two-dimensional objects by means of multiscale features and Hopfield neural network is proposed. The Hopfield network is employed to perform global feature matching. Since the network only guarantee to converge to a local optimal state, the matching results heavily depend on the initial network state determined by the extracted features. To acquire more satisfactory initial matching results, a new feature vector, consisting of the multiscale evolution of the extremal position and magnitude of the wavelet transformed contour orientation, is developed. These features contain the contour primitives information in a multiscale manner, hence good initial states can be obtained. The good initiation enables the method to recognize objects of even heavily occluded, that can not be achieved by using the Nasrabadi-Li’s method. In addition, to make the matching results more insensitive to the threshold value selection of the network, we replace the step-like thresholding function by a ramp-like one. Experimental results have shown that our method is effective even for noisy occluded objects.

UR - http://www.scopus.com/inward/record.url?scp=84947803339&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84947803339&partnerID=8YFLogxK

M3 - Conference contribution

SN - 3540602984

SN - 9783540602989

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 171

EP - 176

BT - Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings

A2 - Vernazza, Gianni

A2 - DeFloriani, Leila

A2 - Braccini, Carlo

PB - Springer Verlag

ER -

Lee JS, Chen C-H, Sun Y-N, Tseng GS. Occluded objects recognition using multiscale features and Hopfield neural networks. In Vernazza G, DeFloriani L, Braccini C, editors, Image Analysis and Processing - 8th International Conference, ICIAP 1995, Proceedings. Springer Verlag. 1995. p. 171-176. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).