Neural and Evolutionary Approaches to Domain Independent Object Recognition


As more and more images are captured in electronic form the need for programs which can find objects of interest in a database of images is increasing. For example, it may be necessary to find all tumors in a database of x-ray images, all cyclones in a database of satellite images or a particular face in a database of photographs. The common characteristic of such problems can be phrased as
Given subpicture_1, subpicture_2...subpicture_n which are examples of the object of interest, find all pictures which contain this object and the locations of all of the objects of interest.
This image shows an example of a problem of this kind. It shows a human retina. We are required to find all of the micro aneurisms and haemorrhages, as indicated by the white squares. (Note: The picture is presented at too coarse a level of resolution for the difference between micro aneurisms and haemorrhages to be evident). In this image we are required to find all occurrences of the head side of the 5 cent coins.

The classical approach to problems of this kind is to examine the objects of interest and determine a set of features, such as number of edges, average brightness and number of perimeter pixels, and write programs to extract the features and hence find the objects. The disadvantage of this approach is that a lot of time is spent looking for useful features and writing the corresponding programs. Furthermore, features and programs that work well on one set of pictures are not very good for another set.

Our approach is to use the pixel values (and more recently pixel statistics) of the objects of interest directly and to learn an object detection program. In this way we hope to achieve domain independence. Learning approaches under investigation include neural networks, genetic algorithms, genetic programming and pulse coded networks.

Current Work

[1]
Minh Luan Nguyen. Evolving Feature Extraction Programs for Weed Detection. PhD Thesis, RMIT, School of Computer Science and Information Technology, 2009. In progress, full time.

[2]
Gayan Wijesinghe. Loops in Genetic Programming. PhD Thesis, RMIT, School of Computer Science and Information Technology, 2006. In progress, full time.

[3]
Brian Lam. Discovery of Texture Features Using Genetic Programming. PhD Thesis, RMIT, School of Computer Science and Information Technology, 2012. (PDF, 2616908 bytes)

Completed Theses

[1]
Vinh Phuong Ha. Texture Detection with One-Class Neural Networks. Honours Thesis, RMIT, School of Computer Science and Information Technology, 2008. (PDF, 1840853 bytes)

[2]
Andrew Innes. Development of an Automated Landmark Detection System for Cephalometric Images. Ph. D. Thesis, RMIT, Department of Mechanical Engineering, Department of Computer Science, 2007.

[3]
Djaka Kurniawan. Image Retrieval Using Texture Segmentation by Genetic Programming. Honours Thesis, RMIT, School of Computer Science and Information Technology, 2006. (PDF, 701767 bytes)

[4]
Gayan Wijesinghe. Landmark Detection on Cephalometric X-Rays using Particle Swarm Optimisation. Honours Thesis, RMIT, School of Computer Science and Information Technology, 2005. (PDF, 494167 bytes)

[5]
Andy Song. Texture Classification: A Genetic Programming Approach. PhD Thesis, RMIT, Department of Computer Science, 2003.

[6]
Mengjie Zhang. A Domain Independent Approach to 2D Object Recognition Based on the Neural and Genetic Paradigms. Ph. D. Thesis, RMIT, Department of Computer Science, 2000. (PDF, 1403828 bytes)

[7]
Nagendra Rai. Pixel Statistics in Neural Networks for Domain Independent Object Detection. Minor Thesis, RMIT, Department of Computer Science, 2000. (Gzipped PostScript, 102 pages, 1583265 bytes)

[8]
Robert Ormiston Smith. Using Pulse Coded Neural Networks to Improve the Accuracy of Object Detection. Minor Thesis, RMIT, Department of Computer Science, 1999. In Progress, Part Time.

[9]
Alf Katz. Blood Cell Discrimination by Computer Vision. Minor Thesis, RMIT, Department of Computer Science, 1999. (PDF, 487044 bytes)

[10]
Tim Burgess. A comparison of Visual Obstacle Detection Methods for Underground Mining Vehicles. Honours Thesis, RMIT, Department of Computer Science, 1998.

[11]
James Newton-Thomas. An Architecture for a Mining Vehicle Controller to perform Autonomous Tramming. Minor thesis, RMIT, Department of Computer Science, 1996.

[12]
Rasha Al-Abbas. A prototype system for off-line signature verification using multilayered feedforward neural networks. Minor thesis, RMIT, Department of Computer Science, Melbourne, March 1994. (Gzipped PostScript, 381996 bytes)

[13]
Dean Couch. Specialized Feed Forward Neural Networks for Image Classification. Honours thesis, RMIT, Department of Computer Science, November 1995.

[14]
Jihan Zhu. An analysis of computer vision and neural network techniques for detecting bacterial growths in microbiological images. Masters thesis, RMIT, Department of Computer Science, Melbourne, June 1992.

Selected Papers Describing This Work

[1]
Andy Song and Vic Ciesielski. Texture segmentation by genetic programming. Evolutionary Computation, 16(4):461-481, Winter 2008.

[2]
Gayan Wijesinghe and Vic Ciesielski. Using restricted loops in genetic programming for image classification. In Proceedings of the 2007 Congress on Evolutionary Computation (CEC2007, pages 4569-4576, Singapore, 2007.

[3]
Vic Ciesielski, Andy Song, and Brian Lam. Visual texture classification and segmentation by genetic programming. In Stefano Cagnoni, Evelyne Lutton, and Gustavo Olague, editors, Genetic and Evolutionary Computation for Image Processing and Analysis, pages 195-213. Hindawi Publishing Company, New York, 2007.

[4]
Vic Ciesielski, Djaka Kurniawan, and Andy Song. Towards image retrieval by texture segmentation with genetic programming. In Ling Guan and Kaoru Hirota, editors, IEEE Symposium on Computational Intelligence in Image and Signal Processing (CIISP 2007),, pages 281-286. IEEE Press, April 2007.

[5]
Vic Ciesielski, Gayan Wijesinghe, Andrew Innes, and Sabu John. Analysis of the superiority of parameter optimization over genetic programming for a difficult object detection problem. In Proceedings of the 2006 Congress on Evolutionary Computation (CEC2006, pages 1264-1271. IEEE Press, July 2006. (PDF)

[6]
Brian Lam and Vic Ciesielski. Applying Genetic Programming to Learn Spatial Differences Between Textures Using A Translation Invariant Representation. In Proceedings of the 2005 Congress on Evolutionary Computation (CEC2005), pages 2202-2209. IEEE Press, September 2005.

[7]
Vic Ciesielski, Andrew Innes, Sabu John, and John Mamutil. Understanding evolved genetic programs for a real world object detection problem. In M. Keijzer et al., editor, Proceedings of the 8th European Conference on Genetic Programming (EuroGP 2005), LNCS 3447, pages 351-360. Springer-Verlag, March 2005. (PDF, 1175418 bytes)

[8]
Andrew Innes, Vic Ciesielski, John Mamutil, and Sabu John. Reducing false alarms using genetic programming in object detection. In Hamid Arabnia, editor, Proceedings of the 2004 International Conference on Artificial Intelligence (IC-AI'04), pages 569-574, June 2004. (PDF, 183765 bytes)

[9]
Brian Lam and Vic Ciesielski. Discovery of human competitive image texture feature extraction programs using genetic programming. In Kalyanmoy Deb et al., editor, Proceedings of the Conference on Genetic and Evolutionary Computation (GECCO2004), volume 2, pages 1114-1125. Springer, June 2004. (PDF, 332343 bytes)

[10]
Andy Song and Vic Ciesielski. Texture analysis by genetic programming. In Garrison Greenwood, editor, Proceedings of the 2004 Congress on Evolutionary Computation (CEC2004), volume 2, pages 2092-2099. IEEE, June 2004. (PDF, 295092 bytes)

[11]
Mengjie Zhang and Victor Ciesielski. Neural networks and genetic algorithms for domain independent multiclass object detection. International Journal of Computational Intelligence and Applications, 4(1):77-108, March 2004.

[12]
Andy Song and Vic Ciesielski. Fast texture segmentation using genetic programming. In Sarker R. et al., editor, Proceedings of the 2003 Congress on Evolutionary Computation (CEC2003), volume 3, pages 2126-2133. IEEE Press, December 2003. (PDF, 206389 bytes)

[13]
Vic Ciesielski, Andrew Innes, John Mamutil, and Sabu John. Landmark detection for cephalometric radiology images using genetic programming. International Journal of Knowledge Based Intelligent Engineering Systems, 7(3):164-171, July 2003. (PDF, 151995 bytes)

[14]
Andrew Innes, Vic Ciesielski, John Mamutil, and Sabu John. Finding templates for cephalometric landmark detection using pulse coupled neural networks and genetic programming. In Hamid Arabnia and Youngsong Mun, editors, Proceedings of the International Conference on Imaging Science, Systems and Technology (CISST'03), volume II, pages 511-517, Las Vegas, June 2003. CSREA Press. (PDF, 241816 bytes)

[15]
Mengjie Zhang, Victor Ciesielski, and Peter Andreae. A domain-independent window approach to multiclass object detection using genetic programming. EURASIP Journal on Applied Signal Processing, Special Issue on Genetic and Evolutionary Computation for Signal Processing and Image Analysis, 2003(8):841-859, July 2003.

[16]
Andrew Innes, Vic Ciesielski, John Mamutil, Sabu John, and Alan Harvey. Landmark detection for cephalometric radiology images using genetic programming. In Ruhul Sarker, Bob McKay, Mitsuo Gen, and Akira Namatame, editors, Proceedings of the 6th Australia-Japan Joint Workshop on Intellignet and Evolutionary Systems, pages 125-132, Canberra, November 2002. (PDF, 398326 bytes)

[17]
Andy Song and Vic Ciesielski. Texture classifiers generated by genetic programming. In Xin Yao, editor, Proceedings of the 2002 Congress on Evolutionary Computation, volume 1, pages 243-248, Honolulu, May 2002. IEEE.

[18]
Andrew Innes, Vic Ciesielski, John Mamutil, and Sabu John. Landmark detection for cephalometric radiology images using pulse coupled neural networks. In Hamid Arabnia and Youngsong Mun, editors, Proceedings of the International Conference on Artificial Intelligence (IC-AI'02), volume 2, pages 511-517, Las Vegas, June 2002. CSREA Press.

[19]
Andy Song, Thomas Loveard, and Vic Ciesielski. Towards genetic programming for texture classification. In Markus Sumpter, Dan Corbett, and Mike Brooks, editors, AI 2001: Advances in Artificial Intelligence, Proceedings of the 14th Australian Joint Conference on Artificial Intelligence,Lecture Notes in Artificial Intelligence 2256, pages 461-472, Berlin, December 2001. Springer-Verlag. (Gzipped PostScript, 12 pages, 305417 bytes)

[20]
Andy Song, Victor Ciesielski, and Peter Rogers. Vision system development by machine learning: Mashing assessment in brewing. Applied Intelligence, Special Issue on Machine Learning in Computer Vision, 15(8):777-795, September 2001. (PDF, 476048 bytes)

[21]
Mengjie Zhang and Victor Ciesielski. Genetic programming for multiple class object detection. In Norman Foo, editor, Proceedings of the 12th Australian Joint Conference on Artificial Intelligence, volume 1747, Lecture Notes in Artificial Intelligence, pages 180-191. Springer, Heidelberg, Dec 1999. (PDF, 257576 bytes)

[22]
Mengjie Zhang and Victor Ciesielski. Using the back propagation algorithm and genetic algorithms to train and refine neural networks for object detection. In T. Bench-Capon, G. Soda, and A.M. Tjoa, editors, Proceedings of 10th International Conference and Workshop on Database and Expert Systems Applications (DEXA99), volume 1677, Lecture Notes in Computer Science, pages 626-635. Springer, Heidelberg, Aug 1999. (Gzipped PostScript, 10 pages, 288763 bytes)

[23]
Victor Ciesielski and Mengjie Zhang. Using genetic algorithms to improve the accuracy of object detection. In Ning Zhong and Lizhu Zhou, editors, Knowledge Discovery and Data Mining --- Research and Practical Experiences. The Third Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD-99), pages 19-24. Tsinghua University Press, Apr 1999. (Gzipped PostScript, 6 pages, 985470 bytes)

[24]
Mengjie Zhang and Victor Ciesielski. Centred weight initialization in neural networks for object detection. In Jenny Edwards, editor, Proceedings of the 22nd Australasian Computer Science Conference, pages 39-50, Singapore, Jan 1999. Springer. (PDF, 258806 bytes)

[25]
Victor Ciesielski and Tak Ho. The synergistic combination of neural and symbolic computation in the recognition of words degraded by noise. In C. Rowles, H. Liu, and N. Foo, editors, Proceedings of the Sixth Australian Joint Artificial Intelligence Conference, pages 131-136, Melbourne, 1993. World Scientific. (Gzipped PostScript, 8 pages, 72862 bytes)

[26]
Victor Ciesielski and Jihan Zhu. A very reliable method for detecting bacterial growths using neural networks. In Proceedings of the International Joint Conference on Neural Networks, pages 62-67, Beijing, Nov 1992. (PDF, 1148290 bytes)

[27]
Victor Ciesielski, Jihan Zhu, John Spicer, and Claire Franklin. A comparison of image processing techniques and neural networks for an automated visual inspection problem. In Anthony Adams and Leon Sterling, editors, Proceedings of the 5th Joint Australian Conference on Artificial Intellignece, pages 147-152, Hobart, Tasmania, 1992. World Scientific. (PDF, 847110 bytes)


Last modified on