EvoWorkshops2006: EvoIASP

8th European Workshop on Evolutionary Computation in Image Analysis and Signal Processing

Evolutionary algorithms have been shown to be tools which can be used effectively in the development of systems (software or hardware) for image analysis and signal processing in complex domains of high industrial and social relevance.

EvoIASP has been the first European event specifically dedicated to the applications of evolutionary computation (EC) to image analysis and signal processing (IASP) and gives European and non-European researchers in those fields, as well as people from industry, an opportunity to present their latest research and to discuss current developments and applications, besides fostering closer future interaction between members of the three scientific communities.

Web address: http://www.evonet.info/eurogp2006

Topics include

Organising Committee

Program Chair
Stefano Cagnoni
cagnoni AT ce DOT unipr DOT it
University of Parma, Italy
EvoWorkshops2006 Chair
Franz Rothlauf
rothlauf AT uni-mannheim DOT de
University of Mannheim, Germany
Local Chair
Anikó Ekárt
ekart AT sztaki DOT hu
Hungarian Academy of Sciences
Publicity Chair
Steven Gustafson
smg AT cs DOT nott DOT ac DOT uk
University of Nottingham, UK

Note: the e-mail addresses are masked for spam protection.

Programme Committee

Lucia Ballerini (Sweden)
Bir Bhanu, University of California, Riverside (USA)
Leonardo Bocchi, University of Florence (Italy)
Alberto Broggi, University of Parma (Italy)
Stefano Cagnoni, University of Parma (Italy)
Ela Claridge, University of Birmingham (UK)
Laura Dipietro, Massachusetts Institute of Technology (USA)
Marc Ebner, University of Wuerzburg (Germany)
Daniel Howard, Qinetiq (UK)
Mario Koeppen, FhG IPK Berlin (Germany)
Evelyne Lutton, INRIA (France)
Gustavo Olague, CICESE (Mexico)
Riccardo Poli, University of Essex (UK)
Stephen Smith University of York (UK)
Giovanni Squillero, Politecnico di Torino (Italy)
Kiyoshi Tanaka, Shinshu University (Japan)
Ankur M. Teredesai, Rochester Institute of Technology (USA)
Andy Tyrrell, University of York (UK)
Leonardo Vanneschi, University of Milan Bicocca (Italy)
Robert Vanyi, Siemens PSE (Hungary)
Mengjie Zhang, Victoria University of Wellington (New Zealand)

Accepted Papers: titles and abstracts

Classifying and Counting Vehicles in Traffic Control Applications
Francesco Archetti, Enza Messina, Daniele Toscani, Leonardo Vanneschi

This paper presents a machine learning system to handle traffic control applications. The input of the system is a set of image sequences coming from a fixed camera. The system can be divided into two main subsystems: the first one, based on Artificial Neural Networks classifies the typology of vehicles moving within a limited image area for each frame of the sequence; the second one, based on Genetic Algorithms, takes as input the frame-by-frame classifications and reconstructs the global traffic scenario by counting the number of vehicles of each typology. This task is particularly hard when the frame rate is low. The results obtained by our system are reliable even for very low frame rate (i.e. four frames per second). Our system is currently used by a company for real-time traffic control.

A Neural Evolutionary Classification Method for Brain-Wave Analysis
Antonia Azzini, Andrea G.B. Tettamanzi

This paper presents an approach to the joint optimization of neural network structure and weights which can take advantage of backpropagation as a specialized decoder. The approach is applied to binary classification of brain waves in the context of brain-computer interfaces.

Image Space Colonization Algorithm
Leonardo Bocchi, Lucia Ballerini

This paper describes an image segmentation method based on an evolutionary approach. Unlike other application of evolutionary algorithms to this problem, our method does not require the definition of a global fitness function. Instead a survival probability for each individual guides the progress of the algorithm. The evolution involves the colonization of a bidimensional world by a number of populations. The individuals, belonging to different populations, compete to occupy all the available space and adapt to the local environmental characteristics of the world. We present various sets of experiments on simulated MR brain images in order to determine the optimal parameter settings. Experimental results on real image are also reported. Images used in this work are color camera photographs of beef meat.

Differential Evolution applied to a multimodal information theoretic optimization problem
Patricia Besson, Jean-Marc Vesin, Vlad Popovici, Murat Kunt

This paper discusses the problems raised by the optimization of a mutual information-based objective function, in the context of a multimodal speaker detection. As no approximation is used, this function is highly nonlinear and plagued by numerous local minima. Three different optimization methods are compared. The Differential Evolution algorithm is deemed to be the best for the problem at hand and, consequently, is used to perform the speaker detection.

Enhancement of an Automatic Fingerprint Identification System Using a Genetic Algorithm and Genetic Programming
Wannasak Wetcharaporn, Nachol Chaiyaratana, Sanpachai Huvanandana

This paper presents the use of a genetic algorithm and genetic programming for the enhancement of an automatic fingerprint identification system (AFIS). The recognition engine within the original system functions by transforming the input fingerprint into a feature vector or fingercode using a Gabor filter bank and attempting to create the best match between the input fingercode and the database fingercodes. A decision to either accept or reject the input fingerprint is then carried out based upon whether the norm of the difference between the input fingercode and the best-matching database fingercode is within the threshold or not. The efficacy of the system is in general determined from the combined true acceptance and true rejection rates. In this investigation, a genetic algorithm is applied during the pruning of the fingercode while the search by genetic programming is executed for the purpose of creating a mathematical function that can be used as an alternative to the norm operator. The results indicate that with the use of both genetic algorithm and genetic programming the system performance has improved significantly.

Artificial Life Models in Lung CTs
Sorin Cristian Cheran, Gianfranco Gargano

With the present paper we introduce a new Computer Assisted Detection method for Lung Cancer in CT images. The algorithm is based on several sub-modules: 3D Region Growing, Active Contour And Shape Models, Centre of Maximal Balls, but the core of our approach are Biological Models of ants known as Artificial Life models. In the first step of the algorithm images undergo a 3D region growing procedure for identifying the ribs cage; then Active Contour Models are used in order to build a confined area for the incoming ants that are deployed to make clean and accurate reconstruction of the bronchial and vascular tree, which is removed from the image just before checking for nodules.

Evolutionary Singularity Filter Bank Optimization for Fingerprint Image Enhancement
Ung-Keun Cho, Jin-Hyuk Hong, Sung-Bae Cho

Singularity is the special feature of fingerprints for identification and classification. Since the performance of singularity extraction depends on the quality of fingerprint images, image enhancement is required to improve the performance. Image enhancement with various image filters might be more useful than a filter, but it is very difficult to find a set of appropriate filters. In this paper, we propose a method that uses the genetic algorithm to find those filters for superior performance of singularity extraction. The performance of the proposed method has been verified by the experiment with NIST DB 4. Moreover, the proposed method does not need any expert knowledge to find the type and order of filters for the target domain, it can be easily applied to other applications of image processing.

Evolutionary Generation of Prototypes for a Learning Vector Quantization Classifier
Luigi Pietro Cordella, Claudio De Stefano, Francesco Fontanella, Angelo Marcelli
(Nominated for Best Paper Award)

An evolutionary computation based algorithm for data classification is presented. The proposed algorithm refers to the learning vector quantization paradigm and is able to evolve sets of points in the feature space in order to find the class prototypes. The more remarkable feature of the devised approach is its ability to discover the right number of prototypes needed to perform the classification task without requiring any a priori knowledge on the properties of the data analyzed. The effectiveness of the approach has been tested on satellite images and the obtained results have been compared with those obtained by using other classifiers.

Automatic Classification of Handsegmented Image Parts with Differential Evolution
Ivanoe De Falco, Antonio Della Cioppa, Ernesto Tarantino

Differential Evolution, a version of an Evolutionary Algorithm, is used to perform automatic classification of handsegmented image parts collected in a seven-class database. Our idea is to exploit it to find the positions of the class centroids in the search space such that for any class the average distance of instances belonging to that class from the relative class centroid is minimized. The performance of the resulting best individual is computed in terms of error rate on the testing set. Then, such a performance is compared against those of other ten classification techniques well known in literature. Results show the effectiveness of the approach in solving the classification task.

Learning High-Level Visual Concepts using Attributed Primitives and Genetic Programming
Krzysztof Krawiec

In this paper, we present a novel approach to genetic learning of high-level visual concepts that works with sets of attributed visual primitives rather than with raster images. The paper presents the approach in detail and verifies it in an experiment concerning locating objects in real-world 3D scenes.

Evolutionary denoising based on an estimation of Hölder exponents with oscillations.
Pierrick Legrand, Evelyne Lutton, Gustavo Olague

In multifractal denoising techniques, the acuracy of the Hölder exponents estimations is crucial for the quality of the outputs. In continuity with the method described in [1], where a wavelet decomposition was used, we investigate the use of another Hölder exponent estimation technique, based on the analysis of the local "oscillations" of the signal. The associated inverse problem to be solved, i.e. finding the signal which is the closest to the initial noisy one but having the prescribed regularity, is then more complex. Moreover, the associated search space is of a different nature as in [1], which necessitates the design of ad-hoc genetic operators. [1] J. Levy Vehel and E. Lutton, "Evolutionary signal enhancement based on Hölder regularity analysis", EVOIASP2001, LNCS 2038, 2001.

Mixed-Integer Evolution Strategies and their Application to Intravascular Ultrasound Image Analysis
Rui Li, Michael T. M. Emmerich, Ernst G.P. Bovenkamp, Jeroen Eggermont, Thomas B¨ack, Jouke Dijkstra, Johan H.C. Reiber
(Nominated for Best Paper Award)

This paper discusses Mixed-Integer Evolution Strategies and their application to an automatic image analysis system for IntraVascular UltraSound (IVUS) images. Mixed-Integer Evolution Strategies can optimize different types of decision variables, including continuous, nominal discrete, and ordinal discrete values. The algorithm is first applied to a set of test problems with scalable ruggedness and dimensionality. The algorithm is then applied to the optimization of an IVUS image analysis system. The performance of this system depends on a large number of parameters that -- so far -- need to be chosen manually by a human expert. It will be shown that a mixed-integer evolution strategy algorithm can significantly improve these parameters compared to the manual settings by the human expert.

The Honeybee Search Algorithm for Three-dimensional Reconstruction
Gustavo Olague, Cesar Puente
(Nominated for Best Paper Award)

This paper investigates the communication system of honeybees with the purpose of obtaining an intelligent approach for three-dimensional reconstruction. A new framework is proposed in which the 3D points communicate between them to achieve an improved sparse reconstruction which could be used reliable in further visual computing tasks. The general ideas that explain the honeybee behaviour are translated into a computational algorithm following the evolutionary computing paradigm. Experiments demonstrate the importance of the proposed communication system to reduce dramatically the number of outliers.

Improving the Segmentation Stage of a Pedestrian Tracking Video-based System by means of Evolution Strategies
´ Oscar P´erez Concha, Miguel ´Angel Patricio, Jes´us Garc´ýa, Jos´e Manuel Molina

Pedestrian tracking video-based systems present particular problems such as the multi fragmentation or low level of compactness of the resultant blobs due to the human shape or movements. This paper shows how to improve the segmentation stage of a video surveillance system by adding morphological post-processing operations so that the subsequent blocks increase their performance. The adjustment of the parameters that regulate the new morphological processes is tuned by means of Evolution Strategies. Finally, the paper proposes a group of metrics to assess the global performance of the surveillance system. After the evaluation over a high number of video sequences, the results show that the shape of the tracks match up more accurately with the parts of interests. Thus, the improvement of segmentation stage facilitates the subsequent stages so that global performance of the surveillance system increases.

Probability Evolutionary Algorithm Based Human Body Tracking
Shuhan Shen, Weirong Chen

A novel evolutionary algorithm called Probability Evolutionary Algorithm (PEA), and a method based on PEA for visual tracking of human body are presented. PEA is inspired by the Quantum computation and the Quantum-inspired Evolutionary Algorithm, and it has a good balance between explora-tion and exploitation with very fast computation speed. In the PEA based hu-man tracking framework, tracking is considered to be a function optimization problem, so the aim is to optimize the matching function between the model and the image observation. Then PEA is used to optimize the matching function. Experiments on synthetic and real image sequences of human motion demon-strate the effectiveness, significance and computation efficiency of the proposed human tracking method.

An Adaptive Stochastic Collision Detection between Deformable Objects using Particle Swarm Optimization
Wang Tianzhu, Li Wenhui, Wang Yi, Ge Zihou, Han Dongfeng

In this paper, we present an efficient method for detecting collisions between highly deformable objects, which is a combination of newly developed stochastic method and Particle Swarm Optimization (PSO) algorithm. Firstly, our algorithm samples primitive pairs within the models to construct a discrete binary search space for PSO, and in this way user can balance performance and detection quality. Besides a particle update process is added in every time step to handle the dynamic environments caused by deformations. Our algorithm is also very general that makes no assumptions about the input models and doesn¡¯t need to store additional data structures either. In the end, we give the precision and efficiency evaluation about the algorithm and find it might be a reasonable choice for complex deformable models in collision detection systems.

Genetic Programming for Automatic Stress Detection in Spoken English
Huayang Xie, Mengjie Zhange, Peter Andreae

This paper describes an approach to the use of genetic programming (GP) for the automatic detection of rhythmic stress in spoken New Zealand English. A linear-structured GP system uses speaker independent prosodic features and vowel quality features as terminals to classify each vowel segment as stressed or unstressed. Error rate is used as the fitness function. In addition to the standard four arithmetic operators, this approach also uses several other arithmetic, trigonometric, and conditional functions in the function set. The approach is evaluated on 60 female adult utterances with 703 vowels and a maximum accuracy of 92.61\% is achieved. The approach is compared with decision trees (DT) and support vector machines (SVM). The results suggest that, on our data set, GP outperforms DT and SVM for stress detection, and GP has stronger automatic feature selection capability than DT and SVM.

Localisation Fitness in GP for Object Detection
Mengjie Zhang, Malcolm Lett
(Nominated for Best Paper Award)

This paper describes two new fitness functions in genetic programming for object detection particularly object localisation problems. Both fitness functions use weighted F-measure of a genetic program and consider the localisation fitness values of the detected object locations, which are the relative weights of these locations to the target object centers. The first fitness function calculates the weighted localisation fitness of each detected object, then uses these localisation fitness values of all the detected objects to construct the final fitness of a genetic program. The second fitness function calculates the average locations of all the detected object centres then calculate the weighted localisation fitness value of the averaged position. The two fitness functions are examined and compared with an existing fitness function on three object detection problems of increasing difficulty. The results suggest that almost all the objects of interest in the large images can be successfully detected by all the three fitness functions, but the two new fitness functions can result in far fewer false alarms and spend much less training time.

Immune Multiobjective Optimization Algorithm for Unsupervised Feature Selection
Xiangrong Zhang, Bin Lu, Shuiping Gou, Licheng Jiao

A feature selection method for unsupervised learning is proposed. Unsupervised feature selection is considered as a combination optimization problem to search for the suitable feature subset and the pertinent number of clusters by optimizing the efficient evaluation criterion for clustering and the number of features selected. Instead of combining these measures into one objective function, we make use of the multiobjective immune clonal algorithm with forgetting strategy to find the more discriminant features for clustering and the most pertinent number of clusters. The results of experiments on synthetic data and real datasets from UCI database show the effectiveness and potential of the method.