Soft Computing Techniques in Vision Science

Efficient algorithms have been developed for mining frequent patterns in traditional data where the content of each transaction is definitely known. It is a core technique used in many mining tasks like sequential pattern mining, correlative mining etc. As we know, fuzzy logic provides a mathematical framework that is compatible with poorly quantitative yet qualitatively significant data. Genetic algorithm (GA) is one of the optimization algorithms, which is invented to mimic some of the processes observed in natural evolution. It is a stochastic search technique based on the mechanism of natural selection and natural genetics. That is a general one, capable of being applied to an extremely wide range of problems. In this paper, we have fuzzified our original dataset and have applied various frequent pattern mining techniques on it. Then the result of a particular frequent pattern mining technique that is frequent pattern (FP) growth is taken into consideration in which we apply the concept of GA. Here, the frequent patterns observed are considered as the set of initial population. For the selection criteria, we consider the mean squared residue score rather using the threshold value. It was observed that out of the three fuzzy based frequent mining techniques and the GA based fuzzy FP growth technique the later finds the best individual frequent patterns. Also, the run time of the algorithm and the number of frequent patterns generated is far better than the rest of the techniques used. To extend our findings we have also compared the results obtained by the GA based fuzzy FP growth with an usual approach on a normalized dataset and then applied the concept of FP growth to find the frequent patterns followed by GA. Then by analyzing the result we found that GA based fuzzy FP growth stills yields the best individual frequent patterns.

Proteins are essential for the biological processes in the human body. They can only perform their functions when they fold into their tertiary structure .Protein structure can be determined experimentally and computationally. Experimental methods are time consuming and high-priced and it is not always feasible to identify the protein structure experimentally. In order to predict the protein structure using computational methods, the problem is formulated as an optimization problem and the goal is to find the lowest free energy conformation. In this paper, Genetic Algorithm (GA) based optimization is used. This algorithm is adapted to search the protein conformational search space to find the lowest free energy conformation. Interestingly, the algorithm was able to find the lowest free energy conformation for a test protein (i.e. Met enkephalin) using ECEPP force fields.

Breast Cancer now becomes a common disease among woman in developing as well as developed countries. Many non-invasive methodologies have been used to detect breast cancer. Computer Aided diagnosis through, Mammography is a widely used as a screening tool and is the gold standard for the early detection of breast cancer. The classification of breast masses into the benign and malignant categories is an important problem in the area of computer-aided diagnosis of breast cancer. We present a new method for complete total image of mammogram analysis. A mammogram is analyzed region by region and is classified as normal or abnormal. We present a hybrid technique for extracting features that can be used to distinguish normal and abnormal regions of a mammogram. We describe our classifier technique that uses a unique re-classification method to boost the classification performance. Our proposed hybrid technique comprises decision tree followed by association rule miner shows most proficient and promising performance with high classification rate compared to many other classifiers. We have tested this technique on a set of ground-truth complete total image of mammograms and the result was quite effective.

In a country like India where more number of scripts are in use, automatic identification of printed and handwritten script facilitates many important applications including sorting of document images and searching online archives of document images. In this paper, a multiple feature based approach is presented to identify the script type of the collection of handwritten documents. Eight popular Indian scripts are considered here. Features are extracted using Gabor filters, Discrete Cosine Transform, and Wavelets of Daubechies family. Experiments are performed to test the recognition accuracy of the proposed system at line level for bilingual scripts and later extended to trilingual scripts. We have obtained 100% recognition accuracy for bi-scripts at line level. The classification is done using k-nearest neighbour classifier.

This paper proposes Handwritten Character Recognition method using 2D view and Support Vector Machine (SVM). In this all the character images are normalized using line density based nonlinear normalization, which are further used for feature extraction using two dimensional (2D) views. Each character is considered from five different views, and from each view 16 features are extracted and combined to obtain 80 features. Using these features, Radial Basis Function (RBF) of SVM classifier is trained to separate different classes of characters. Handwritten Character database is used for training and testing of SVM classifier. Support Vector Machine is promising recognition method, which is alternative to Neural Network (NN). Experiments show that the proposed method can provide a good recognition result using Support Vector Machines at a recognition rate 82.33%.

This paper presents a novel approach for the automatic localization of pupil in which multiscale edge detection approach has been employed as a preprocessing step to efficiently localize the pupil followed by a new feature extraction technique which is based on a combination of some multiscale feature extraction techniques. Then pupil is localized using histogram thresholding and filter mask which looks for the region that has the highest probability of having pupil. Here some effort has given for the removal of the effect of hairs on eyelashes and eye brows by the help of a region based averaging filtering. The proposed method is tested on CASIA database. Experimental results show that this method is comparatively accurate.

Image processing has a great impact in the field of medical science. The engineering application spreads over various applications and equally it shows the effective performance. In current research, the medical diagnosis as well as the medical data analysis is most challenging job, as it is very complex task. The complexity is tried to reduced by the help of image processing in this approach. Colloidal Cyst, located in the third ventricle of the human brain is considered in this work for the purpose of detection at the time of diagnosis. Image Processing especially useful for detection, recognition and classification etc. In this chapter, a simple as well as a novel method has been applied for the colloidal cyst detection. The novelty is the structuring element is considered in such a manner that a better result is obtained as compared to traditional and basic morphological methods. The structuring elements used as gradient operator and also has been considered in their complementary forms which produces better results than the initial structuring elements.

This work focuses on a method to automatically localize the heart region, and segment the left ventricle from 4D cardiac cine MR images with no user input. Initial estimate of the heart region, relies on temporal variation of intensity values over one cardiac cycle along with the prior knowledge on spatial geometry. The estimated region is threshold and a sequence of morphological operations applied to crop the region of interest. Level set based segmentation framework developed in this paper is fully automatic, and does not require manually drawn initial contour. The method was evaluated on MRI short axis cine slices of 15 subjects from MICCAI 2009 LV challenge database.

In earlier days, the brain MR images classification and tumor detection was done by human inspection. But this classification method is impractical for large amounts of data and is also non-reproducible. MR images always contain a noise caused by operator performance which leads to serious inaccurate classification. Hence automated classification is preferred for accuracy. The use of artificial intelligence techniques, for instance, neural networks, fuzzy logic, neuro fuzzy have shown great improvement in this field. Hence, in this paper the ANFIS is applied for classification and detection purposes. Decision making was performed in two stages: feature extraction using the principal component analysis (PCA) and the ANFIS trained with the back propagation gradient descent method in combination with the least squares method. The performance of the ANFIS classifier is evaluated in terms of training performance and classification accuracies and the results will confirm that the proposed ANFIS classifier has potential in detecting the tumors.

A new algorithm meant for content based image retrieval (CBIR) is presented in this paper. First the image is segmented into regions using fuzzy c-means algorithm (FCM), and then the local region of image is represented by local binary patterns (LBP), which are evaluated by taking into consideration of local difference between the center pixel and its neighbors. LBP extracts the information based on distribution of edges in an image. Two experiments have been carried out for proving the worth of our algorithm. It is further mentioned that the database considered for experiments are Corel 1000 database (DB1), and Corel 2450 database (DB2). The results after being investigated shows a significant improvement in terms of their evaluation measures as compared to LBP and other existing transform domain techniques.

In this work, a simplified Artificial Neural Network (ANN) based approach for recognition of various objects is explored using multiple features. The objective is to configure and train an ANN to be capable of recognizing an object using a feature set formed by Principal Component Analysis (PCA), Frequency Domain and Discrete Cosine Transform (DCT) components. The idea is to use these varied components to form a unique hybrid feature set so as to capture relevant details of objects for recognition using a ANN which for the work is a Multi Layer Perceptron (MLP) trained with (error) Back Propagation learning.

Protein secondary structure prediction is the method of extracting locally defined protein structures from the sequence of amino acids. It is a challenging and elucidating part of the field of bioinformatics. Several methods are attempting to meet these challenges. But the Artificial Neural Network (ANN) technique is turning out to be the most successful. In this work, an ANN based multi level classifier is designed for predicting secondary structure of the proteins. In this method ANNs are trained to make them capable of recognizing amino acids in a sequence following which from these amino acids secondary structures are derived. Then based on the majority of the secondary structure final structure is derived. This work shows the prediction of secondary structure of proteins employing ANNs though it is restricted initially to four structures only.

Since existing Intrusion Detection Systems (IDS) including misuse detection and anomoly detection are generally incapable of detecting new type of attacks. However, all these systems are capable of detecting intruders with high false alarm rate. It is an urgent need to develop IDS with very high Detection rate and with low False alarm rate. To satisfy this need we propose a new intelligent agent based IDS using Fuzzy Rough Set based outlier detection and Fuzzy Rough set based SVM. In this proposed model we intorduced two different inteligent agents namely feature selection agent to select the required feature set using fuzzy rough sets and decision making agent manager for making final decision. Moreover, we have introduced fuzzy rough set based outlier detection algorithm to detect outliers. We have also adopted Fuzzy Rough based SVM in our system to classify and detect anomalies efficiently. Finally, we have used KDD Cup 99 data set for our experiment, the experimental result show that the proposed intelligent agent based model improves the overall accuracy and reduces the false alarm rate.

Ant Colony Optimization (ACO) algorithm has been applied to data mining recently. In this paper an algorithm for data mining called Ant-Miner is used(Ant Colony Algorithm-based Data Miner). The goal of Ant-Miner is to extract classification rules from data. The algorithm is inspired by both research on the behavior of real ant colonies and some data mining concepts and principles. In this paper the application of Ant Miner Algorithm for classification of data for the weather dataset is proposed using dotnet technology. Result shows that the slightly modified Ant Miner algorithm is capable of classifying the weather dataset more efficiently.

Over the last few years, the interest in studying the Artificial Immune System (AIS) is increasing because of its properties such as uniqueness, recognition of foreigners, anomaly detection, distributed detection, noise tolerance, reinforcement learning and memory. Previous research work has proved that AIS model can apply to behavior-based robotics, but implementation of idiotypic selection in these fields are very few. The present research aims to implement a simple system architecture for a mobile robot navigation problem working with artificial immune system based on the idiotypic effects among the antibodies and the antigens. In this architecture environmental conditions are modeled as antigens and the set of action strategies by the mobile robot are treated as antibodies. These antibodies are selected on the basis of providing the robot with the ability to move in a number of different directions by avoiding obstacles in its environment. Simulation results showed that the robot is capable to reach goal effectively by avoiding obstacles and escape traps in its maze environment.

Robot path planning is about finding a collision free motion from one position to another. Efficient algorithms for solving problems of this type have important applications in areas such as: industrial robotics, computer animation, drug design, and automated surveillance. In this paper, a modified A* algorithm is used for optimizing the path. Different from the approaches that only choose the shortest routes, this method estimates the energy consumption and chooses the most energy efficient routes. As mobile robots are powered by batteries, their energy is limited. Therefore, how to minimize energy consumption is an important problem. The basic idea is to minimize unnecessary stops and turns for mobile robots that cause acceleration and deceleration and consumes significant energy. Simulation results are presented on various environments with different levels of complexity depending on the density of the obstacles. The effectiveness of the proposed approach is evaluated in terms of number of movement steps, path length, energy consumption, number of turns and time. The experimental results show that our approach can provide effective path by reducing the number of turns compared to A*, thus saving energy. All paths generated were optimal in terms of length and smoothness.

Planning of the optimal path has always been the target pursued by many researchers since last five decade. Its application on mobile robot is one of the most important research topics among the scientist and researcher. This paper aims to plan the obstacle-avoiding path for mobile robots based on the Artificial Immune Algorithm (AIA) developed from the immune principle. An immunity algorithm adapting capabilities of the immune system is proposed and enable robot to reach the target object safely and successfully fulfill its task through optimal path and with minimal rotation angle efficiency. Finally, we have compared with the GA based path planning with the AIA based path planning. Simulation results show that the mobile robot is capable of avoiding obstacles, escaping traps, and reaching the goal efficiently and effectively by using AIA than GA.

Facial feature recognition has received much attention among the researchers in computer vision. This paper presents a new approach for facial feature extraction. The work can be broadly classified into two stages, face acquisition and feature extraction. Face acquisition is done by a 4D stereo camera system from Dimensional Imaging and the data is available in ‘obj’ files generated by the camera system. The second stage illustrates extraction of important facial features. The algorithm developed for this purpose is inspired from the natural biological shape and structure of human face. The accuracy of identifying the facial points has been shown using simulation results. The algorithm is able to identify the tip of the nose, the point where nose meets the forehead, and near corners of both the eyes from the faces acquired by the camera system.