Aggregating Local Descriptors for Epigraphs Recognition

In this paper, we consider the task of recognizing epigraphs in images such as photos taken using mobile devices. Given a set of 17,155 photos related to 14,560 epigraphs, we used a k-NearestNeighbor approach in order to perform the recognition. The contribution of this work is in evaluating state-of-the-art visual object recognition techniques in this specific context. The experimental results conducted show that Vector of Locally Aggregated Descriptors obtained aggregating SIFT descriptors is the best choice for this task.

Developmental commonalities between object and face recognition in adolescence

In the visual perception literature, the recognition of faces has often been contrasted with that of non-face objects, in terms of differences with regard to the role of parts, part relations and holistic processing. However, recent evidence from developmental studies has begun to blur this sharp distinction. We review evidence for a protracted development of object recognition that is reminiscent of the well-documented slow maturation observed for faces. The prolonged development manifests itself in a retarded processing of metric part relations as opposed to that of individual parts and offers surprising parallels to developmental accounts of face recognition, even though the interpretation of the data is less clear with regard to holistic processing. We conclude that such results might indicate functional commonalities between the mechanisms underlying the recognition of faces and non-face objects, which are modulated by different task requirements in the two stimulus domains.

Thermal imaging as a biometrics approach to facial signature authentication

This dissertation develops an image processing framework with unique feature extraction and similarity measurements for human face recognition in the thermal mid-wave infrared portion of the electromagnetic spectrum. The goals of this research is to design specialized algorithms that would extract facial vasculature information, create a thermal facial signature and identify the individual. The objective is to use such findings in support of a biometrics system for human identification with a high degree of accuracy and a high degree of reliability. This last assertion is due to the minimal to no risk for potential alteration of the intrinsic physiological characteristics seen through thermal infrared imaging. The proposed thermal facial signature recognition is fully integrated and consolidates the main and critical steps of feature extraction, registration, matching through similarity measures, and validation through testing our algorithm on a database, referred to as C-X1, provided by the Computer Vision Research Laboratory at the University of Notre Dame. Feature extraction was accomplished by first registering the infrared images to a reference image using the functional MRI of the Brain’s (FMRIB’s) Linear Image Registration Tool (FLIRT) modified to suit thermal infrared images. This was followed by segmentation of the facial region using an advanced localized contouring algorithm applied on anisotropically diffused thermal images. Thermal feature extraction from facial images was attained by performing morphological operations such as opening and top-hat segmentation to yield thermal signatures for each subject. Four thermal images taken over a period of six months were used to generate thermal signatures and a thermal template for each subject, the thermal template contains only the most prevalent and consistent features. Finally a similarity measure technique was used to match signatures to templates and the Principal Component Analysis (PCA) was used to validate the results of the matching process. Thirteen subjects were used for testing the developed technique on an in-house thermal imaging system. The matching using an Euclidean-based similarity measure showed 88% accuracy in the case of skeletonized signatures and templates, we obtained 90% accuracy for anisotropically diffused signatures and templates. We also employed the Manhattan-based similarity measure and obtained an accuracy of 90.39% for skeletonized and diffused templates and signatures. It was found that an average 18.9% improvement in the similarity measure was obtained when using diffused templates. The Euclidean- and Manhattan-based similarity measure was also applied to skeletonized signatures and templates of 25 subjects in the C-X1 database. The highly accurate results obtained in the matching process along with the generalized design process clearly demonstrate the ability of the thermal infrared system to be used on other thermal imaging based systems and related databases. A novel user-initialization registration of thermal facial images has been successfully implemented. Furthermore, the novel approach at developing a thermal signature template using four images taken at various times ensured that unforeseen changes in the vasculature did not affect the biometric matching process as it relied on consistent thermal features.

Fuzzy angel: uma arquitetura distribuída de telemedicina para monitoramento de pacientes com esclerose lateral amiotrófica

The Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease characterized by progressive muscle weakness that leads the patient to death, usually due to respiratory complications. Thus, as the disease progresses the patient will require noninvasive ventilation (NIV) and constant monitoring. This paper presents a distributed architecture for homecare monitoring of nocturnal NIV in patients with ALS. The implementation of this architecture used single board computers and mobile devices placed in patient’s homes, to display alert messages for caregivers and a web server for remote monitoring by the healthcare staff. The architecture used a software based on fuzzy logic and computer vision to capture data from a mechanical ventilator screen and generate alert messages with instructions for caregivers. The monitoring was performed on 29 patients for 7 con-tinuous hours daily during 5 days generating a total of 126000 samples for each variable monitored at a sampling rate of one sample per second. The system was evaluated regarding the rate of hits for character recognition and its correction through an algorithm for the detection and correction of errors. Furthermore, a healthcare team evaluated regarding the time intervals at which the alert messages were generated and the correctness of such messages. Thus, the system showed an average hit rate of 98.72%, and in the worst case 98.39%. As for the message to be generated, the system also agreed 100% to the overall assessment, and there was disagreement in only 2 cases with one of the physician evaluators.

Fuzzy angel: uma arquitetura distribuída de telemedicina para monitoramento de pacientes com esclerose lateral amiotrófica

The Amyotrophic Lateral Sclerosis (ALS) is a neurodegenerative disease characterized by progressive muscle weakness that leads the patient to death, usually due to respiratory complications. Thus, as the disease progresses the patient will require noninvasive ventilation (NIV) and constant monitoring. This paper presents a distributed architecture for homecare monitoring of nocturnal NIV in patients with ALS. The implementation of this architecture used single board computers and mobile devices placed in patient’s homes, to display alert messages for caregivers and a web server for remote monitoring by the healthcare staff. The architecture used a software based on fuzzy logic and computer vision to capture data from a mechanical ventilator screen and generate alert messages with instructions for caregivers. The monitoring was performed on 29 patients for 7 con-tinuous hours daily during 5 days generating a total of 126000 samples for each variable monitored at a sampling rate of one sample per second. The system was evaluated regarding the rate of hits for character recognition and its correction through an algorithm for the detection and correction of errors. Furthermore, a healthcare team evaluated regarding the time intervals at which the alert messages were generated and the correctness of such messages. Thus, the system showed an average hit rate of 98.72%, and in the worst case 98.39%. As for the message to be generated, the system also agreed 100% to the overall assessment, and there was disagreement in only 2 cases with one of the physician evaluators.

Uma interface para o controle de robôs móveis por intermédio de gestos

The main objective of this work was to enable the recognition of human gestures through the development of a computer program. The program created captures the gestures executed by the user through a camera attached to the computer and sends it to the robot command referring to the gesture. They were interpreted in total ve gestures made by human hand. The software (developed in C ++) widely used the computer vision concepts and open source library OpenCV that directly impact the overall e ciency of the control of mobile robots. The computer vision concepts take into account the use of lters to smooth/blur the image noise reduction, color space to better suit the developer's desktop as well as useful information for manipulating digital images. The OpenCV library was essential in creating the project because it was possible to use various functions/procedures for complete control lters, image borders, image area, the geometric center of borders, exchange of color spaces, convex hull and convexity defect, plus all the necessary means for the characterization of imaged features. During the development of the software was the appearance of several problems, as false positives (noise), underperforming the insertion of various lters with sizes oversized masks, as well as problems arising from the choice of color space for processing human skin tones. However, after the development of seven versions of the control software, it was possible to minimize the occurrence of false positives due to a better use of lters combined with a well-dimensioned mask size (tested at run time) all associated with a programming logic that has been perfected over the construction of the seven versions. After all the development is managed software that met the established requirements. After the completion of the control software, it was observed that the overall e ectiveness of the various programs, highlighting in particular the V programs: 84.75 %, with VI: 93.00 % and VII with: 94.67 % showed that the nal program performed well in interpreting gestures, proving that it was possible the mobile robot control through human gestures without the need for external accessories to give it a better mobility and cost savings for maintain such a system. The great merit of the program was to assist capacity in demystifying the man set/machine therefore uses an easy and intuitive interface for control of mobile robots. Another important feature observed is that to control the mobile robot is not necessary to be close to the same, as to control the equipment is necessary to receive only the address that the Robotino passes to the program via network or Wi-Fi.

Fast 2D/3D object representation with growing neural gas

This work presents the design of a real-time system to model visual objects with the use of self-organising networks. The architecture of the system addresses multiple computer vision tasks such as image segmentation, optimal parameter estimation and object representation. We first develop a framework for building non-rigid shapes using the growth mechanism of the self-organising maps, and then we define an optimal number of nodes without overfitting or underfitting the network based on the knowledge obtained from information-theoretic considerations. We present experimental results for hands and faces, and we quantitatively evaluate the matching capabilities of the proposed method with the topographic product. The proposed method is easily extensible to 3D objects, as it offers similar features for efficient mesh reconstruction.

Face recognition using a unified 3D morphable model

We address the problem of 3D-assisted 2D face recognition in scenarios when the input image is subject to degradations or exhibits intra-personal variations not captured by the 3D model. The proposed solution involves a novel approach to learn a subspace spanned by perturbations caused by the missing modes of variation and image degradations, using 3D face data reconstructed from 2D images rather than 3D capture. This is accomplished by modelling the difference in the texture map of the 3D aligned input and reference images. A training set of these texture maps then defines a perturbation space which can be represented using PCA bases. Assuming that the image perturbation subspace is orthogonal to the 3D face model space, then these additive components can be recovered from an unseen input image, resulting in an improved fit of the 3D face model. The linearity of the model leads to efficient fitting. Experiments show that our method achieves very competitive face recognition performance on Multi-PIE and AR databases. We also present baseline face recognition results on a new data set exhibiting combined pose and illumination variations as well as occlusion.

Automatically Evolving Rotation-invariant Texture Image Descriptors by Genetic Programming

In computer vision, training a model that performs classification effectively is highly dependent on the extracted features, and the number of training instances. Conventionally, feature detection and extraction are performed by a domain-expert who, in many cases, is expensive to employ and hard to find. Therefore, image descriptors have emerged to automate these tasks. However, designing an image descriptor still requires domain-expert intervention. Moreover, the majority of machine learning algorithms require a large number of training examples to perform well. However, labelled data is not always available or easy to acquire, and dealing with a large dataset can dramatically slow down the training process. In this paper, we propose a novel Genetic Programming based method that automatically synthesises a descriptor using only two training instances per class. The proposed method combines arithmetic operators to evolve a model that takes an image and generates a feature vector. The performance of the proposed method is assessed using six datasets for texture classification with different degrees of rotation, and is compared with seven domain-expert designed descriptors. The results show that the proposed method is robust to rotation, and has significantly outperformed, or achieved a comparable performance to, the baseline methods.

Anomaly Detection in Noisy Images

Finding rare events in multidimensional data is an important detection problem that has applications in many fields, such as risk estimation in insurance industry, finance, flood prediction, medical diagnosis, quality assurance, security, or safety in transportation. The occurrence of such anomalies is so infrequent that there is usually not enough training data to learn an accurate statistical model of the anomaly class. In some cases, such events may have never been observed, so the only information that is available is a set of normal samples and an assumed pairwise similarity function. Such metric may only be known up to a certain number of unspecified parameters, which would either need to be learned from training data, or fixed by a domain expert. Sometimes, the anomalous condition may be formulated algebraically, such as a measure exceeding a predefined threshold, but nuisance variables may complicate the estimation of such a measure. Change detection methods used in time series analysis are not easily extendable to the multidimensional case, where discontinuities are not localized to a single point. On the other hand, in higher dimensions, data exhibits more complex interdependencies, and there is redundancy that could be exploited to adaptively model the normal data. In the first part of this dissertation, we review the theoretical framework for anomaly detection in images and previous anomaly detection work done in the context of crack detection and detection of anomalous components in railway tracks. In the second part, we propose new anomaly detection algorithms. The fact that curvilinear discontinuities in images are sparse with respect to the frame of shearlets, allows us to pose this anomaly detection problem as basis pursuit optimization. Therefore, we pose the problem of detecting curvilinear anomalies in noisy textured images as a blind source separation problem under sparsity constraints, and propose an iterative shrinkage algorithm to solve it. Taking advantage of the parallel nature of this algorithm, we describe how this method can be accelerated using graphical processing units (GPU). Then, we propose a new method for finding defective components on railway tracks using cameras mounted on a train. We describe how to extract features and use a combination of classifiers to solve this problem. Then, we scale anomaly detection to bigger datasets with complex interdependencies. We show that the anomaly detection problem naturally fits in the multitask learning framework. The first task consists of learning a compact representation of the good samples, while the second task consists of learning the anomaly detector. Using deep convolutional neural networks, we show that it is possible to train a deep model with a limited number of anomalous examples. In sequential detection problems, the presence of time-variant nuisance parameters affect the detection performance. In the last part of this dissertation, we present a method for adaptively estimating the threshold of sequential detectors using Extreme Value Theory on a Bayesian framework. Finally, conclusions on the results obtained are provided, followed by a discussion of possible future work.

RICH AND EFFICIENT VISUAL DATA REPRESENTATION

Increasing the size of training data in many computer vision tasks has shown to be very effective. Using large scale image datasets (e.g. ImageNet) with simple learning techniques (e.g. linear classifiers) one can achieve state-of-the-art performance in object recognition compared to sophisticated learning techniques on smaller image sets. Semantic search on visual data has become very popular. There are billions of images on the internet and the number is increasing every day. Dealing with large scale image sets is intense per se. They take a significant amount of memory that makes it impossible to process the images with complex algorithms on single CPU machines. Finding an efficient image representation can be a key to attack this problem. A representation being efficient is not enough for image understanding. It should be comprehensive and rich in carrying semantic information. In this proposal we develop an approach to computing binary codes that provide a rich and efficient image representation. We demonstrate several tasks in which binary features can be very effective. We show how binary features can speed up large scale image classification. We present learning techniques to learn the binary features from supervised image set (With different types of semantic supervision; class labels, textual descriptions). We propose several problems that are very important in finding and using efficient image representation.

A three-dimensional representation method for noisy point clouds based on growing self-organizing maps accelerated on GPUs

The research described in this thesis was motivated by the need of a robust model capable of representing 3D data obtained with 3D sensors, which are inherently noisy. In addition, time constraints have to be considered as these sensors are capable of providing a 3D data stream in real time. This thesis proposed the use of Self-Organizing Maps (SOMs) as a 3D representation model. In particular, we proposed the use of the Growing Neural Gas (GNG) network, which has been successfully used for clustering, pattern recognition and topology representation of multi-dimensional data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models, without considering time constraints. It is proposed a hardware implementation leveraging the computing power of modern GPUs, which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). The proposed methods were applied to different problem and applications in the area of computer vision such as the recognition and localization of objects, visual surveillance or 3D reconstruction.