Feature Detection

A method for automatic facial expression recognition in image sequences, is introduced which make use of Candide wire frame model and active appearance algorithm for tracking, and Bayesian classifier for classification. On the first frame of face image sequence, Candide wire frame model is adapted properly. In subsequent frames of image sequence, facial features are tracked using active appearance algorithm. The algorithm adapts Candide wire frame model to the face in each of the frames and tracks the grid in consecutive video frames over time. Last frame of image sequence corresponds to greatest facial expression intensity. The difference of the node coordinates between the first and the greatest facial expression intensity frame, called the geometrical displacement of Candide wire frame nodes is used as an input to a classifier, which classifies facial expression into one of the class such as happy, surprise, sad, anger, disgust and fear. The experimental results show that the proposed method is better in classification correctness in comparison with binary SVM tree classifier.

Detection and location of the face as well as ex traction of facial features from images is an important stage for numerous facial image interpretation tasks. Detection of facial feature points, such as corners of eyes, lip corners, nostrils from the images are crucial. In this paper a method for automatic facial feature point detection in image sequences, is introduced. The method uses image normalization, and thresholding tech niques to detect 14 facial feature points. Algorithm proposed by Wolf Kienzle is used for face recognition. The detected face region is then divided into 5 relevant regions of interest, each of which is examined separately, further to detect the location of the facial feature points. In each region image is normalized with respect to brightness. Suitable threshold is set for each region, using which image is converted into binary image. Then in each region extreme ends of binary image wiD locate the facial feature points. In the eye region horizontal and vertical histograms are analyzed to detect eyebaUs. This method when tested on Cobn-Kanade database results in recognition rate of 86%. Moreover, when tested on Informatics and Mathematical Modeling (IMM) face database which consists of tilted faces around y axis, we achieved average recognition rate of 83 %.

Localization is a key research topic in mobile robotics, responsible to assist robotic systems equipped with sensors, to navigate with certain autonomy. Unfortunately, the sensors shows frequently reading errors that disturb its location. In this paper, we describe the development of a computer vision system for autonomous navigation of a robot in a simulated environment. The system uses a unattached camera to detect the robot, concentrating the localization problem in the feature extraction of the images. Also, it uses Artificial Intelligence algorithms to determine the path in order to find the best solution. Our results show that the robot was able to follow the path and reach the goal, validating the proposed method.

We discuss an airborne experiment aimed to establish the potential of the PARIS concept (PAssive Reflectometry and Interferometry System) to retrieve small features in the sea surface topography. The date and location were chosen to coincide with a TOPEX/POSEIDON (T/P) overflight. The signals of the Global Positioning System (GPS) reflected off the sea surface are tracked and compared to the directly received ones, to compute the relative delays. The features detected in the peak tracking are likely caused by topographic and sea roughness variations. While very promising, these results open the challenge to use additional information to appropriately separate both contributions.

Abstract. The paper addresses an important and difficult problem of object recognition in poorly constrained environments and with objects having large variability. This research uses genetic programming (GP) to develop automatic object detectors. The task is to detect vehicles in ...

In this paper, novel methods for facial emotion recognition in facial image sequences are presented. Our facial emotional feature detection and extracting based on Active Appearance Models (AAM) with Ekman's Facial Action Coding System (FACS). Our approach to facial emotion recognition lies in the dynamic and probabilistic framework based on Dynamic Bayesian Network (DBN) with Kalman Filter for modeling and understanding the temporal phases of facial expressions in image sequences. By combining AAM and DBN, the proposed method can achieve a higher recognition performance level compare with other facial expression recognition methods. The result on the BioID dataset show a recognition accuracy of more than 90% for facial emotion reasoning using the proposed method.

Detection of boundaries of materials stored in transparent vessels is essential for identifying properties such as liquid level and phase boundaries, which are vital for controlling numerous processes in the industry and chemistry laboratory. This work presents a computer vision method for identifying the boundary of materials in transparent vessels using the graph-cut algorithm. The method receives an image of a transparent vessel containing a material and the contour of the vessel in the image. The boundary of the material in the vessel is found by the graph cut method. In general the method uses the vessel region of the image to create a graph, where pixels are vertices, and the cost of an edge between two pixels is inversely correlated with their intensity difference. The bottom 10% of the vessel region in the image is assumed to correspond to the material phase and defined as the graph and source. The top 10% of the pixels in the vessels are assumed to correspond to the air phase and defined as the graph sink. The minimal cut that splits the resulting graph between the source and sink (hence, material and air) is traced using the max-flow/min-cut approach. This cut corresponds to the boundary of the material in the image. The method gave high accuracy in boundary recognition for a wide range of liquid, solid, granular and powder materials in various glass vessels from everyday life and the chemistry laboratory, such as bottles, jars, Glasses, Chromotography colums and separatory funnels.

In this paper, we propose an original manner to employ a tangential cover algorithm -minDSS -in order to geometrically reconstruct noisy digital contours. To do so, we exploit the representation of graphical objects by maximal primitives we have introduced in previous works. By calculating multi-scale and irregular isothetic representations of the contour, we obtained 1-D (one-dimensional) intervals, and achieved afterwards a decomposition into maximal line segments or circular arcs. By adapting minDSS to this sparse and irregular data of 1-D intervals supporting the maximal primitives, we are now able to reconstruct the input noisy objects into cyclic contours made of lines or arcs with a minimal number of primitives. In this work, we explain our novel complete pipeline, and present its experimental evaluation by considering both synthetic and real image data. We also show that this is a robust approach, with respect to selected references from state-of-the-art, and by considering a multi-scale noise evaluation process.

In this paper, we propose an original manner to employ a tangential cover algorithm -minDSS -in order to geometrically reconstruct noisy digital contours. To do so, we exploit the representation of graphical objects by maximal primitives we have introduced in previous works. By calculating multi-scale and irregular isothetic representations of the contour, we obtained 1-D (one-dimensional) intervals, and achieved afterwards a decomposition into maximal line segments or circular arcs. By adapting minDSS to this sparse and irregular data of 1-D intervals supporting the maximal primitives, we are now able to reconstruct the input noisy objects into cyclic contours made of lines or arcs with a minimal number of primitives. In this work, we explain our novel complete pipeline, and present its experimental evaluation by considering both synthetic and real image data. We also show that this is a robust approach, with respect to selected references from state-of-the-art, and by considering a multi-scale noise evaluation process.

In this paper, the results of a pilot project carried out to estimate the speeds of moving vehicles from video sequences are presented. The software pre-processes video images using gray scale and median filters. To extract moving vehicles from a traffic scene, the difference between three consecutive video frames were taken and filtered through an edge detection filter. Blob counting technique was applied to identify the position and size of the moving vehicles in the images. To estimate the trajectories of moving vehicles, the ...

The Nernst coefficient of CeCoIn5 displays a distinct anomaly at H K ∼ 23 T. This feature is reminiscent of what is observed at 7.8 T in CeRu2Si2, a well-established case of metamagnetic transition. New frequencies are observed in de Haas-van Alphen oscillations when the field exceeds 23 T, which may indicate a modification of the Fermi surface at this field.

The Nernst coefficient of CeCoIn5 displays a distinct anomaly at H K ∼ 23 T. This feature is reminiscent of what is observed at 7.8 T in CeRu2Si2, a well-established case of metamagnetic transition. New frequencies are observed in de Haas-van Alphen oscillations when the field exceeds 23 T, which may indicate a modification of the Fermi surface at this field.

The Nernst coefficient of CeCoIn5 displays a distinct anomaly at H K ∼ 23 T. This feature is reminiscent of what is observed at 7.8 T in CeRu2Si2, a well-established case of metamagnetic transition. New frequencies are observed in de Haas-van Alphen oscillations when the field exceeds 23 T, which may indicate a modification of the Fermi surface at this field.

The animation found in frame/video content is usually characterized by specific color distribution and texture. The existence of animation within a frame can be described by carefully selected features. We analyzed both fully and partially animated frames. Here, we present our initial results related to automatic animation interpretation using different types of video sequences.

This paper presents a novel, robust and flexible method for extracting four mouth features (top of the upper lip, bottom of the lower lip, left and right mouth corners) from facial image sequences. While robustness is referred to subject variability, pose, and image quality, flexibility is begotten by efficient fusion of several information sources and expounding the certainty of the generated results.

At time of crisis, when fraud permanently frightens the basis of modern societies, the existence of effective tools to prevent it, or just to identify it in time, is critical. However, the detection of fraud is naturally impaired (among other issues) by the difficulty on labelling data, due to the cost of identifying and attest fraud. Moreover, the inability to incorporate domain knowledge in the mining process makes classifiers to use inadequate attributes to distinguish entities, ignoring most of existing relevant information, like entities' social relations. In this paper, we propose a new methodology for enriching semi-supervised clustering with information collected through the analysis of social networks. The methodology is then applied on tax fraud detection and assessed by measuring the impact of this enrichment in the accuracy of semi-supervised clustering methods.

Indirect SLAM (Simultaneous Localization and
Mapping) algorithms primarily rely upon image features to
estimate poses and to perform loop closure. Traditional image
key point extraction algorithms fail to detect features in the
image in some cases due to various reasons like low exposure,
lack of contours and hardware constraints. To tackle this issue,
we propose a U-Net neural network based image keypoint
extraction algorithm that works on low cost camera hardware.
We utilize an unconventional approach by applying image
segmentation techniques to extract Shi-Tomashi features for
training the neural network. It is an attempt to building
innovative approaches for Deep learning-based visual odometry
problem. The primary focus of this work is to establish a new
research direction for deep learning based keypoint extraction
tailored for real-time SLAM on low-cost and low-power devices.
We aim to lay the foundations for further investigation into
segmentation-based key-point modeling and its impact on
critical SLAM components like feature tracking, mapping, and
loop closure detection.

Actors are computational entities that communicate with other actors and the environment by passing tokens via theirs input and output ports. Actors have state and parameters and when an actor is fired, input tokens are consumed tokens, output tokens are produced, and the internal state is updates. Actors are connected to form models or applications. It is highly desirable to be able to investigate properties, such as bounded memory and absence of deadlock, for such applications. One key for doing this is to have full knowledge of the production and consumption patterns of the actors. Another issue for actor applications is to generate code for embedded systems. To do so is necessary to have a good understanding of control flow and memory management. Caltrop is novel actor language designed to facilitate the extraction of needed properties in order handle those issues. It is a small domain specific language for implementation of actors. It is designed to be embedded in richer environment, providing data type, operations and function libraries.

Range measuring sensors can play an extremely important role in robot navigation. All range measuring devices rely on a 'detection criterion' made in the presence of noise, to determine when the transmitted signal is considered detected and hence a range reading is obtained. In commonly used sensors, such as laser range finders and polaroid sonars, the criterion under which successful detection is assumed, is kept hidden from the user. However, 'detection decisions' on the presence of noise still take place within the sensor. This paper integrates signal detection probabilities into the map building process which provides the most accurate interpretation of such sensor data. To facilitate range detection analysis, map building with a frequency modulated continuous wave millimetre wave radar (FMCW MMWR), which is able to provide complete received power-range spectra for multiple targets down range is considered. This allows user intervention in the detection process and although not directly applicable to the commonly used 'black-box' type range sensors, provides insight as to how not only range values, but received signal strength values should be incorporated into the map building process. This paper presents two separate methods of map building with sensors which return both range and received signal power information. The first is an algorithm which uses received signal-to-noise power to make an estimates of the range to multiple targets down range, without any signal distribution assumptions. We refer to this as feature detection based on target presence probability (TPP). In contrast to the first method, the second method does use assumptions on the statistics of the signal in target presence and absence scenarios to formulate a probabilistic likelihood detector. This allows for an increased rate of convergence to ground truth. Evidence theory is then introduced to model and update successive observations in a recursive fashion. Both methods are then compared using real MMWR data sets from indoor and outdoor experiments.

Riassunto -Questo lavoro di tesi presenta l'estimazione delle misure delle caratteristice geometriche della propagazione di un incendio di vegetazione, usando un sistema di stereovisione multimodale equipaggiato su un drone (UAV). A partire dalle immagini stereoscopiche nel campo del visibile e dell'infrarosso, i punti 3D sono calcolati e le caratteristiche geometriche dell'incendio quali posizione, velocità di propagazione, altezza, lunghezza, larghezza, angolo d'inclinazione della fiamma e superficie sono estimate. Il primo importante contributo di questa tesi é lo sviluppo di un sistema di stereovisione multimodale portabile da un drone. Questo dispositivo integra telecamere che lavorano nel campo visibile e infrarosso, un computer Rasperry PI, batterie elettriche, GPS e un'unità di misura inerziale. Questo sistema permette di ottenere immagini stereoscopiche multimodali e georeferziate. Il secondo importante contributo di questa tesi é il metodo per l'estimazione delle caratteristiche geometriche di un incendio a partire da immagini stereoscopiche aeree. Il framework per l' estimazione delle caratteristiche geometriche é stato validato su un'auto di dimensioni conosciute e i risultati ottenuti confermano la buona accuratezza del dispositivo. È stato anche testato con successo un fuoco in esterno e senza propagazione. I risultati ottenuti da un incendio di vegetazione in propagazione su un terreno con differenti inclinazioni sono presentati in dettaglio.

This work describes how evolutionary computation can be used to synthesize low-level image operators that detect interesting points on digital images. Interest point detection is an essential part of many modern computer vision systems that solve tasks such as object recognition, stereo correspondence, and image indexing, to name but a few. The design of the specialized operators is posed as an optimization/search problem that is solved with genetic programming (GP), a strategy still mostly unexplored by the computer vision community. The proposed approach automatically synthesizes operators that are competitive with state-of-the-art designs, taking into account an operator's geometric stability and the global separability of detected points during fitness evaluation. The GP search space is defined using simple primitive operations that are commonly found in point detectors proposed by the vision community. The experiments described in this paper extend previous results (Trujill...

In Fall 2008 NASA selected a large international consortium to produce a comprehensive automated feature-recognition system for the Solar Dynamics Observatory (SDO). The SDO data that we consider are all of the Atmospheric Imaging Assembly (AIA) images plus surface magnetic-field images from the Helioseismic and Magnetic Imager (HMI). We produce robust, very efficient, professionally coded software modules that can keep up with the SDO data stream and detect, trace, and analyze numerous phenomena, in-The Solar Dynamics Observatory

The task of detection requires that at least one target component (i.e. ''feature") be detected, while the task of identification requires the detection and integration of multiple features into a recognizable object. Enumeration seems to involve aspects of both feature detection and object identification. As in identification, it requires the detection of multiple features, but as in detection, it does not require the explicit encoding of a global form. Enumeration of briefly presented objects is accurate up to the ''subitizing capacity" of 3-5 items. We discuss the relation of enumeration to visual detection and identification by considering the effect of target visibility on subitizing capacity. We found that while the distribution of enumeration responses changes with contrast, subitizing capacity is generally invariant with contrast until it nears detection threshold. These results suggest that component detection (associated with number estimation) and component integration (associated with subitizing) behaved differentially as contrast was manipulated. We speculate that subitizing capacity is linked to the approximate number of detected features adequate for recognizing shapes.

A letter in the peripheral visual field is much harder to identify in the presence of nearby letters. This is "crowding." Both crowding and ordinary masking are special cases of "masking," which, in general, refers to any effect of a "mask" pattern on the discriminability of a signal. Here we characterize crowding, and propose a diagnostic test to distinguish it from ordinary masking. In ordinary masking, the signal disappears. In crowding, it remains visible, but is ambiguous, jumbled with its neighbors. Masks are usually effective only if they overlap the signal, but the crowding effect extends over a large region. The width of that region is proportional to signal eccentricity from the fovea and independent of signal size, mask size, mask contrast, signal and mask font, and number of masks. At 4 deg eccentricity, the threshold contrast for identification of a 0.32 deg signal letter is elevated (up to six-fold) by mask letters anywhere in a 2.3 deg region, 7 times wider than the signal. In ordinary masking, threshold contrast rises as a power function of mask contrast, with a shallow log-log slope of 0.5 to 1, whereas, in crowding, threshold is a sigmoidal function of mask contrast, with a steep log-log slope of 2 at close spacing. Most remarkably, although the threshold elevation decreases exponentially with spacing, the threshold and saturation contrasts of crowding are independent of spacing. Finally, ordinary masking is similar for detection and identification, but crowding occurs only for identification, not detection. More precisely, crowding occurs only in tasks that cannot be done based on a single detection by coarsely coded feature detectors. These results (and observers' introspections) suggest that ordinary masking blocks feature detection, so the signal disappears, while crowding (like "illusory conjunction") is excessive feature integration -detected features are integrated over an inappropriately large area because there are no smaller integration fields -so the integrated signal is ambiguous, jumbled with the mask. In illusory conjunction, observers see an object that is not there made up of features that are. A survey of the illusory conjunction literature finds that most of the illusory conjunction results are consistent with the spatial crowding described here, which depends on spatial proximity, independent of time pressure. The rest seem to arise through a distinct phenomenon that one might call "temporal crowding," which depends on time pressure ("overloading attention"), independent of spatial proximity.

Modern studies of the Sun involve coordinated observations collected from a collage of instruments on the ground and in orbit. Each instrument has its own constraints, such as field of view, duty cycle, and scheduling and commanding windows, that must both be coordinated during operations and be discoverable for analyses of the resulting data. Details on the observed solar features, i.e. sunspots or filaments, and solar events, i.e. flares or coronal mass ejections, are also incorporated to help guide data discovery and data analysis pipelines. The Heliophysics Coverage Registry (HCR) provides a standards-based system for collecting and presenting observations collected by distributed, ground and space based solar observatories which form an integrated Heliophysics system. The HCR currently supports all instruments on the Interface Region Imaging Spectrograph (IRIS) and Hinode missions as well as associated ground-based observatories. Here we present an overview of the HCR along wit...