No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Segmentation from motion of non-rigid objects by neuronal lateral interaction
Abstract
The problem we are stating is the discrimination of non-rigid objects capable of holding our attention in a scene. Motion allows gradually obtaining all moving objects shapes. We introduce an algorithm that fuses spots obtained by means of neuronal lateral interaction in accumulative computation.
... It is convenient to highlight that features associated to pixels are easier to obtain than features associated to elements. In the latter case an additional problem arises, namely the one of segmenting the objects present in the scene [7,13,43,44]. Thus, obtaining features associated with image pixels as well as scene segmentation into different elements are fundamental to the solution adopted in this paper. ...
... @BULLET At pixel level the chosen features are grey level, motion detection, velocity and acceleration. This decision is supported by the fact that these features have largely been considered as interesting in lots of applications10111213141516 where selection and segmentation are crucial. @BULLET At object level the selection has been much more complicate . ...
... Later on these objects parts (also called zones, patches or spots) will be treated as whole objects. In previous papers from our research team some algorithms for the segmentation of the image in different objects have been proposed based on the detection of motion, the permanency effect and lateral interaction [13,26]. Thus, based on the satisfactory results of the algorithms commented, we propose, in order to solve the current problem, to incorporate mechanisms of charge and discharge (based on the permanency effect), as well as mechanisms of lateral interaction. ...
A new computational architecture of dynamic visual attention is introduced in this paper. Our approach defines a model for the generation of an active attention focus on a dynamic scene captured from a still or moving camera. The aim is to obtain the objects that keep the observer’s attention in accordance with a set of predefined features, including color, motion and shape. The solution proposed to the selective visual attention problem consists in decomposing the input images of an indefinite sequence of images into its moving objects, by defining which of these elements are of the user’s interest, and by keeping attention on those elements through time. Thus, the three tasks involved in the attention model are introduced. The Feature-Extraction task obtains those features (color, motion and shape features) necessary to perform object segmentation. The Attention-Capture task applies the criteria established by the user (values provided through parameters) to the extracted features and obtains the different parts of the objects of potential interest. Lastly, the Attention-Reinforcement task maintains attention on certain elements (or objects) of the image sequence that are of real interest.
... Therefore, techniques for estimating the velocity field (optical flow field) are of great interest for enhancing scene segmentation (Gautama and Van Hulle, 2002). In previous works, our research team has taken advantage of motion information in segmenting (Fernández-Caballero et al., 2001 ) and classifying moving objects () through accumulative computation (Fernández et al., 1995) and algorithmic lateral inhibition (Mira et al., 2004) by means of a series of charge maps related to pixel-wise ''motion presence'' information. But, up to this moment, we had not directly used calculated motion parameters—e.g. ...
... We work with 256 grey-level input images and transform them into a lower number of levels n. In particular, good results are obtained with eight levels in normal illumination indoor and outdoor scenes (Caballero et al., 2001). A higher value rarely gives better results, whilst lower values (say, 2 or 4) may be used for night vision . ...
... Our approach starts obtaining the objectÕs parts from their grey level bands. Later on these objects parts (also called zones, patches or spots) will be treated as whole objects incorporating lateral interaction methods (Caballero et al., 2001 Ló pez et al., 2003). In this proposal, the patches present in the Working Memory are constructed from the Interest Map compared with the Grey Level Bands Map. ...
A new computational model for active visual attention is introduced in this paper. The method extracts motion and shape features from video image sequences, and integrates these features to segment the input scene. The aim of this paper is to highlight the importance of the motion features present in our algorithms in the task of refining and/or enhancing scene segmentation in the method proposed. The estimation of these motion parameters is performed at each pixel of the input image by means of the accumulative computation method, using the so-called permanency memories. The paper shows some examples of how to use the “motion presence”, “module of the velocity” and “angle of the velocity” motion features, all obtained from accumulative computation method, to adjust different scene segmentation outputs in this dynamic visual attention method.
... [24], Ettlinger-Tor in Karlsruhe (http: //i21www.ira.uka.de/image_sequences/) [24,33], TwoWalkNew (University of Maryland) [32,33]. ...
... [24], Ettlinger-Tor in Karlsruhe (http: //i21www.ira.uka.de/image_sequences/) [24,33], TwoWalkNew (University of Maryland) [32,33]. ...
Many researchers have explored the relationship between recurrent neural networks and finite state machines. Finite state machines constitute the best-characterized computational model, whereas artificial neural networks have become a very successful tool for modeling and problem solving. The neurally-inspired lateral inhibition method, and its application to motion detection tasks, have been successfully implemented in recent years. In this paper, control knowledge of the algorithmic lateral inhibition (ALI) method is described and applied by means of finite state machines, in which the state space is constituted from the set of distinguishable cases of accumulated charge in a local memory. The article describes an ALI implementation for a motion detection task. For the implementation, we have chosen to use one of the members of the 16-nm Kintex UltraScale+ family of Xilinx FPGAs. FPGAs provide the necessary accuracy, resolution, and precision to run neural algorithms alongside current sensor technologies. The results offered in this paper demonstrate that this implementation provides accurate object tracking performance on several datasets, obtaining a high F-score value (0.86) for the most complex sequence used. Moreover, it outperforms implementations of a complete ALI algorithm and a simplified version of the ALI algorithm—named “accumulative computation”—which was run about ten years ago, now reaching real-time processing times that were simply not achievable at that time for ALI.
... Nuestro grupo lleva varios años desarrollando sus investigaciones en el estudio del movimiento en secuencias de imágenes. Todas estas investigaciones nos han permitido afrontar aplicaciones como el reconocimiento de siluetas de objetos móviles en entornos ruidosos [Fer03a], la clasificación de móviles según sus características del movimiento, como su velocidad o su aceleración [Fer01a] y [Fer01b], y en aplicaciones relacionadas con la atención selectiva visual [Lop03a], [Lop04]. En todos estos trabajos se ha abordado la solución a estos problemas usando una serie de métodos de inspiración biológica basados en dos mecanismos fundamentales: (1) la computación acumulativa, [Fer92], [Fer95a], [Fer97] y [Mir03b]; y (2) una versión generalizada del cálculo realizado por las redes de inhibición lateral algorítmica (ALI) [Mir01], [Fer01a], [Fer01b], [Del02], [Fer03b] y [Fer03c]. ...
... Todas estas investigaciones nos han permitido afrontar aplicaciones como el reconocimiento de siluetas de objetos móviles en entornos ruidosos [Fer03a], la clasificación de móviles según sus características del movimiento, como su velocidad o su aceleración [Fer01a] y [Fer01b], y en aplicaciones relacionadas con la atención selectiva visual [Lop03a], [Lop04]. En todos estos trabajos se ha abordado la solución a estos problemas usando una serie de métodos de inspiración biológica basados en dos mecanismos fundamentales: (1) la computación acumulativa, [Fer92], [Fer95a], [Fer97] y [Mir03b]; y (2) una versión generalizada del cálculo realizado por las redes de inhibición lateral algorítmica (ALI) [Mir01], [Fer01a], [Fer01b], [Del02], [Fer03b] y [Fer03c]. ...
... Accumulative computation has now been largely applied to moving objects detection, classification and tracking in indefinite sequences of images (e.g. [16], [17], [18], [19]). The more general modality of accumulative computation is the charge/discharge mode, which may be described by means of the following generic formula: ...
... In concrete, good results use to be obtained with 8 levels. These 8 level images are called images segmented into 8 grey level bands and are stored in the Grey Level Bands Map [16], [18], as stated in Equation 2: ...
A new method for active visual attention is briefly introduced in this paper. The method extracts motion and shape features
from indefinite image sequences, and integrates these features to segment the input scene. The aim of this paper is to highlight
the importance of the accumulative computation method for motion features extraction in the active selective visual attention
model proposed. We calculate motion presence and velocity at each pixel of the input image by means of accumulative computation.
The paper shows an example of how to use motion features to enhance scene segmentation in this active visual attention method.
... [1], [2]) and their activities [3] is performed both in visible (e.g. [4], [5], [6]) and infrared spectrum (e.g. [7], [8]). ...
This paper introduces a solution for detecting humans in smart spaces through computer vision. The approach is valid both for images in visible and infrared spectra. Histogram of oriented gradients (HOG) is used for feature extraction in the human detection process, whilst linear support vector machines (SVM) are used for human classification. A set of tests is conducted to find the classifiers which optimize recall in the detection of persons in visible video sequences. Then, the same classifiers are used to detect people in infrared video sequences obtaining excellent results.
... La interacción lateral en computación acumulativa [5], [6], [7] (de aquí en adelante, ILCA), es un método conducido por datos, capaz de obtener con bastante claridad los objetos deformables presentes en una secuencia de imágenes indefinida, independientemente del tipo de movimiento. La ILCA se implementa como una red neuronal multicapa inspirada en dos modelos: la computación acumulativa local [8] y la interacción lateral recurrente [11]. ...
Resumen El problema de la segmentación de objetos en movimiento en secuencias de vídeo ha sido abordado desde varias aproximaciones. Aumenta un grado la dificultad cuando los objetos monitorizados poseen una apariencia deformable. El método usado en este documento utiliza una red neuronal, explotando la mecánica de la computación acumulativa en conjunción con la interacción lateral recurrente. A pesar de los resultados contrastados en anteriores trabajos, realizamos en este artículo un estudio para mejorar la segmentación sin recurrir a conocimiento de alto nivel. Los módulos propuestos incluyen un filtrado de los objetos según características de tamaño y compacidad y un algoritmo genético capaz de aprender los parámetros que se comportan de un modo mejor.
... Por tanto, no podemos despreciar la influencia que puede tener la consulta de un nodo en los nodos relacionados con éste. Para ello, se usan mecanismos de interacción lateral [10], en los que los términos relacionados se ven influidos por la aparición de otros. Mediante la conjunción de ambos mecanismos, interacción lateral en computación acumulativa [8][9], la aparición de un término debería aumentar la prioridad de otros términos relacionados, en mayor medida cuanto mayor sea el grado de proximidad, siendo el máximo incremento para el propio término. ...
RESUMEN A lo largo del documento se plantea cómo llevar los principios de adaptatividad de interfaces de usuario al terreno de las aplicaciones web. Para ello necesitaremos ser capaces de establecer ciertas características de estos usuarios, características que serán empleadas en el proceso de adaptación. Basándonos en estas ideas, hemos implementado un portal web capaz de recopilar información de la sesión de navegación de los usuarios. Para ello, consultará los metadatos asociados a los diferentes documentos. A partir de esta información, la interfaz recomendará aquellas páginas del portal que se ajusten mejor a los intereses del usuario.
... In this paper we introduce a model of dynamic visual attention that combines bottom-up and top-down processes. Bottom-up is related to the first step of the architectures proposed, where the input image is segmented using dynamic criteria by means of neurally inspired accumulative computation234 and lateral interaction5678. The observer may indicate how to tune system parameters to define the attention focus using top-down processes. ...
A model for dynamic visual attention is briefly introduced in this paper. A PSM (problem-solving method) for a generic “Dynamic
Attention Map Generation” task to obtain a Dynamic Attention Map from a dynamic scene is proposed. Our approach enables tracking objects that keep attention in accordance with a set of characteristics
defined by the observer. This paper mainly focuses on those subtasks of the model inspired in neuronal mechanisms, such as
accumulative computation and lateral interaction. The subtasks which incorporate these biologically plausible capacities are
called “Working Memory Generation” and “Thresholded Permanency Calculation”.
... Firstly, AC was applied to the problem of the classification of moving objects in long image sequences [5], showing its capacity to be implemented in real-time [7]. Later on the combination of AC and ALI was used in the resolution of the problem of segmenting moving silhouettes in video sequences [8], [9]. Its neuronal nature was described in detail [10], as well as the model for motion detection [11] and the influence of each parameter of the combination between AC and ALI [12]. ...
Certainly, one of the prominent ideas of Professor Mira was that it is absolutely mandatory to specify the mechanisms and/or
processes underlying each task and inference mentioned in an architecture in order to make operational that architecture.
The conjecture of the last fifteen years of joint research of Professor Mira and our team at University of Castilla-La Mancha
has been that any bottom-up organization may be made operational using two biologically inspired methods called “algorithmic
lateral inhibition”, a generalization of lateral inhibition anatomical circuits, and “accumulative computation”, a working
memory related to the temporal evolution of the membrane potential. This paper is dedicated to the computational formulations
of both methods, which have led to quite efficient solutions of problems related to motion-based computer vision.
... Its application to moving objects detection and labeling has also been described (Ló pez, FernándezCaballero, Mira, Delgado, & Fernández, 2006), as well as its approach to visual surveillance (Ló pez, FernándezCaballero, Fernández, Mira, & Delgado, 2006a). We have approached the solution to the DSVA problem with a series of biologically inspired methods based on two fundamental mechanisms: (1) accumulative computation (Fernández et al., 1995) and (2) a generalized version of the calculation done by lateral inhibition networks called algorithmic lateral inhibition (Fernández-Caballero, Mira, Fernández, & Ló pez, 2001; Fernández-Caballero, Mira, , & Fernández, 2003). Our approach is similar to the attention model for dynamic vision presented by Backer and Mertsching (2003), where the characteristics extracted are symmetry, eccentricity, color, contrast and depth. ...
Depth inclusion as an important parameter for dynamic selective visual attention is presented in this article. The model introduced in this paper is based on two previously developed models, dynamic selective visual attention and visual stereoscopy, giving rise to the so-called dynamic stereoscopic selective visual attention method. The three models are based on the accumulative computation problem-solving method. This paper shows how software reusability enables enhancing results in vision research (video segmentation) by integrating earlier works. In this article, the first results obtained for synthetic sequences are included to show the effectiveness of the integration of motion and shape features with depth parameter in video segmentation.
... The aim of subtask Motion feature extraction is to calculate the dynamic (motion) features of the image pixels, that is to say, in our case, the presence of motion. Due to our experience363738394041 we know some methods to get that information. Indeed, to diminish the effects of noise due to the changes in illumination in motion detection, variation in gray level bands at each image pixel is performed. ...
This paper describes a method for visual surveillance based on biologically motivated dynamic visual attention in video image sequences. Our system is based on the extraction and integration of local (pixels and spots) as well as global (objects) features. Our approach defines a method for the generation of an active attention focus on a dynamic scene for surveillance purposes. The system segments in accordance with a set of predefined features, including gray level, motion and shape features, giving raise to two classes of objects: vehicle and pedestrian. The solution proposed to the selective visual attention problem consists of decomposing the input images of an indefinite sequence of images into its moving objects, defining which of these elements are of the user's interest at a given moment, and keeping attention on those elements through time. Features extraction and integration are solved by incorporating mechanisms of charge and discharge—based on the permanency effect—, as well as mechanisms of lateral interaction. All these mechanisms have proved to be good enough to segment the scene into moving objects and background.
... Lateral interaction in accumulative computation has recently been introduced15161718, as well as its application to segmentation from motion [19]. For it, a generic model based on a neural architecture was presented. ...
To be able to understand the motion of non-rigid objects, techniques in image processing and computer vision are essential for motion analysis. Lateral interaction in accumulative computation for extracting non-rigid shapes from an image sequence has recently been presented, as well as its application to segmentation from motion. In this paper, we introduce a modified version of the first multi-layer architecture. This version uses the basic parameters of the LIAC model to spatio-temporally build up to the desired extent the shapes of all moving objects present in a sequence of images. The influences of LIAC model parameters are explained in this paper, and we finally show some examples of the usefulness of the model proposed.
... In this paper, our study is focused on a segmentation method based on lateral interaction in accumulative calculation (LIAC) [8] [9] [10] that can be generically included in the image difference methods. LIAC is a multi-layer artificial neuronal network (ANN) inspired in two models: local accumulative computation [12] and recurrent lateral interaction [27]. ...
Segmentation of moving objects is an essential component of any vision system. However, its accomplishment is hard due to some challenges such as the occlusion treatment or the detection of objects with deformable appearance. In this paper an artificial neuronal network approach for moving object segmentation, called lateral interaction in accumulative computation (LIAC), which uses accumulative computation and recurrent lateral interaction is revisited. Although the results reported for this approach so far may be considered relevant, the problems faced each time (environment, objects of interest, etc.) make that the system outcome varies. Hence, our aim is to improve segmentation provided by LIAC in a double sense: by removing the detected objects not matching some size or compactness constraints, and by learning suitable parameters that improve the segmentation behavior through a genetic algorithm.
... Some outstanding approaches to motion detection are biologically (neurally) inspired (e.g.,45678). Also in the last few years, the neurally inspired accumulative computation (AC) method9101112 and its application to motion detection have been introduced131415. Currently our research team is involved in implementing the method into real-time in order to provide efficient performance in visual surveillance applications161718. ...
The neurally inspired accumulative computation (AC) method and its application to motion detection have been introduced in the past years. This paper revisits the fact that many researchers have explored the relationship between neural networks and finite state machines. Indeed, finite state machines constitute the best characterized computational model, whereas artificial neural networks have become a very successful tool for modeling and problem solving. The article shows how to reach real-time performance after using a model described as a finite state machine. This paper introduces two steps towards that direction: (a) A simplification of the general AC method is performed by formally transforming it into a finite state machine. (b) A hardware implementation in FPGA of such a designed AC module, as well as an 8-AC motion detector, providing promising performance results. We also offer two case studies of the use of AC motion detectors in surveillance applications, namely infrared-based people segmentation and color-based people tracking, respectively.
... In input modality we have: if ((p(t-∆t) == 1) && (p(t) == 0)) then Q(t) = Q max else begin Q(t) = Q(t-∆t) -δQ; if Q(t) < Q min then Q(t) = Q min ; end; Finally, the more general charge/discharge modality is shown (figure 2f). This one has already been successfully used in some previous papers of the authors of this work [6][9]. These papers are about moving objects detection, classification and tracking in indefinite image sequences. ...
This paper explores the computational capacity of a novel local computational model that expands the conventional analogical and logical dynamic neural models, based on the charge and discharge of a capacity or in the use of a D flip-flop. The local memory capacity is augmented to behave as an S states automaton and some control elements are added to the memory. The analogical or digital calculus equivalent part of the balance between excitation and inhibition is also generalised to include the measure of specific spatio- temporal features over temporal expansions of the input space (dendritic field). This model is denominated as accumulative computation and is inspired in biological short-term memory mechanisms. The work describes the model's general specifications, including its architecture, the different working modes and the learning parameters. Then, some possible software and hardware implementations (using FPGAs) are proposed, and, finally, its potential usefulness in real time motion detection tasks is illustrated.
... The system uses a camera mounted on a bridge which captures real traffic images. These images, which are processed in a 256-color-gray scale format, are segmented according to [3] [4] [5] model in such a way that each object's movement captured in the real image may be observed. The segmented image shows the image's background in black, whereas the motion detected in the vehicles is represented by pixels in different gray scales (seeFig. ...
This article presents a knowledge-based application to study and analyze traffic behavior on major roads, using as the main
surveillance artefact a video camera mounted on a relatively high place with a significant image analysis field. The system
described presents something new which is the combination of both traditional traffic monitoring systems, that is, monitoring
to get information on different traffic parameters and monitoring to detect accidents automatically. Therefore, we present
a system in charge of compiling information on different traffic parameters. It also has a surveillance module, which can
detect a wide range of the most significant incidents on a freeway or highway.
... Recently the algorithmic lateral inhibition (ALI) method and its application to the motion detection task have been introduced [1][5]. And, currently our research team is involved in implementing the method into real-time in order to provide efficient response time in visual surveillance applications [6][7]. ...
Many researchers have explored the relationship between recurrent neural networks and finite state machines. Finite state
machines constitute the best characterized computational model, whereas artificial neural networks have become a very successful
tool for modeling and problem solving. Recently, the neurally-inspired algorithmic lateral inhibition (ALI) method and its
application to the motion detection task have been introduced. The article shows how to implement the tasks directly related
to ALI in motion detection by means of a formal model described as finite state machines. Automata modeling is the first step
towards real-time implementation by FPGAs and programming of ”intelligent” camera processors.
... Later on the combination of AC and ALI (introducing a second time scale into the equation) was used in the resolution of the problem of segmenting moving silhouettes in video sequences [10,11]. Its neuronal nature was described in detail [12], as well as the model for motion detection [13] and the influence of each parameter of the combination between AC and ALI [14]. ...
Certainly, one of the prominent ideas of Professor José Mira was that it is absolutely mandatory to specify the mechanisms and/or processes underlying each task and inference mentioned in an architecture in order to make operational that architecture. The conjecture of the last fifteen years of joint research has been that any bottom-up organization may be made operational using two biologically inspired methods called “algorithmic lateral inhibition”, a generalization of lateral inhibition anatomical circuits, and “accumulative computation”, a working memory related to the temporal evolution of the membrane potential. This paper is dedicated to the computational formulation of both methods. Finally, all of the works of our group related to this methodological approximation are mentioned and summarized, showing that all of them support the validity of this approximation.
... These methods are the permanency effect and the lateral inhibition (Fernández-Caballero, Mira, Fernández, & López, 2001;. Based on the satisfactory results of these algorithms (Fernández-Caballero, Mira, Férnandez & Delgado, 2003), in this paper we propose to use mechanisms of charge and discharge together with mechanisms of lateral inhibition to solve the current task of DSVA. ...
Abstract In a recent article, knowledge modelling at the knowledge level for the task of moving objects detection in image sequences has been introduced. In this paper, the algorithmic lateral inhibition (ALI) method is now applied in the generic dynamic and selective visual attention (DSVA) task with the objective of moving objects detection, labelling and further tracking. The four basic subtasks, namely feature extraction, feature integration, attention building and attention reinforcement in our proposal of DSVA are described in detail by inferential CommonKADS schemes. It is shown that the ALI method, in its various forms, that is to say, recurrent and non-recurrent, temporal, spatial and spatial-temporal, may,perfectly be used as a problem-solving-method,in most of the subtasks involved in the DSVA task. q 2005 Elsevier Ltd. All rights reserved. Keywords: Dynamic and selective visual attention; Algorithmic lateral inhibition; Feature extraction; Feature integration 1. Modelling the dynamic,and selective visual attention task
... In the last few years, lateral inhibition in accumulative computation (LIAC) method (Ferná ndez-Caballero et al., 2003a, 2003b, 2007, 2008aMira et al., 2004) and its application to the motion detection task have been introduced (Ferná ndez- Caballero et al., 2001Caballero et al., , 2003cMartínez-Cantos et al., 2008). LIAC is a neurally inspired method based on biologically based accumulative computation (AC) and lateral inhibition (LI). ...
Many researchers have explored the relationship between recurrent neural networks and finite state machines. Finite state machines constitute the best characterized computational model, whereas artificial neural networks have become a very successful tool for modeling and problem solving. In the few last years, the neurally inspired lateral inhibition in accumulative computation (LIAC) method and its application to the motion detection task have been introduced. The article shows how to implement the tasks directly related to LIAC in motion detection by means of a formal model described as finite state machines. This paper introduces two steps towards that direction: (a) A simplification of the general LIAC method is performed by formally transforming it into a finite state machine. (b) A hardware implementation of such a designed LIAC module, as well as an 8×8 LIAC module, has been tested on several video sequences, providing promising performance results.
Traffic jams are a huge problem for all road users and are caused by increasing traffic intensity and poor quality of traffic management systems. Systems that control traffic flows and decide to change control parameters must receive reliable and up-to-date data on traffic intensity. In order to accurately determine the intensity of traffic, a system of automated video data processing from video surveillance cameras of the traffic lane was developed. The traffic intensity is determined by the developed method of obtaining the traffic congestion index (TLCR) according to the data obtained by processing the video frame using the U-Net neural network. The research results show that the proposed method is able to calculate vehicles with an accuracy of 99.35%.
In recent years, convolutional neural networks (CNNs) have shown their advantages on MR image super-resolution (SR) tasks. Many current SR models, however, have heavy demands on computation and memory, which are not friendly to magnetic resonance imaging (MRI) where computing resource is usually constrained. On the other hand, a basic consideration in most MRI experiments is how to reduce scanning time to improve patient comfort and reduce motion artifacts. In this work, we ease the problem by presenting an effective and lightweight model that supports fast training and accurate SR inference. The proposed network is inspired by the lateral inhibition mechanism, which assumes that there exist inhibitory effects between adjacent neurons. The backbone of our network consists of several lateral inhibition blocks, where the inhibitory effect is explicitly implemented by a battery of cascaded local inhibition units. When model scale is small, explicitly inhibiting feature activations is expected to further explore model representational capacity. For more effective feature extraction, several parallel dilated convolutions are also used to extract shallow features directly from the input image. Extensive experiments on typical MR images demonstrate that our lateral inhibition network (LIN) achieves better SR performance than other lightweight models with similar model scale.
Concentrate on the computational model of visual attention, in this paper, we present the electric circuit model of the human retina cells on the base of lateral inhibit network and then deduce the mathematic formulas to define the static and dynamic processes of the network. The learning algorithm is also discussed to get right model parameters. The model reveals the case of the human biological character on image and video processing. Specifically on video processing, the convolution integral function which shows the neural computing process of the human eyes is developed based on the dynamical equation and the circuit principle. Some examples are also introduced to demonstrate the neural computing process of the model on the static image and video processing.
Resumen En Inteligencia Artiflcial esta ampliamente aceptado el im- portante papel que desempe~na el razonamiento espacial. Por un lado, el razonamiento cualitativo espacial puede tratarse mediante teor¶‡as pura- mente existenciales en forma de sistemas de satisfaccion de restricciones sobre un conjunto de relaciones mutuamente excluyentes y disjuntas. Por otro lado, las logicas modales y de primer orden proporcionan un poder expresivo mayor a costa de un peor comportamiento computacional. En este art¶‡culo, se compara una logica modal, llamada Spatial Propositio- nal Neighborhood Logic (SpPNL), con el ¶ Algebra de Rectangulos, que es la extension bidimensional del ¶ Algebra de Intervalos de Allen para el razonamiento temporal. Se demuestra como es posible comprobar la consistencia de una red de restricciones del ¶ Algebra de Rectangulos me- diante una formula en SpPNL. Tambien se muestra como dicha logica permite expresar ciertas restricciones espaciales intuitivas que no pueden representarse en el ¶ Algebra de Rectangulos.
Resumen En los últimos años, hemos sido testigos del importante papel que la lógica difusa o borrosa (fuzzy logic) ha jugado en el desarrollo de sofisticadas aplicaciones software en campos tan diversos como los sistemas expertos, inteligencia artificial, control industrial, medicina, etc. Con el objetivo de facilitar el desarrollo de tales aplicaciones, mucho más recientemente ha surgido el interés por diseñar lenguajes declara- tivos (funcionales, lógicos y/o difusos) que incorporen entre sus recursos expresivos el tratamiento de información imprecisa de forma natural, al tiempo que puedan ejecutarse de forma eficiente mediante, por ejem- plo, el uso de mecanismos de evaluación perezosos. Para que ello sea factible, es necesario definir previamente una noción de igualdad apropi- ada que integre de forma natural todas estas características en un marco uniforme. En este trabajo, proponemos una potente noción de igualdad en este sentido, que es capaz de dar cuenta de propiedades perezosas y borrosas en un contexto declarativo integrado, y que además resulta facilmente implementable a un alto nivel de abstracción. Palabras clave: Igualdad, Programación Declarativa, Lógica Difusa
An important problem in artificial intelligence (AI) is to find calculation procedures to save the semantic gap between the analytic formulations of the neuronal models and the concepts of the natural language used to describe the cognitive processes. In this work we explore a way of saving this gap for the case of the attentional processes, consisting in (1) proposing in first place a conceptual model of the attention double bottom-up/top-down organization, (2) proposing afterwards a neurophysiological model of the cortical and sub-cortical involved structures, (3) establishing the correspondences between the entities of (1) and (2), (4) operationalizing the model by using biologically inspired calculation mechanisms (algorithmic lateral inhibition and accumulative computation) formulated at symbolic level, and, (5) assessing the validity of the proposal by accommodating the works of the research team on diverse aspects of attention associated to visual surveillance tasks. The results obtained support in a reasonable way the validity of the proposal and enable its application in surveillance tasks different from the ones considered in this work. In particular, this is the case when linking the geometric descriptions of a scene with the corresponding activity level.
In this article knowledge modelling at the knowledge level for the task of moving objects detection in image sequences is introduced. Three items have been the focus of the approach: (1) the convenience of knowledge modelling of tasks and methods in terms of a library of reusable components and in advance to the phase of operationalization of the primitive inferences; (2) the potential utility of looking for inspiration in biology; (3) the convenience of using these biologically inspired problem-solving methods (PSMs) to solve motion detection tasks.After studying a summary of the methods used to solve the motion detection task, the moving targets in indefinite sequences of images detection task is approached by means of the algorithmic lateral inhibition (ALI) PSM. The task is decomposed in four subtasks: (a) thresholded segmentation; (b) motion detection; (c) silhouettes parts obtaining; and (d) moving objects silhouettes fusion. For each one of these subtasks, first, the inferential scheme is obtained and then each one of the inferences is operationalized. Finally, some experimental results are presented along with comments on the potential value of our approach.
This article presents a visual application which allows a study and analysis of traffic behavior on major roads (more specifically freeways and highways), using as the main surveillance artefact a video camera mounted on a relatively high place (such as a bridge) with a significant image analysis field. The system described presents something new which is the combination of both traditional traffic monitoring systems, that is, monitoring to get information on different traffic parameters and monitoring to detect accidents automatically. Therefore, we present a system in charge of compiling information on different traffic parameters. It also has a surveillance module for that traffic, which can detect a wide range of the most significant incidents on a freeway or highway.
Visual attention is the ability to rapidly detect the interest- ing parts of a given scene on which higher level computer vision tasks can focus. This paper reports a computational model of dynamic visual attention which combines static and dynamic features to detect salient locations in natural image sequences. Therefore, the model computes a map of interest - saliency map - related to static features and a saliency map derived from dynamic scene features and then combines them into a final saliency map, which topographically encodes stimulus saliency. The information provided by the model of attention is then used by a tracking method to attentively track the interesting features in the scene. The experimental results, reported in this work refer to real color image sequences. They clearly validate the reported model of dynamic visual attention and show its usefulness in guiding the tracking task.
To be able to understand the motion of non-rigid objects, techniques in image processing and computer vision are essential
for motion analysis. Lateral interaction in accumulative computation for extracting non-rigid blobs and shapes from an image
sequence has recently been presented, as well as its application to segmentation from motion. In this paper we show an architecture
consisting of five layers based on spatial and temporal coherence in visual motion analysis with application to visual surveillance.
The LIAC method used in general task ”spatio-temporal coherent shape building” consists in (a) spatial coherence for brightness-based
image segmentation, (b) temporal coherence for motion-based pixel charge computation, (c) spatial coherence for charge-based
pixel charge computation, (d) spatial coherence for charge-based blob fusion, and, (e) spatial coherence for charge-based
shape fusion. In our case, temporal coherence (in accumulative computation) is understood as a measure of frame to frame motion
persistency on a pixel, whilst spatial coherence (in lateral interaction) is a measure of pixel to neighbouring pixels accumulative
charge comparison.
Digital image analysis appears to be more and more relevant to the study of physical phenomena involving fluid motion, and of their evolution over time. In that context, 2D deformable motion analysis is one of the important issues to be investigated. The interpretation of such deformable 2D flow fields can generally be stated as the characterization of linear models provided that first order approximations are considered in an adequate neighborhood of so-called singular points, where the velocity becomes null. This paper describes an efficient method, based on a statistical approach, which explicitly addresses these problems, and allows us to locate, characterize and track such singular points in an image sequence. It does not require the prior computation of the velocity field. The method has been validated by experiments carried out with synthetic and real examples corresponding to meteorological image sequences. In fact, the described approach can be of interest in different applications dealing with the characterization of vector fields
In this paper, we propose a new statistical framework for modeling
and extracting 2D moving deformable objects from image sequences. The
object representation relies on a hierarchical description of the
deformations applied to a template. Global deformations are modeled
using a Karhunen Loeve expansion of the distortions observed on a
representative population. Local deformations are modeled by a
(first-order) MarKov process. The optimal bayesian estimate of the
global and local deformations is obtained by maximizing a non-linear
joint probability distribution using stochastic and deterministic
optimization techniques. The use of global optimization techniques
yields robust and reliable segmentations in adverse situations such as
low signal-to-noise ratio, non-gaussian noise or occlusions. Moreover,
no human interaction is required to initialize the model. The approach
is demonstrated on synthetic as well as on real-world image sequences
showing moving hands with partial occlusions
Recent developments are reviewed in the computation of motion and
structure of objects in a scene from a sequence of images. Two distinct
paradigms are highlighted: (i) the feature-based approach and (ii) the
optical-flow-based approach. The comparative merits/demerits of these
approaches are discussed. The current status of research in these areas
is reviewed and future research directions are indicated
W 4 is a real time visual surveillance system for detecting and tracking people and monitoring their activities in an outdoor environment. It operates on monocular grayscale video imagery, or on video imagery from an infrared camera. Unlike many of systems for tracking people, W 4 makes no use of color cues. Instead, W 4 employs a combination of shape analysis and tracking to locate people and their parts (head, hands, feet, torso) and to create models of people's appearance so that they can be tracked through interactions such as occlusions. W 4 is capable of simultaneously tracking multiple people even with occlusion. It runs at 20 Hz for 320x240 resolution images on a dual-pentium 200 PC. 1 Overview of the W 4 System W 4 is a real time system for tracking people and their body parts in monochromatic imagery. It constructs dynamic models of people's movements to answer questions about what they are doing, and where and when they act. It constructs appearance models of t...
A snake is an energy-minimizing spline guided by external constraint forces and influenced by image forces that pull it toward features such as lines and edges. Snakes are active contour models: they lock onto nearby edges, localizing them accurately. Scale-space continuation can be used to enlarge the capture region surrounding a feature. Snakes provide a unified account of a number of visual problems, including detection of edges, lines, and subjective contours, motion tracking, and stereo matching. The authors have used snakes successfully for interactive interpretation, in which user-imposed constraint forces guide the snake near features of interest.
This paper discusses the problem of determining the three-dimensional
model and movement of an object from a sequence of two-dimensional
images. A solution to this problem depends on solving a system of
nonlinear equations using a modified least-squared error method. Two
views of six points or three views of four points are needed to provide
an overdetermined set of equations when the images are noisy. It is
shown, however, that this numerical method is not very accurate unless
the images of considerably more points are used.
Usually we have assumed that the neuron was the atom in the architecture of the Neurons System. However, there is such a wealth of drendo-dendritic connections and synaptic mechanisms that it seems essential to distinguish different styles of analog microcomputation.
In this paper we look inside the synaptic structure after a local process of accumulation of persistent activity and their discharge towards the spike trigger zone. To illustrate the usefulness of this information processing behaviour in image motion analysis, and architecture for extraction and selection of length velocity ratio invariants (LVR) is proposed, simulated and partially evaluated.
The anatomy and physiology of biological neurons is revisited looking at a minimum set of computational requirements to be included in new and more complex models of self-contained local computation ANN. Some of these functionalities are then integrated and the corresponding model is evaluated. Properties included are: (1) locality and autonomy in all the computations including the learning algorithms. (2) a layered architecture with high-order recurrent neurons, (3) self and external programming via input spaces and (4) fault tolerance after physical lesion, or even elimination of one or more neurons.
Given the case that the improvements included in this proposal are of some value, it is nonetheless clear to us that the more genuine properties of biological computation are still lost. If we were to look for implementation, we would say that a minimal model of a neuron should include at least the computational capacity of a microprocessor with self-programming and fault tolerance facilities as addenda.
A snake is an energy-minimizing spline guided by external constraint forces and influenced by image forces that pull it toward features such as lines and edges. Snakes are active contour models: they lock onto nearby edges, localizing them accurately. Scale-space continuation can be used to enlarge the capture region surrounding a feature. Snakes provide a unified account of a number of visual problems, including detection of edges, lines, and subjective contours; motion tracking; and stereo matching. We have used snakes successfully for interactive interpretation, in which user-imposed constraint forces guide the snake near features of interest.
I deseribe a system that first segments an image into parts, and then fits deformable models to range data associated with the image. Models are represented by using modal dynamics applied to volumetric primitives, which significantly improves the computational complexity of both model recovery and subsequent processing. The segmentation procedure uses two simple mechanisms: a filtering operation to produce a large set of potential object parts, followed by a quadratic integer optimization that searches among these part hypotheses to produce a maximum-likelihood estimate of the image's part structure. Once a segmentation has been produced, a volumetric description is obtained by a fitting procedure that minimizes squared error between the range measurements an the model's visible surface. For simple part shapes, it is possible to compute the deformable model's parameters using only the shape of its symmetry axes.
Thesis (Ph. D.)--Massachusetts Institute of Technology, 1977. Bibliography: leaves 248-254. Microfilm.
The elastic properties of real materials provide constraint on the
types of non-rigid motion that can occur, and thus allow overconstrained
estimates of 3-D non-rigid motion from optical flow data. It is shown
that by modeling and simulating the physics of non-rigid motion it is
possible to obtain good estimates of both object shape and velocity.
Examples using grey-scale and X-ray imagery are presented, including an
example of tracking a complex articulated figure
A class of physically based models suitable for animating flexible objects in simulated physical environments was proposed earlier by the authors (1987). The original formulation works as well in practice for models whose shapes are moderately to highly deformable, but it tends to become numerically ill conditioned as the rigidity of the models is increased. An alternative formulation of deformable models is presented in which deformations are decomposed into a reference component, which may represent an arbitrary shape, and a displacement component, allowing deformation away from this reference shape. The application of the deformable models to a physically based computer animation project is illustrated.< >
Aspectos B asicos de la Inteligencia Arti®cial Cooperative processes at the symbolic level in cerebral dynamics: reliability and fault tolerance
- J Mira
- A E Delgado
- J G Boticario
- F J Diez
- Torres
- Sl
- Madrid
- J Mira
- A E Delgado
- A Manjarres
- S Ros
- J R Alvarez
Mira, J., Delgado, A.E., Boticario, J.G., Diez, F.J., 1995. Aspectos B asicos de la Inteligencia Arti®cial. Editorial Sanz y Torres, SL, Madrid. Mira, J., Delgado, A.E., Manjarres, A., Ros, S., Alvarez, J.R., 1996. Cooperative processes at the symbolic level in cerebral dynamics: reliability and fault tolerance. In: D õaz, R.M., Mira, J. (Eds.), Brain Processes Theories and Models. MIT Press, Cambridge, MA, pp. 244±255. Moreno, D.R, Rubio, F., Mira, J., 1969. Aplicaci on de las transformaciones integrales al proceso de datos en la retina. Rev. de Autom atica 5, 7±17.
Automatic extraction of deformable part models Determining the motion of objects from a sequence of images
- A P Pentland
Pentland, A.P., 1990. Automatic extraction of deformable part models. Internat. J. Comput. Vision 4, 107±126. Roach, J.W., Aggarwal, J.K., 1980. Determining the motion of objects from a sequence of images. IEEE Trans. Pattern Anal. Machine Intell.
Computaci on neural en el camino visual
- J Mira
Mira, J., 1993. Computaci on neural en el camino visual. Notas de Visi on y Apuntes sobre la Ingenier õa del Software. Colecci on
Estudios, 24, UCLM, pp. 175±197.
Cooperative processes at the symbolic level in cerebral dynamics: reliability and fault tolerance
- J Mira
- A E Delgado
- A Manjarres
- S Ros
- J R Alvarez
Mira, J., Delgado, A.E., Manjarres, A., Ros, S., Alvarez, J.R., 1996. Cooperative processes at the symbolic level in cerebral dynamics:
reliability and fault tolerance. In: D õaz, R.M., Mira, J. (Eds.), Brain Processes Theories and Models. MIT Press, Cambridge, MA,
pp. 244±255.
Computación neural en el camino visual. Notas de Visión y Apuntes sobre la Ingenierı́a del Software
- J Mira
Aspectos Básicos de la Inteligencia Artificial
- J Mira
- A E Delgado
- J G Boticario
- F J Diez
Cooperative processes at the symbolic level in cerebral dynamics: reliability and fault tolerance
- Mira