Helder Araujo | University of Coimbra (original) (raw)
Papers by Helder Araujo
Pattern Recognition Letters
IEEE Computational Intelligence Magazine
International journal of intelligent robotics and applications, 2017
Since the development of capsule endoscopy technology, medical device companies and research grou... more Since the development of capsule endoscopy technology, medical device companies and research groups have made significant progress to turn passive capsule endoscopes into robotic active capsule endoscopes. However, the use of robotic capsules in endoscopy still has some challenges. One such challenge is the precise localization of the actively controlled robot in real-time. In this paper, we propose a non-rigid map fusion based direct simultaneous localization and mapping method for endoscopic capsule robots. The proposed method achieves high accuracy for extensive evaluations of pose estimation and map reconstruction performed on a non-rigid, realistic surgical EsophagoGastroDuodenoscopy Simulator and outperforms state-of-the art methods.
International journal of intelligent robotics and applications, 2017
A reliable, real time localization functionality is crutial for actively controlled capsule endos... more A reliable, real time localization functionality is crutial for actively controlled capsule endoscopy robots, which are an emerging, minimally invasive diagnostic and therapeutic technology for the gastrointestinal (GI) tract. In this study, we extend the success of deep learning approaches from various research fields to the problem of sensor fusion for endoscopic capsule robots. We propose a multi-sensor fusion based localization approach which combines endoscopic camera information and magnetic sensor based localization information. The results performed on real pig stomach dataset show that our method achieves sub-millimeter precision for both translational and rotational movements.
PSICOLOGIA, 2014
Aborda-se um dos temas da IA, o da codificação da informação em imagens, vulgo visão computaciona... more Aborda-se um dos temas da IA, o da codificação da informação em imagens, vulgo visão computacional, discutindo o seu âmbito e objectivo. E, situa-se o tema nas aplicações da robótica industrial.
Proceedings of the 10th International Conference on Distributed Smart Camera - ICDSC '16, 2016
In this paper we analyze and compare approaches for 3D reconstruction from low-resolution (250x25... more In this paper we analyze and compare approaches for 3D reconstruction from low-resolution (250x250), high radial distortion stereo images, which are acquired with small baseline (approximately 1mm). These images are acquired with the system NanEye Stereo manufactured by CMOSIS/AWAIBA. These stereo cameras have also small apertures, which means that high levels of illumination are required. The goal was to develop an approach yielding accurate reconstructions, with a low computational cost, i.e., avoiding non-linear numerical optimization algorithms. In particular we focused on the analysis and comparison of radial distortion models. To perform the analysis and comparison, we defined a baseline method based on available software and methods, such as the Bouguet toolbox [2] or the Computer Vision Toolbox from Matlab. The approaches tested were based on the use of the polynomial model of radial distortion, and on the application of the division model. The issue of the center of distortion was also addressed within the framework of the application of the division model. We concluded that the division model with a single radial distortion parameter has limitations.
As a consequence of the aperture problem, when a moving sine-grating is viewed through an apertur... more As a consequence of the aperture problem, when a moving sine-grating is viewed through an aperture, the perceived motion is ambiguous. However, when we superimpose two of this gratings, the resulting pattern usually moves in a coherent way. But crossed gratings do not always cohere into a single moving pattern. It has been shown that the orientation, speed, contrast and spatial frequency of each plaid are parameters that influence the perception of coherence. We have studied the stimulus conditions under which coherence does and does not occur, using human observers who assigned a probability of coherence to each sequence of images. We then tried to find a computational method whose result would correlate with the results from the human observers. The proposed method consists on computing the optical flow field on the image sequence, using then the direction of the optical flow in each point on the image to build an histogram. This histogram can provide information about the probability of coherence of the image sequences. If the histogram presents two well marked peaks this corresponds to the situation where we perceive transparency of motion. On the other hand when we perceive coherence the corresponding histogram presents only one peak. The width of this peak is related with the probability of coherence, i.e., the narrower the peak the highest the probability of coherence. We show that the proposed method provides results very similar to the ones obtained with the human observers in terms of probability of coherence.
To assure a cost-effective maintenance of (concrete) structures during their lifetime is importan... more To assure a cost-effective maintenance of (concrete) structures during their lifetime is important. Monitoring concrete deterioration is the best way to achieve this aim. Some of the most important anomalies in concrete are originated by mechanical, chemical and biological attacks, namely: cracking, carbonation and chloride ingress, and biological colonization, the latter often associated with moisture. For the pathological characterization of the structure, periodic detection, mapping and measuring of concrete anomalies is needed. Normally, scheduled visual inspections are conducted and non-destructive tests are used when any sign of alarm is observed. Nevertheless, these traditional methods are time-consuming and, often, subjective results are obtained. In this paper, the authors describe the development of an innovative method, Intelligent Concrete Health Monitoring, based in digital processing of multi-spectral images, containing both visible and near infra red spectrums, to monitor concrete pathology. Concrete and mortar specimens were tested in laboratory to calibrate and validate the proposed method, having results also been compared with results obtained with conventional contact and visual methods. It was concluded that the developed method presents several advantages in relation to traditional methods, namely: increase in speed, efficiency and reliability.
Omnidirectional imaging is highly advantageous for several applications such as surveillance, mod... more Omnidirectional imaging is highly advantageous for several applications such as surveillance, model acquisition for virtual reality and visual control of motion. Central catadioptric systems are imaging systems that use mirrors to enhance the field of view while keeping a unique center of projection. A general model for central catadioptric image formation has already been established. The present paper proposes a modified version of the model made up of three functions: a linear function mapping the world into an oriented projective plane, a nonlinear transformation between two oriented projective planes, and a collineation in the plane. The mapping model is used to study the projection of lines. The equations and geometric properties of general catadioptric line imaging are derived. We show that it is possible to obtain the position of the effective viewpoint and the absolute conic in the catadioptric image plane from the images of three lines. The type of catadioptric system and the pose of the camera with respect to the mirror can also be determined without further information. Moreover for hyperbolic/elliptical systems it is proved that if both the shape of the mirror and the camera pose are known then two line images are enough to obtain the intrinsic parameters of the catadioptric system. The application of the theory derived to calibration purposes is discussed. Experimental results on tracking lines in the catadioptric image plane are presented.
2008 19th International Conference on Pattern Recognition, 2008
![Research paper thumbnail of An introduction to the log-polar mapping [image sampling]](https://attachments.academia-assets.com/71703682/thumbnails/1.jpg)
](Proceedings II Workshop on Cybernetic Vision, 1997
One interesting feature of the human visual system is the topological transformation of the retin... more One interesting feature of the human visual system is the topological transformation of the retinal image into its cortical projection. The excitation of the cortex can be approximated by a log-polar mapping of the eye's retinal image. In this tutorial paper we describe the log-polar mapping and its main properties.
A special architecture for texture analysis and recognition that is inherently parallel and aimed... more A special architecture for texture analysis and recognition that is inherently parallel and aimed at hardware implementation is described. It has the advantage of making possible real-time recognition (at frame rates). The architecture is based on the concept of associative memory, which has learning and recognizing capabilities. This architecture achieved high rates of recognition. With image processing, over 90% correct
Lecture Notes in Computer Science, 2008
In this article is introduced a framework for the development of reusable software components for... more In this article is introduced a framework for the development of reusable software components for Human Robot Interaction (HRI): the Experimental Robotics Framework. Normally human-robot interfaces are discarded as they stop being useful and because of that lots of work put into those architectures is lost. Our software plans to change that, and provide a platform that will enable interfaces to be reused. While most interfaces are static and 2D our interface uses a careful selection of technology and libraries to make sure it stays updated and free for a long time. We explain the architecture and design rationale of the framework, and demonstrate it with some use cases. The developed framework is available on-line with a LGPL license.
Lecture Notes in Computer Science, 2014
Lecture Notes in Computer Science, 2005
Artificial Intelligence in Industrial Decision Making, Control and Automation, 1995
Planes are important geometric features and can be used in a wide range of applications. This wor... more Planes are important geometric features and can be used in a wide range of applications. This work aims to illustrate a homography-based method to detect planes using the affine model. Using two image frames from a monocular sequence, a set of match pairs of interest points is obtained using Harris corner detector combined with the Scale Invariant Feature Transform (SIFT) as local descriptor. A Delaunay triangulation is performed on the set of detected corners of the first image and we use a triangle to calculate an affine homography that represents a plane in the image. An algorithm was developed to cluster interest points belonging to the same plane based on the reprojection error of the affine homography. Tests are performed in different sequences of indoor and outdoor images and results are shown.
Autonomous mobile robots rely on multiple sensors to perform a varied number of tasks in a given ... more Autonomous mobile robots rely on multiple sensors to perform a varied number of tasks in a given environment. Different tasks may need different sensors to estimate different subsets of world state. Also, different sensors can cooperate in discovering common subsets of world state. This paper presents a new approach to multimodal sensor fusion using dynamic Bayesian networks and an occupancy grid. The environment in which the robot operates is represented with an occupancy grid. This occupancy grid is asynchronously updated using ...
2. The Basis Of The Method In this paper we present a texture analysis and classification method ... more 2. The Basis Of The Method In this paper we present a texture analysis and classification method based on the computation of highorder statistics of gray-levels or selected features. The computation of these high-order statistics can be performed in parallel in non-overlapping windows that cover the full image. Classification is performed by neural networks. In this extended summary results are presented for back-propagation networks only, but several other types of neural nets were also tested.
Pattern Recognition Letters
IEEE Computational Intelligence Magazine
International journal of intelligent robotics and applications, 2017
Since the development of capsule endoscopy technology, medical device companies and research grou... more Since the development of capsule endoscopy technology, medical device companies and research groups have made significant progress to turn passive capsule endoscopes into robotic active capsule endoscopes. However, the use of robotic capsules in endoscopy still has some challenges. One such challenge is the precise localization of the actively controlled robot in real-time. In this paper, we propose a non-rigid map fusion based direct simultaneous localization and mapping method for endoscopic capsule robots. The proposed method achieves high accuracy for extensive evaluations of pose estimation and map reconstruction performed on a non-rigid, realistic surgical EsophagoGastroDuodenoscopy Simulator and outperforms state-of-the art methods.
International journal of intelligent robotics and applications, 2017
A reliable, real time localization functionality is crutial for actively controlled capsule endos... more A reliable, real time localization functionality is crutial for actively controlled capsule endoscopy robots, which are an emerging, minimally invasive diagnostic and therapeutic technology for the gastrointestinal (GI) tract. In this study, we extend the success of deep learning approaches from various research fields to the problem of sensor fusion for endoscopic capsule robots. We propose a multi-sensor fusion based localization approach which combines endoscopic camera information and magnetic sensor based localization information. The results performed on real pig stomach dataset show that our method achieves sub-millimeter precision for both translational and rotational movements.
PSICOLOGIA, 2014
Aborda-se um dos temas da IA, o da codificação da informação em imagens, vulgo visão computaciona... more Aborda-se um dos temas da IA, o da codificação da informação em imagens, vulgo visão computacional, discutindo o seu âmbito e objectivo. E, situa-se o tema nas aplicações da robótica industrial.
Proceedings of the 10th International Conference on Distributed Smart Camera - ICDSC '16, 2016
In this paper we analyze and compare approaches for 3D reconstruction from low-resolution (250x25... more In this paper we analyze and compare approaches for 3D reconstruction from low-resolution (250x250), high radial distortion stereo images, which are acquired with small baseline (approximately 1mm). These images are acquired with the system NanEye Stereo manufactured by CMOSIS/AWAIBA. These stereo cameras have also small apertures, which means that high levels of illumination are required. The goal was to develop an approach yielding accurate reconstructions, with a low computational cost, i.e., avoiding non-linear numerical optimization algorithms. In particular we focused on the analysis and comparison of radial distortion models. To perform the analysis and comparison, we defined a baseline method based on available software and methods, such as the Bouguet toolbox [2] or the Computer Vision Toolbox from Matlab. The approaches tested were based on the use of the polynomial model of radial distortion, and on the application of the division model. The issue of the center of distortion was also addressed within the framework of the application of the division model. We concluded that the division model with a single radial distortion parameter has limitations.
As a consequence of the aperture problem, when a moving sine-grating is viewed through an apertur... more As a consequence of the aperture problem, when a moving sine-grating is viewed through an aperture, the perceived motion is ambiguous. However, when we superimpose two of this gratings, the resulting pattern usually moves in a coherent way. But crossed gratings do not always cohere into a single moving pattern. It has been shown that the orientation, speed, contrast and spatial frequency of each plaid are parameters that influence the perception of coherence. We have studied the stimulus conditions under which coherence does and does not occur, using human observers who assigned a probability of coherence to each sequence of images. We then tried to find a computational method whose result would correlate with the results from the human observers. The proposed method consists on computing the optical flow field on the image sequence, using then the direction of the optical flow in each point on the image to build an histogram. This histogram can provide information about the probability of coherence of the image sequences. If the histogram presents two well marked peaks this corresponds to the situation where we perceive transparency of motion. On the other hand when we perceive coherence the corresponding histogram presents only one peak. The width of this peak is related with the probability of coherence, i.e., the narrower the peak the highest the probability of coherence. We show that the proposed method provides results very similar to the ones obtained with the human observers in terms of probability of coherence.
To assure a cost-effective maintenance of (concrete) structures during their lifetime is importan... more To assure a cost-effective maintenance of (concrete) structures during their lifetime is important. Monitoring concrete deterioration is the best way to achieve this aim. Some of the most important anomalies in concrete are originated by mechanical, chemical and biological attacks, namely: cracking, carbonation and chloride ingress, and biological colonization, the latter often associated with moisture. For the pathological characterization of the structure, periodic detection, mapping and measuring of concrete anomalies is needed. Normally, scheduled visual inspections are conducted and non-destructive tests are used when any sign of alarm is observed. Nevertheless, these traditional methods are time-consuming and, often, subjective results are obtained. In this paper, the authors describe the development of an innovative method, Intelligent Concrete Health Monitoring, based in digital processing of multi-spectral images, containing both visible and near infra red spectrums, to monitor concrete pathology. Concrete and mortar specimens were tested in laboratory to calibrate and validate the proposed method, having results also been compared with results obtained with conventional contact and visual methods. It was concluded that the developed method presents several advantages in relation to traditional methods, namely: increase in speed, efficiency and reliability.
Omnidirectional imaging is highly advantageous for several applications such as surveillance, mod... more Omnidirectional imaging is highly advantageous for several applications such as surveillance, model acquisition for virtual reality and visual control of motion. Central catadioptric systems are imaging systems that use mirrors to enhance the field of view while keeping a unique center of projection. A general model for central catadioptric image formation has already been established. The present paper proposes a modified version of the model made up of three functions: a linear function mapping the world into an oriented projective plane, a nonlinear transformation between two oriented projective planes, and a collineation in the plane. The mapping model is used to study the projection of lines. The equations and geometric properties of general catadioptric line imaging are derived. We show that it is possible to obtain the position of the effective viewpoint and the absolute conic in the catadioptric image plane from the images of three lines. The type of catadioptric system and the pose of the camera with respect to the mirror can also be determined without further information. Moreover for hyperbolic/elliptical systems it is proved that if both the shape of the mirror and the camera pose are known then two line images are enough to obtain the intrinsic parameters of the catadioptric system. The application of the theory derived to calibration purposes is discussed. Experimental results on tracking lines in the catadioptric image plane are presented.
2008 19th International Conference on Pattern Recognition, 2008
Proceedings II Workshop on Cybernetic Vision, 1997
One interesting feature of the human visual system is the topological transformation of the retin... more One interesting feature of the human visual system is the topological transformation of the retinal image into its cortical projection. The excitation of the cortex can be approximated by a log-polar mapping of the eye's retinal image. In this tutorial paper we describe the log-polar mapping and its main properties.
A special architecture for texture analysis and recognition that is inherently parallel and aimed... more A special architecture for texture analysis and recognition that is inherently parallel and aimed at hardware implementation is described. It has the advantage of making possible real-time recognition (at frame rates). The architecture is based on the concept of associative memory, which has learning and recognizing capabilities. This architecture achieved high rates of recognition. With image processing, over 90% correct
Lecture Notes in Computer Science, 2008
In this article is introduced a framework for the development of reusable software components for... more In this article is introduced a framework for the development of reusable software components for Human Robot Interaction (HRI): the Experimental Robotics Framework. Normally human-robot interfaces are discarded as they stop being useful and because of that lots of work put into those architectures is lost. Our software plans to change that, and provide a platform that will enable interfaces to be reused. While most interfaces are static and 2D our interface uses a careful selection of technology and libraries to make sure it stays updated and free for a long time. We explain the architecture and design rationale of the framework, and demonstrate it with some use cases. The developed framework is available on-line with a LGPL license.
Lecture Notes in Computer Science, 2014
Lecture Notes in Computer Science, 2005
Artificial Intelligence in Industrial Decision Making, Control and Automation, 1995
Planes are important geometric features and can be used in a wide range of applications. This wor... more Planes are important geometric features and can be used in a wide range of applications. This work aims to illustrate a homography-based method to detect planes using the affine model. Using two image frames from a monocular sequence, a set of match pairs of interest points is obtained using Harris corner detector combined with the Scale Invariant Feature Transform (SIFT) as local descriptor. A Delaunay triangulation is performed on the set of detected corners of the first image and we use a triangle to calculate an affine homography that represents a plane in the image. An algorithm was developed to cluster interest points belonging to the same plane based on the reprojection error of the affine homography. Tests are performed in different sequences of indoor and outdoor images and results are shown.
Autonomous mobile robots rely on multiple sensors to perform a varied number of tasks in a given ... more Autonomous mobile robots rely on multiple sensors to perform a varied number of tasks in a given environment. Different tasks may need different sensors to estimate different subsets of world state. Also, different sensors can cooperate in discovering common subsets of world state. This paper presents a new approach to multimodal sensor fusion using dynamic Bayesian networks and an occupancy grid. The environment in which the robot operates is represented with an occupancy grid. This occupancy grid is asynchronously updated using ...
2. The Basis Of The Method In this paper we present a texture analysis and classification method ... more 2. The Basis Of The Method In this paper we present a texture analysis and classification method based on the computation of highorder statistics of gray-levels or selected features. The computation of these high-order statistics can be performed in parallel in non-overlapping windows that cover the full image. Classification is performed by neural networks. In this extended summary results are presented for back-propagation networks only, but several other types of neural nets were also tested.