Iñigo Barandiaran

Real-time optical markerless tracking for augmented reality applications

Augmented reality (AR) technology consists in adding computer-generated information (2D/3D) to a real video sequence in such a manner that the real and virtual objects appear coexisting in the same world. To get a realistic illusion, the real and virtual objects must be properly aligned with respect to each other, which requires a robust real-time tracking strategy—one of the bottlenecks of AR applications. In this paper, we describe the limitations and advantages of different optical tracking technologies, and we present our customized implementation of both recursive tracking and tracking by detection approaches. The second approach requires the implementation of a classifier and we propose the use of a Random Forest classifier. We evaluated both approaches in the context of an AR application for design review. Some conclusions regarding the performance of each approach are given.

A Perspective on Knowledge Based and Intelligent Systems Implementation in Industrie 4.0

Abstract A worldwide trend in advanced manufacturing countries is defining Industrie 4.0, Industrial Internet and Factories of the Future as a new wave that can revolutionize the production and its associated services. Cyber-Physical Systems (CPS) are central to this vision and are entitled to be part of smart machines, storage systems and production facilities able to exchange information with autonomy and intelligence. Such systems should be able to decide and trigger actions, and control each other independently and for such reason it is required the use of Knowledge based and intelligent information approaches. In this paper we present our perspective on how to support Industrie 4.0 with Knowledge based and intelligent systems. We focus in the conceptual model, architecture and necessary elements we believe are required for a real world implementation. We base our conceptualization in the experiences gathered during the participation in different ongoing research projects where the presented architecture is being implemented.

Realtime Dense Stereo Matching with Dynamic Programming in CUDA

Real-time depth extraction from stereo images is an important process in computer vision. This paper proposes a new implementation of the dynamic programming algorithm to calculate dense depth maps using the CUDA architecture achieving real-time performance with consumer graphics cards. We compare the running time of the algorithm against CPU implementation and demonstrate the scalability property of the algorithm by testing it on different graphics cards.

An automatic segmentation and reconstruction of mandibular structures from CT-data

In any medical data analysis a good visualization of specific parts or tissues are fundamental in order to perform accurate diagnosis and treatments. For a better understanding of the data, a segmentation process of the images to isolate the area or region of interest is important to be applied beforehand any visualization step. In this paper we present a method for mandibular structure surface extraction and reconstruction from CT-data images. We tested several methods and algorithms in order to find a fast and feasible approach that could be applicable in clinical procedures, providing practical and efficient tools for mandibular structures analysis.

Real-time depth map generation architecture for 3D videoconferencing

In this paper we present a reliable depth estimation system which works in real-time with commodity hardware. The system is specially intended for 3D visualization using autostereoscopic displays. The core of this work is an implementation of a modified version of the adaptive support-weight algorithm that includes highly optimized algorithms for GPU, allowing accurate and stable depth map generation. Our approach overcomes typical problems of live depth estimation systems such as depth noise and flickering. Proposed approach is integrated within the versatile GStreamer multimedia software platform. Accurate depth map estimation together with real-time performance make proposed approach suitable for 3D videoconferencing.

Statistical tuning of adaptive-weight depth map algorithm

In depth map generation, the settings of the algorithm parameters to yield an accurate disparity estimation are usually chosen empirically or based on unplanned experiments. A systematic statistical approach including classical and exploratory data analyses on over 14000 images to measure the relative influence of the parameters allows their tuning based on the number of bad pixels. Our approach is systematic in the sense that the heuristics used for parameter tuning are supported by formal statistical methods. The implemented methodology improves the performance of dense depth map algorithms. As a result of the statistical based tuning, the algorithm improves from 16.78% to 14.48% bad pixels rising 7 spots as per the Middlebury Stereo Evaluation Ranking Table. The performance is measured based on the distance of the algorithm results vs. the Ground Truth by Middlebury. Future work aims to achieve the tuning by using significantly smaller data sets on fractional factorial and surface-response designs of experiments.

Extending Marching Cubes with Adaptative Methods to Obtain More Accurate Iso-surfaces

This work proposes an extension of the Marching Cubes algorithm, where the goal is to represent implicit functions with higher accuracy using the same grid size. The proposed algorithm displaces the vertices of the cubes iteratively until the stop condition is achieved. After each iteration, the difference between the implicit and the explicit representations is reduced, and when the algorithm finishes, the implicit surface representation using the modified cubical grid is more accurate, as the results shall confirm. The proposed algorithm corrects some topological problems that may appear in the discretization process using the original grid.

AN EMPIRICAL EVALUATION OF INTEREST POINT DETECTORS

Image interest point extraction and matching across images is a commonplace task in computer vision–based applications, across widely diverse domains, such as 3D reconstruction, augmented reality, or tracking. We present an empirical evaluation of state-of-the-art interest point detection algorithms measuring several parameters, such as efficiency, robustness to image domain geometric transformations—that is, similarity—affine or projective transformations, as well as invariance to photometric transformations such as light intensity or image noise.

Exploring simple visual languages for real time human-computer interaction

We want to explore the expressive power of visual languages that can be recognized in real time with off-the-shelf components. Real time constraints imply that we are only able to detect and track color blobs in the image. We test two basic applications: remote control of a mobile robot and gesture based music synthesis. The former requires accuracy in movements, and on the contrary, creativity is important in the latter.

ROBOTIC RESEARCH PLATFORM FOR IMAGE-GUIDED SURGERY ASSISTANCE

In the context of surgery, it is very common to face challenging scenarios during the preoperative plan implementation. The surgical technique’s complexity, the human anatomical variability and the occurrence of unexpected situations generate issues for the intervention’s goals achievement. To support the surgeon, robotic systems are being integrated to the operating room. However, current commercial solutions are specialized for a particular technique or medical application, being difficult to integrate with other systems. Thus, versatile and modular systems are needed to conduct several procedures and to help solving the problems that surgeons face. This article aims to describe the implementation of a robotic research platform prototype that allows novel applications in the field of image-guided surgery. In particular, this research is focused on the topics of medical image acquisition during surgery, patient registration and surgical/medical equipment operation. In this paper, we address the implementation of the general purpose teleoperation and path following modes of the platform, which constitute the base of future developments. Also, we discuss relevant aspects of the system, as well as future directions and application fields to investigate.