Nikolaos Kyriakoulis

Color-Based Monocular Visuoinertial 3-D Pose Estimation of a Volant Robot

The pose estimation from visual sensors is widely practiced nowadays. The pose vector is estimated by means of homographies and projection geometry. The integration of visual and inertial measurements is getting more attractive due to its robustness and flexibility. The cooperation of visual with inertial sensors for finding a robots pose bears many advantages, as it exploits their complementary attributes. Most of the visual pose estimation systems identify a geometrically known planar target to extract the pose vector. In this paper, the pose is estimated from a set of colored markers arranged in a known geometry, fused with the measurements of an inertial unit. The utilization of an extended Kalman filter (EKF) compensates the error and fuses the two heterogeneous measurements. The novelty of the proposed system is the use of low-cost colored post-it markers, along with the capability of handling different frames of reference, as the camera and the inertial unit are mounted on different mobile subsystems of a sophisticated volant robotic platform. The proposed system is computationally inexpensive, operates in real time, and exhibits high accuracy.

Dense disparity estimation using a hierarchical matching technique from uncalibrated stereo vision

In motion estimation, the sub-pixel matching technique involves the search of sub-sample positions as well as integer-sample positions between the image pairs, choosing the one that gives the best match. Based on this idea, the proposed disparity estimation algorithm performs a 2-D correspondence search using a hierarchical search pattern. The disparity value is then defined using the distance of the matching position. Therefore, the proposed algorithm can process non-rectified stereo image pairs, maintaining the computational load within reasonable levels.

Comparison of data fusion techniques for robot navigation

This paper proposes and compares several data fusion techniques for robot navigation. The fusion techniques investigated here are several topologies of the Kalman filter. The problem that had been simulated is the navigation of a robot carrying two sensors, one Global Positioning System (GPS) and one Inertial Navigation System (INS). For each of the above topologies, the statistic error and its, mean value, variance and standard deviation were examined.

A recursive fuzzy systemfor efficient digital image stabilization

Correspondence should be addressed to Antonios Gasteratos, gasteratos@ieee.orgReceived 14 March 2008; Accepted 23 May 2008Recommended by Zne-Jung LeeA novel digital image stabilization technique is proposed in this paper. It is based on a fuzzy Kalman compensation of the globalmotion vector (GMV), which is estimated in the log-polar plane. The GMV is extracted using four local motion vectors (LMVs)computed on respective subimages in the logpolar plane. The fuzzy Kalman system consists of a fuzzy system with the Kalmanfilter’s discrete time-invariant definition. Due to this inherited recursiveness, the output results into smoothed image sequences.The proposed stabilization system aims to compensate any oscillations of the frame absolute positions, based on the motionestimation in the log-polar domain, filtered by the fuzzy Kalman system, and thus the advantages of both the fuzzy Kalman systemand the log-polar transformation are exploited. The described technique produces optimal results in terms of the output qualityand the level of compensation.Copyright

A Rotational and Translational Image Stabilization System for Remotely Operated Robots

Remotely operated robots equipped with on board cameras, apart from providing video input to operators, perform optical measurements to assist their navigation as well. Such image processing algorithms require image sequences, free of high frequency unwanted movements, in order to generate their optimal results. Image stabilization is the process which removes the undesirable position fluctuations of a video sequence improving, therefore, its visual quality. In this paper, we introduce the implementation of an image stabilization system that utilizes input from an on board camera and a gyrosensor. The frame sequence is processed by an optic flow algorithm and the inertial data is processed by a discrete Kalman filter. The compensation is performed using two servo motors for the pan and tilt movements and frame shifting for the vertical and horizontal movements. Experimental results of the robot head, have shown fine stabilized image sequences and a system capable of processing 320 times 240 pixel image sequences at approximately 10 frames/sec, with a maximum acceleration of A deg/sec2.

An adaptive fuzzy system for the control of the vergence angle on a robotic head

An important issue in realizing robots with stereo vision is the efficient control of the vergence angle. In an active robotic vision system the vergence angle along with the pan and tilt ones determines uniquely the fixation point in the 3D space. The vergence control involves the adjustment of the angle between the two cameras’ axes towards the fixation point and, therefore, it enables the robot to perceive depth and to compute obstacle maps. Vergence movement is directly related to the binocular fusion. Additionally, the decision for convergence or divergence is extracted either by motion affine models or by mathematical ones. In this paper, a new method for extracting the cameras’ movement direction is presented. The movement decision is performed by an adaptive fuzzy control system, the inputs of which are the zero-mean normalized cross correlation (ZNCC) and the depth estimations at each time step. The proposed system is assessed on a 4 d.o.f. robotic head, yet it can be utilized in any active binocular system, since it is computationally inexpensive and it is independent to a priori camera calibration.

Pose estimation of a volant platform with a monocular visuo-inertial system

One of the serious problems in robotics applications is the estimation of the robots pose. A lot of research effort has been put on finding the pose via inertial and proximity sensors. However, the last decades many systems adopt vision to estimate the pose, by using homographies and projection geometry. In this paper the pose estimation is achieved by the identification of a geometrically known platform from one camera and from the measurements of an inertial unit. The extended Kalman filter (EKF) is used for data fusion and error compensation. The novelty of this system is that the visual sensor and the inertial unit are mounted on different mobile systems. The proposed pose estimation system exhibits high accuracy in real-time.

On Visuo-Inertial Fusion for Robot Pose Estimation Using Hierarchical Fuzzy Systems

This article presents a novel application of a three level fuzzy hierarchical system for the fusion of visual and inertial pose estimations. The goal is to provide accurate and robust pose measurements of an indoor volant robot, which operates in real working conditions, in order to facilitate its control. The first level corrects the error of the inertial measurement unit based on the acceleration measurements. The second level fuses the measurements from the visual sensor with the ones from the first level. Finally, the output of the fuzzy hierarchical system is available at the third level; the inputs of which are the previous systems state and the one of the second level. The achieved results are compared to ground truth pose estimations, which are used to analyze the behavior of the proposed fusion system against possible changes. The system provides accurate and precise measurements, while its straightforward design allows real-time implementation.

Multi-camera 3D scene reconstruction from vanishing points

In this paper we present a multi-camera system, which serves as the primary vision apparatus for a ceiling based swinging service robotic platform. To facilitate a smooth and unimpeded movement of the swinging platform the knowledge of the working environment is essential. To this end we examine the 3D scene reconstruction by concurrent multiple views from the distinct cameras located at the upper corners of a room for volume calculation of objects in everyday indoor environments. The 3D scene reconstruction is used to determine the working volume where the swinging robot will be able to operate. At first, we detect lines across the views using the hough transformation while the computation of multiple vanishing points serves as the platform to distinguish regions across the multi camera system. We thereafter obtain correspondences across the multiple views of the test room and finally we determine the 3D volumes of objects across the scene. The preliminary experimental results show that the volume computation is very accurate and reasonably time consuming.

An efficient method to reconstruct prismatic rigid objects by multiple camera arrangements

An efficient method to reconstruct prismatic rigid objects by multiple camera arrangements is presented in this paper. The proposed method estimates accurately the dominant vanishing points in remarkably short time. First, concurrent multiple views from the distinct cameras located at the upper corners of a room are used as the input to the algorithm. An implementation of the Radon transform is then applied in order to detect line segments across the respective views. An initial hypothesis about the existing vanishing points is made. The J-linkage algorithm is utilized to cluster the different groups of line segments and to refine the calculation of the respective vanishing points. The regions across the multi-camera system are then distinguished and the correspondences across the multiple views of the test room are obtained. Last, the volume of objects is estimated accurately with multiple cross ratios along the images and is represented by an octree decomposition technique.