Using depth in visual simultaneous localisation and mapping

2013

J Intell Robot Syst

Self Cite

We present a quadrotor Micro Aerial Vehicle (MAV) equipped with four cameras, which are arranged in two stereo configurations. The MAV is able to perform stereo matching for each camera pair on-board and in real-time, using an efficient sparse stereo method. In case of the camera pair that is facing forward, the stereo matching results are used for a reduced stereo SLAM system. The other camera pair, which is facing downwards, is used for ground plane detection and tracking. Hence, we are able to obtain a full 6DoF pose estimate from each camera pair, which we fuse with inertial measurements in an extended Kalman filter. Special care is taken to compensate various drift errors. In an evaluation we show that using two instead of one camera pair significantly increases the pose estimation accuracy and robustness.

Section: Related Workmentioning

confidence: 99%

“…2 The successfully matched features are used for estimating the current MAV pose. A SLAM system is employed for this task, which is based on the method proposed in [17]. This method is an adaptation of PTAM that incorporates depth information, which we receive from stereo matching.…”

Section: Processing Of Forward-facing Camerasmentioning

confidence: 99%

On-Board Dual-Stereo-Vision for the Navigation of an Autonomous MAV

Schauwecker

2013

J Intell Robot Syst

Self Cite

“…KinectFusion [3], which uses coarse-to-fine iterative closest point with projective data association implemented on graphics hardware for registration with the map. The authors themselves in [4] introduced bundle adjustment with depth constraints, which allows us to easily extend monocular visual SLAM systems like the very efficient PTAM system [5] to also utilize depth measurements of RGBD data. This method was shown to enable autonomous flight of a MAV with a stereo camera in [6], but has some systematic limitations: For optimizing the full map, PTAM relies on global bundle adjustment, which quickly becomes computationally infeasible in real time for large numbers of keyframes.…”

Section: Related Workmentioning

confidence: 99%

“…Deviating from [4], we need to modify our bundle adjustment formulation to accomodate the relative representation chosen for this work. Given two keyframes, let us call map points that were measured not only in their source keyframe S but also in the other keyframe O relevant map points.…”

Section: B Relative Bundle Adjustment With Depth Constraintsmentioning

confidence: 99%

See 1 more Smart Citation

Efficient onbard RGBD-SLAM for autonomous MAVs

Scherer

2013 IEEE/RSJ International Conference on Intelligent Robots and Systems

2013

Self Cite

Abstract-We present a computationally inexpensive RGBD-SLAM solution taylored to the application on autonomous MAVs, which enables our MAV to fly in an unknown environment and create a map of its surroundings completely autonomously, with all computations running on its onboard computer. We achieve this by implementing efficient methods for both tracking its current location with respect to a heavily processed previously seen RGBD image (keyframe) and efficient relative registration of a set of keyframes using bundle adjustment with depth constraints as a front-end for pose graph optimization. We prove the accuracy and efficiency of our system based on a public benchmark dataset and demonstrate that the proposed method enables our quadrotor to fly autonomously.

A Comparison of 3D Sensors for Wheeled Mobile Robots

Rauscher

Dubé

Intelligent Autonomous Systems 13

2015

Self Cite

3D Sensors are used for many different applications, e.g. scene reconstruction, object detection and mobile robots, etc. Several studies on usability and accuracy have been done for different sensors. However, all these studies have used different settings for the different sensors. For this reason we compare five 3D sensors, including the structured light sensors Microsoft Kinect and ASUS Xtion Pro Live and the time of flight sensors Fotonic E70P, IFM O3D200 and Nippon Signal FX6, using the same settings. The sensor noise, absolute error and point detection rates are compared for different depth values, environmental illumination and different surfaces. Also simple models of the noise depending on the measured depth are proposed. It is found that the structured light sensors are very accurate for close ranges. The time of flight sensors have more noise, but the noise does not increase as strongly with the measured distance. Further, it is found that these sensors can be used for outdoor applications.