Engineering topics

Cigla, Cevahir

Publications and source records attributed to Cigla, Cevahir.

2 records

NASA NTRS2018

Onboard Stereo Vision for Drone Pursuit or Sense and Avoid

Wedescribeanew,on-board,shortrangeperceptionsystem that enables micro aerial vehicles (MAVs) to detect, track, and follow or avoid nearby drones (within 2-20 meters) in GPS-denied environments. Each vehicle is able to sense its neighborhood and adapt its motion accordingly without use of centralized reasoning or inter-vehicle communication. To enable a lightweight, low power solution, on-board stereo cameras are used for detection and tracking with depth images, while a downward-looking camera and an inertial measurement unit are used to estimate the position of the observer without use of GPS. We illustrate the robustness and accuracy of this approach through real-time, outdoor leader-follower experiments with three quadrotors. Our experiments show that state-of-art trackers are far less robust in detection against cluttered background. This demonstrates that stereo vision is a highly effective approach to perception for safe navigation of multiple MAVs in close proximity.

Matthies, Larry↗

NASA NTRS2017

Gaussian Mixture Models for Temporal Depth Fusion

Sensing the 3D environment of a moving robot is essential for collision avoidance. Most 3D sensors produce dense depth maps, which are subject to imperfections due to various environmental factors. Temporal fusion of depth maps is crucial to overcome those. Temporal fusion is traditionally done in 3D space with voxel data structures, but it can be approached by temporal fusion in image space, with potential benefits in reduced memory and computational cost for applications like reactive collision avoidance for micro air vehicles. In this paper, we present an efficient Gaussian Mixture Models based depth map fusion approach, introducing an online update scheme for dense representations. The environment is modeled from an ego-centric point of view, where each pixel is represented by a mixture of Gaussian inverse-depth models. Consecutive frames are related to each other by transformations obtained from visual odometry. This approach achieves better accuracy than alternative image space depth map fusion techniques at lower computational cost.

Matthies, Larry↗