Lavanya Karthikeyan, Department of Computer Science and Engineering, Amrita School of Engineering-Bangalore, India; Malavika R. Nair, Department of Computer Science and Engineering, Amrita School of Engineering-Bangalore, India; S.V. Apoorva, BMS Institute of Technology and Management, India; Vinod Kumar, UR Rao Satellite Centre, India

View Abstract Sign in for premium content

Abstract:

Space exploration requires autonomous operations of space vehicles. This involves guidance and navigation of space crafts. On the Earth several navigation techniques are available viz. global navigation satellite system (GNSS), radar-based techniques etc. On other planets these conventional methods of navigation are not available, presently. In such a situation, computer vision-based techniques come to rescue. Digital image processing or computer vision plays a vital role pertaining to horizontal velocity estimation of a descent lander craft. In landing missions, it is very much essential to estimate the horizontal velocity of a lander accurately in order to achieve safe and successful landing. The attempt to reduce the errors produced by conventional methods is what led to visionbased velocity estimation methods. This project thus proposes a horizontal velocity estimation model that tries to reduce the hardware by implementing computer intelligence techniques and aims at presenting more accurate results. Our algorithm uses images obtained from the Lunar Reconnaissance Orbiter Camera (LROC) and Lunar Planetary Institute (LPI) as input and applies digital image processing algorithms to extract the SURF (Speeded up Robust Features) features. Once the key points are matched, Random Sample Consensus (RANSAC) algorithm is used in order to remove the vertical motion and the outliers. Following this, horizontal motion vectors are estimated by applying Lucas-Kanade algorithm using optical flow. Finally, the horizontal velocity of the descent lander is estimated between two consecutive images. Two images are taken at an altitude of 45.705 km with a time interval of 2.02s. The magnitude of the motion vector obtained is 1.9743-pixel shift between the two images. This gives us the horizontal velocity of 336.64m/s. Further Monte Carlo analysis is carried out, for the thirty-four sets of image pairs at different altitudes varying from 40 to 112 km. The estimated horizontal velocity at an altitude of 45, 47 and 54 km is obtained as 212.58, 246.58 and 358.75 m/s, respectively. In addition to this, horizontal velocity of 26.2894 m/s is estimated from the two images taken at an altitude of 1000m and 500m. This value is estimated to be 5.28 m/s for the images taken at an altitude of 200m and 100 m; from LROC images with a time interval of 4s each.