Deep Learning based approach for Range Estimation

Range analysis is one of the most sought after topics in evaluation of airborne flight vehicles. Deterministic and statistical methods have been used for estimation of the range of flight vehicles from a point on ground. The present industrial implementation involves use of redundant sensors and apparatus at two far-off sites which has always been a problem. In this paper, we find a method to estimate the range of a flight vehicle with the use of a single camera. Deep Learning based depth mapping has provided appreciable results for shorter ranges, but this method remains unexplored for longer ranges. A Convolutional Neural Networks(CNN) based approach fares equally well with the other methods of monocular depth estimation such as linear regression, support vector machine based regression, decision tree regression by treating the range estimation problem as a regression problem. A comparison has been done among various methods that have been used to approach the problem.