A Revised Monte Carlo Method for Target Location with UAV

Abstract

Target location using UAV equipped with vision system has played an important role in many applications but there remain challenges. One of the principal difficulties is to position a target with a high accuracy, particularly in some specific conditions. There are many factors impacting location accuracies, such as turret setup process, sensors intrinsic properties, movement noise and GPS data precision. The most common and notable factors are the movement noise and sensors noise, which are tricky to be eliminated or compensated. Solutions to dealing with noise are mainly from methods such as recursive least square method, least square and Kalman filtering methods. But these routine methods will meet their bottlenecks when locating some plane based targets, a common scenario in target location applications. In this case, the usual methods are subject to target pointing bias of line of sight owing to the specific geometric condition. To eliminate this kind of location bias, an improved Monte Carlo method is proposed in this paper which first estimates the bias of pointing deviation for each measurement with statistical methods and then subtracts the estimated biases in a variance optimization process. Relevant experiments are conducted showing an obvious advantage of the proposed method over the other methods.