Attitudes estimation by machine vision

In today´s life Inertial Navigation System (INS) is the main source of finding the attitudes i.e. roll pitch and yaw of a flying object for navigation purposes. Due to the complexity and high cost of this system vision based navigation systems are introduced in this field for exact localization. In both systems rate of change of one function relative to other function is determined. A very sophisticated INS system is required to give an accurate instant values of roll, pitch and yaw of an object at any state during flight, which is much expensive. Vision based techniques are used for localization and aerial navigation. It constitutes straightforward-cheap method to estimate the object location. A single consumer grade camera can replace a typical expensive suit (encoders, IMU etc.). The visual information from successive aerial images is used to estimate the movements of the flying object by applying phase correlation techniques. With the help of these correlation methods the images are registered with each other. The requirements for a successful registration are sufficient illumination in the environment, dominance of static scene over moving objects, enough texture to allow apparent motion to be extracted, and sufficient scene overlap between consecutive frames. In the proposed method modified normalized phase correlation has been used. In particular Gram polynomial basis functions are applied to remove the Gibbs error problem. When flying object changes its direction or AGL (altitude above ground level) the pixel change between two consecutive images is calculated and converted to degrees along and across the flight direction. Which results in the roll and yaw by machine vision. The proposed method is applied on a real time data set and the attitudes results are compared with available INS system´s results. As per available information of the INS system, the attitude error growth was typically in the range of 2 to 3 degrees/ hr i.e. 1 sigma. The results achie- ed by the proposed method are comparable with the available system.