Members: Ido Azulai, Asaf Granica

Supervisor: Prof. Tzevi Batus

The orientation of a vehicle, such as a rocket, is typically measured using an Inertial Measurement Unit (IMU), which integrates angular rates to calculate orientation. This method is susceptible to the inherent drift of IMU sensors, necessitating additional measurement methods to improve estimation accuracy.

To overcome this limitation, we have developed an optical system for orientation measurement via horizon detection. This system is designed to be part of the two-stage supersonic rocket developed by the Hebrew University Rocketry Club, capable of flying vertically to an altitude of 4.5 km and returning.

Our system consists of four cameras in a panoramic configuration, connected to a real-time embedded computer. We implemented a contour detection algorithm, which combines live video streams from the camera set and calculates the vehicle's orientation in terms of yaw, pitch and roll angles.

The approaches tested included various image processing techniques and machine learning models. The primary algorithm uses color filtering in the HSV color space to distinguish sky and ground, followed by edge detection and contour analysis to identify the horizon line. Linear regression is applied to fit a line to the detected contour, allowing the algorithm to estimate the vehicle's orientation by continuously updating and refining this line.

We estimate that our system can predict pitch and yaw angles with an error margin within 0.5° and exhibits less drift bias compared to the IMU. The main challenges in this work were achieving real-time processing speeds and ensuring the system's robustness under the variable conditions of rocket flight.

Ultimately, the system's data and the IMU data will be integrated to cancel out each other's biases, aiming to provide precise orientation data, which is crucial for the accurate navigation and control of rockets, thereby enhancing the capabilities of current aerospace technology.