High-precision ground validation of TVC systems using super-resolution enhanced stereo vision

Abstract

Ground validation of Thrust Vector Control (TVC) systems is crucial for ensuring accurate nozzle movement and positioning. Traditionally, verifying these systems requires expensive mechanical setups or high-end metrology tools like Laser Trackers, which can be costly and difficult to operate. To solve this problem, we present a cost-effective, non-contact stereo vision method to determine the absolute position and orientation of the nozzle. Our approach uses a commercial ZED stereo camera and Aruco markers to map the nozzle plane and body plane. A key part of our method involves using a deep learning-based super-resolution technique (LapSRN) combined with adaptive thresholding. This effectively solves the issue of resolution loss when measuring from a distance, ensuring precise depth estimation. We tested the system on a functional electromechanical TVC unit and compared the results against a Leica Absolute Tracker and high-precision encoders. The tests showed a maximum angular error of 0.1642° and estimated the gimbal’s rotation center with an accuracy of approximately 1.12 mm. These findings prove that our stereo vision framework is a reliable and affordable alternative to traditional verification techniques. It offers significant potential for Ground Support Equipment (GSE), especially for TVC designs where the rotation center is difficult to pinpoint, such as flexible-joint systems.