Computer Vision Based Realtime Opto-Mechatronic Tracking System

For purpose of high accuracy image acquisition of moving objects with a finite shutter speed of the observing camera, a camera-tracking mount is required to avoid blurred objects during the exposure time. The tracking mount could have one-, two- or multi-axis kinematics with a fixed location and camera perspective or variable location with the requirement of calibration of the relative camera coordinate system. The calibration calculates the correlation of camera position, kinematics of the mount and the movement of observed objects. For testing, in this thesis a two-axis equatorial mount is used to track astronomical objects such as stars and satellites. The exact solution of the forward kinematics is calculated, which includes the coordinate transformation of the projection of astronomical objects, represented by the camera data, and the mount coordinate system. Further a reduced kinematics is derived from the exact solution, to provide a generic two-axis object tracking system using inverse kinematics. A high-performance optimised multiple object tracking software is developed, which detects objects and calculates the two-axis movement of the mount in real-time, based on the two-dimensional deviation in the acquired image. For this reason, optimised computer-vision based algorithms on a SoC (system-on-a-chip) system are used together with a newly developed electronic control interface.