STUDY OF COMPUTER VISION ALGORITHMS FOR SPACE TRACKING SYSTEMS IN TYPICAL MODES OF THEIR FUNCTIONING

Abstract

The paper proposes an algorithm for an object capturing and tracking in a frame for automatic phase focusing of a single-objective camera with a mirror obturator. The capture is performed by the object identification through the filtering of boundaries and edges by the Canny operator; then the Hough transformation is used to detect the characteristic lines of the object. Reliable support of the object is provided by a predictor/equalizer based on the Kalman filter. The proposed calculation algorithm makes it possible to achieve optimal performance, sufficient for the object capturing and tracking in the frame and its timely exposure. This algorithm was implemented to track an object within the scene, the trajectory and speed of which are unknown in advance that excludes the possibility of creation a self-learning algorithm. The Canny operator provides the most accurate detection of object boundaries from existing filtering methods that simplifies the subsequent image processing. The application of an additional Hough transformation makes it possible to speed up the calculations by reducing the amount of data processed, confirming the overall speed of the algorithm as compared to the classical filtering method. The usage of the Kalman filter as a predictor/equalizer gives the possibility to pre-determine the point for focusing at the next time. The proposed calculation algorithm makes it possible to achieve optimal performance, sufficient for the object capturing and tracking on the stage, and also sufficient for timely exposure of the frame.

Keywords

• tracking system
• boundary detector
• object tracking
• Hough transformation
• Kalman filter