Study On The Pose Prediction Method For Moving Helmet Mounted Sight

Helmet mounted sight is a sighting device installed in the pilot’s helmet, which isplaying an increasingly important role in modern airfight of combat aircrafts.However, accurate and quick positioning for pilot’s helmet is an importantprerequisite for the helmet mounted sight performing its function adequately. In themonocular vision positioning system for helmet mounted sight, as the limitation offrame rate of area-array CCD, the measurement speed of helmet’s pose is limitedseriously. Consequently, how to improve the measuring speed of helmet mountedsight pose has become more and more important in the present study. To solve thisproblem, theoretical as well as experimental studies of pose prediction methods formoving helmet mounted sight, based on recursive least squares and Kalman filteringtheory respectively, are studied in this paper. The specific work are as follows:1. The significance of pose prediction for moving helmet mounted sight isanalyzed. The prediction methods for moving object pose, as well as its researchsituations at home and abroad, are studied. Finally, the research value and purpose ofthis paper is described.2. A motion-matrix prediction method based on recursive least-squares theory isstudied. The observations are chosen as the pose measurement values based on thepoint-feature-positioning method. The fixed-time prediction model for helmet’s poseis set up using recursive motion matrix, and corresponding computer simulation isprovided, too.3. To improve the poor capability of anti-jamming for helmet’s posemeasurement noise in motion-matrix prediction method, another pose predictionmethod based on adaptive extended Kalman filtering is studied. The camera’sperspective transformation model is combined with the state-space model in Kalmanfiltering, and the pose measurement process is skipped, too. Therefor, the helmet posecan be predicted and estimated directly by the adaptive extended Kalman filteringthrough the ideal image coordinates in the CCD image plane of the positioning featurepoints on helmet mounted sight.4. To overcome the shortcomings of low prediction and estimation precision for helmet pose in the prediction method based on adaptive extended Kalman filtering, anovel pose prediction method based on adaptive Kalman filtering is studied. Theobservations are also chosen as the pose measurement values based on thepoint-feature-positioning method. The helmet’s pose prediction model is established,and then, fixed-time prediction for its poses at future timestamps is achieved.5. To validate all the studied prediction methods for helmet pose in this paper, asimulated experimental system is set up. The experimental results show that theprediction method based on adaptive Kalman filtering can overcome the defects of themethods using motion matrix and adaptive extended Kalman filtering. Furthermore,the adaptive Kalman filtering prediction method can realize the fixed-time predictionfor helmet’s poses at future timestamps accurately, as a result, the measuring speed ofhelmet mounted sight pose can be improved effectively, and its measuring precisioncan be improved partly, too.