Method and apparatus for computer vision analysis of cannon-launched artillery video

摘要

An automated method to quantify the pitching and yawing motion of a projectile during ballistic flight using two camera/tracker video systems. Image processing tools are used to segment the shape of the projectile in each frame of a launch video, which allows the location and observed pitch angle to be calculated with sub-pixel accuracy. Subsequent automated analysis uses the history of the projectile location and the pitching behavior to calculate estimates for the epicyclic motion, as well as other ballistic parameters such as aeroballistic coefficients. Using two cameras located at different orthographic views of the line-of-fire (LOF) allows the pitching and yawing motion history of the projectile to be calculated in three dimensions (3D). In addition, input of the camera locations, cannon trunnion location, and the cannon pointing direction allows for automatic correction for camera misalignment.

主权项

1. A method and apparatus for computer vision analysis of video taken of a marked projectile launched from a gun, the method comprising:
receiving a first projectile launch video recording of a marked projectile launched from a gun generated by a first video camera system, said first projectile launch video having a plurality of first image frames, at least a portion of said plurality of first image frames containing a projectile shape corresponding with said marked projectile; receiving a second projectile launch video recording of said marked projectile launched from said gun generated by a second video camera system, said second video camera system being positioned at a different location from said first video camera system, said second projectile launch video having a plurality of second image frames, at least a portion of said plurality of second image frames containing a projectile shape corresponding with said marked projectile; segmenting a projectile shape, if any, in each of said first image frames of said plurality of first image frames, and generating a first pitch history; segmenting a projectile shape, if any, in each of said second image frames of said plurality of second image frames, and generating a second pitch history; receiving first mirror rotation data corresponding with said first projectile launch video recording, correcting said first pitch history for skew angle and camera misalignment, and generating a first corrected pitch history; receiving second mirror rotation data corresponding with said second projectile launch video recording, correcting said second pitch history for skew angle and camera misalignment, and generating a second corrected pitch history; receiving said first corrected pitch history and said second corrected pitch history and geodetic information associated with the location geometry of said first video camera system, said second video camera system, and said gun, and determining the pitch and the yaw of said projectile based on said first corrected pitch history, said second corrected pitch history, and said geodetic information; and outputting said pitch and said yaw of said projectile.