Algorithm for Point Target Detection With Extreme Wide Angle Staring Infrared Imaging System

摘要

The paper builds the models of background, noise and target in fish-eye lenses or extreme wide-angle staring Imaging fields of view. It uses difference method every two frames to generate new consecutive frames and self adapting threshold division To remove White noise and rationally segment image after suppressing the background,. According to characteristics of noise is random distribution, degree of correlation is low, and target is high correlative, the paper describes a Logical And method to separate target from residual noise, which can cumulate point target trajectories, strengthen target contrast intensity, improve its sign-to-noise ratio. Experiment and simulation have been done in the field and laboratory. The result indicates this algorithm is effective to detect multi-pixel targets whose velocities are between 0.2 to 1 pixels per frame and sign-to-noise ratio are 3 or less.