Point target detection methods based on multi-frame association

Noise and clutter could seriously degrade performance of point target detection, and multi-frame association methods can be used to improve the probability of detection. To figure out detection model and applicability of different methods with multi-frame data, research on SNR (signal to noise ratio) of multi-frame superposition, multi-frame difference and joint probability methods are carried out. Signal gain and clutter restrain coefficient are proposed to revise general model based on affection of image registration. By analyzing signal variation in multi-frame process, influence of coefficient range and frame number to SNR is obtained. The study concluded that detection performance can be improved by multi-frame association significantly. Different methods are proper for specific scenes. Superposition method is applicable to general clutter scene, difference method for severe clutter scene, and joint probability for rapid changes. Superposition and difference methods are sensitive to image registration and better performance can be achieved at sub-pixel precision. The conclusion of this paper can support index design and detection method selection.