Method for detection of moving object of approximately known size in conditions of low signal-to-noise ratio

A method for detection of a moving object of approximately known size in conditions of low signal-to-noise ratio comprising:

• specifying a set of hypotheses (pairs “azimuth—speed”) that describe all possible states of the object in a frame of data points representing a field of view;
• updating regularly the frame representing the field of view;
• smoothing, on each update, the brightness of data points of a new frame by replacing the brightness of each data point with the average brightness of data points surrounding this point and representing a hypothetical object;
• recalculating, on each update, the accumulated frame of each specified hypothesis as the sum of the accumulated frame of this hypothesis from the preceding update and the new smoothed frame shifted in accordance with the transfer of the object between updates;
• determining, on each update, a hypothesis that represent the actual state of the object the best as a hypothesis that produces the accumulated frame with the brightest point;
• displaying the accumulated frame of the best hypothesis after each update;
• determining the moment when the object is detected as the moment when the best hypothesis stabilizes in a sequence of updates and the movement of the brightest point in the screen becomes consistent with the azimuth and speed of this hypothesis;
• reporting the detected coordinates of the object, its azimuth and speed to a surveillance system.