The Analysis of Target Detection Probability In Sequence Image

摘要

The detection of infrared small and puny target is the critical problem for target tracking and recognition based on imaging sensors. Being limited to the complicated environment factors, measurement scale and precision of the sensors, the measurements are uncertain, imprecise or incomplete on certain level, and thus brings the difficulty for detecting target in real image. Usually the binary hypothesis method is used to examine the potential target from background in a single frame, then the detection probability can be raised and the false alarm probability can be reduced through the method of detection multi-frame of the sequence image (namely K/N rule). In this paper, the target detection probability P_d and the false alarm probability P_f for a single image is calculated on the basis of the probabiliry density function p (Z|H_0) and p (Z|H_1) on judgement regions Z_0 and Z_1. In N times measurement of a sequence image, the target is detected at least K times, then it can determine that the target is existent. The P_d and P_f of K/N rule are analyzed according to single frame. Applying K/N rule, the detecting probability and the false alarm probability can meet the demands of the system if choosing the value of N and K properly.