On-line Ships Detecting Based on Probability Velocity Field
基于概率速度场的实时船舶检测

In the paper, an on|line ship detecting system based on the probability velocity field is presented. First all environment factors that ships go through the ship lock are analyzed, such as lighting, shading, waving and so on; Second, a Gaussian probability distribution model(probability field) is generated from gray|level histogram to diminish the affection of shades and speckles, since shades and speckles' gray|levels are distributed in both sides of gray|level histogram, after histogram is fitted with the Gaussian distribution function, the probability of them is smaller than that of other gray|levels. Therefore a probability velocity field is defined and derived based on the optical flow to detect ships in lock on|line. The velocity field is calculated by the traditional optical flow estimation method using the probability of gray|levels, not the gray|levels. Finally the affection of shading and lighting is reduced greatly. The method has succeeded in testing a larger number of image sequences that ships pass through the temporary ship lock of the three gorges project(TGP) and ship lock of the Gezhou dam.