Theory And Application For Second Order Motion Perception

Computer vision plays more and more important roles in the domain of industry and military,. Compared with present computer visual system, human visual system could percept second order motion defined by the characteristics except for lightness, and with excellent ability in computing motion. The main research of this thesis is as follow:Based on Reichardt correlation model, we model first order motion system, here Gabor function is used as the spatial filters due to its multilple center frequencies and orientations, and Gamma function is used as the temporal filters. In every spatial frequency-orientation channel, make decomposition signals with same spatial location but different phases multiply with each other's temporal delay, the difference of them indicates the motion strength of the signal, where the value's signal represent direction. In every spatial location, take the vector addition of all the orientations in the same spatial frequency channel as the output, then combine outputs of all the channels with the principle of Winner-Take-All, so the motion strength in this location will be gained.Second order motion system is composed of two parts: texture grabber and motion energy (ME) , the texture grabber consists of three stages: an isotropic linear spatial filter using difference of two Gausse functions, a temporal bandpass filter constructed by difference of two Gamma functions and full-wave rectification, ME is the first order motion system.Research the second order sitmulus in the natural scene both in spatial and frequency domain, and sort them due to their frequency spectrum. Detect second order stimulus analysed in this thesis with second motion system. From the experiment results ,we could know the model is effective.