Abstract:Similitude velocity analysis based on stack or correlation principleresults in very good velocity estimations of separate seismic events,but unsatisfactory velocity estimations of both the adjacent coherent seismic events and the seismic events which have Iow SJN ratio or reverse polarities due to AVO phenomenon. In recentyears,the covariance measurement used usually in radar and sonar techniques are applied to seismic velocity analysis. People have found that this method is more accurate than the similitude rnethod,and that both of them are based on stacking rnode1 in velocity analysis. The covariance measurement theory is used in AVO rnodeling,where AVO factor is introduced and covariance coefficient is proved to be only related to the squared AVO factor,so that the inverse-polarity AVO phenomenon is removed. To reduce desirably the influence of AVO factor,normalized correlation matrix is used in velocity analysis instead of covariance rnatrix,so that not only thecoherent and low S/N seismic events show higher resolution but also the velocity spectrum remains accurate in the case of inverse-polarity AVO phenomenon. Modelink and real data processing show that the accurate velocity analysis will find wider wider application than stack velocity analysis.