Abstract:The available pattern recognition techniques is to pick up attributes based on poststack data,in which the information of attributes changing with offsets lost because of stack.For that reason,we presented using prestack seismic data to pick up various attributes,then,computing changing gradient or average value of attributes relative to offsets in all traces of each CDP gather,and using cluster analysis and classification judgment to carry out pattern recognition for picked seismic attributes;we also presented that pickup of attributes is first carried out for the targets of traces near the well,then,using cluster and logging data to specify the classification of samples.It is shown by practical data tests that the method can recognize the change of reservoir in both vertical and lateral directions.