CLASSIFICATION OF RICKER WAVELETS AND THE DETECTION OF BRIGHT SPOTS USING A TREE CALSSIFIER.

The Tree classification method is proposed for the classification of Ricker wavelets and the detection of bright spots. The Envelope and instantaneous frequency are used as features. From the analysis of zero-phase Ricker wavelet, a tree classification technique is adopted. The analytic signal analysis of Gaussian noise as the ground roll motion provides references for the tree classifier design. Envelope is selected as the first-level feature and instantaneous frequency is selected as the second-level feature in the tree classification. For every input seismic trace and seismogram, the data are not known as bright spots or non-bright spots. The advantages of the tree classifiers are that they are easy to design and computationally efficient.