Effective detection of seismic events by non-classical receptive field visual cognitive modelling

The detection and up-picking pf the seismic events are critical for seismic dat analysis and interpretation. Events picking can be used for sequence stratigraphic analysi reservoir feature extraction, the determining of the subsequent reflection interface, th improving of the SNR and the storage prediction. The research of the events picking i very significant for the seismic exploration. In order to overcome the existing event picking methods have the same sensitivity to noise. we propose a non-classical receptiv field visual cognitive method for the events picking up. Vision is an important functiona organ for human beings to obseve and recognize the world. About 80% of th information obtained by human beings from the outside world comes from the visua system. which fully shows that visual information is huge. and also shows that huma beings have a high utilization rate of visual information. How to transfer some typica information processing mechanism and target recognition function of human vision t machine is one of the most important and urgent tasks in the field of computer vision an artificial intelligence. The introduction of computer vision technology into geophysica prospecting is still in its infancy in the field of seismic exploration, our research fill th blank of this field, where the use of visual features to improve the seismic data processin and rapid realization of oil and gas exploration, will become the vane of the futur direction of research and development.
As a basic research work in the crossing field. this paper has made a breakthrough i the research methods and ideas, and the research content can be summarized as th following four aspects:1. The proposed method implements the function of environmental suppression an spatial enhancement of the bio-visual primary visual cortex. which is applies to th pre-stack seismic data, as pre-stack seismic data contains abundant information such a amplitude and frequency to reflect tiny structures of the formation.
2. The seismic data is preprocessed to obtain the wavelet fusion of the envelope peal instantaneous frequency (EPIF) and the slant stack peak amplitude (SSPA), which can maximum the limit to provide optimal quality data.
3. An adaptive Gabor filter direction selection method is proposed to provide a reliabl angle range and improve the recognition rate of filter decomposition. In addition. b adopting an anisotropic environmental suppression method, our method can detect edg variability more accurately than the isotropic method.
4. With the enhanced contour aggregation. cocircular constraint is adopted and combine with the characteristics of low curvature and continuous changing curvature, which i unique to the seismic events. to establish a consistent spatial structure perception model The events picked by our method is more continuous and accurate than the existin methods. and doesn't require human interaction, which is beneficial for subsequent seismic interpretation and reservoir prediction.