Detection of truncations on satellite images and seismic slices: application to channels and point bars.

Leonore Gallot and Nicolas Clausolles and Guillaume Caumon. ( 2019 )
in: 2019 Ring Meeting, ASGA

Abstract

Detecting channels and the associated sedimentary architectural elements on a picture can be useful for understanding the evolution of channels over time and to model the heterogeneity of alluvial deposits. This work proposes two methods to automatically identify depositional gaps by extracting truncation patterns on satellite images. The first approach adapts a seismic attribute (the unconformity likelihood) to detect where the point bars, the channels, the abandonedloops and the meander scars intersect. This attribute relies on the comparison of the variations of local orientations on the image. The variations of local orientations are estimated with the structure tensor and its eigen decomposition. We smooth twice the structure tensor to remove the background noise and enhance the unconformities: horizontally by a structure-oriented smoothing and then vertically by a gradient-oriented smoothing. Finally, a script extracts the edges of the detected unconformities thanks to an hysteresis thresholding. The second method uses a brute-force search of unconformities by splitting the neighborhood of each pixel in two parts separated by a line of variable orientation. We search the orientation which maximizes the difference between the structure tensors computed in each part. Then we calculate an anisotropy attribute with a discretized smoothing perpendicular to the variable orientation. The results of this comparison show that structures are detected more precisely with the second method.

Download / Links

BibTeX Reference

@inproceedings{GallotRM2019,
 abstract = { Detecting channels and the associated sedimentary architectural elements on a picture can be useful for understanding the evolution of channels over time and to model the heterogeneity of alluvial deposits. This work proposes two methods to automatically identify depositional gaps by extracting truncation patterns on satellite images. The first approach adapts a seismic attribute (the unconformity likelihood) to detect where the point bars, the channels, the abandonedloops and the meander scars intersect. This attribute relies on the comparison of the variations of local orientations on the image. The variations of local orientations are estimated with the structure tensor and its eigen decomposition. We smooth twice the structure tensor to remove the background noise and enhance the unconformities: horizontally by a structure-oriented smoothing and then vertically by a gradient-oriented smoothing. Finally, a script extracts the edges of the detected unconformities thanks to an hysteresis thresholding. The second method uses a brute-force search of unconformities by splitting the neighborhood of each pixel in two parts separated by a line of variable orientation. We search the orientation which maximizes the difference between the structure tensors computed in each part. Then we calculate an anisotropy attribute with a discretized smoothing perpendicular to the variable orientation. The results of this comparison show that structures are detected more precisely with the second method. },
 author = { Gallot, Leonore AND Clausolles, Nicolas AND Caumon, Guillaume },
 booktitle = { 2019 Ring Meeting },
 publisher = { ASGA },
 title = { Detection of truncations on satellite images and seismic slices: application to channels and point bars. },
 year = { 2019 }
}