Data Mining Large Volumes of Seismic Data for Oil Information – 3D Orientation Histogram
a report by
Maria Petrou
Professor of Signal Processing, Imperial College London, and Director, Informatics and Telematics Institute, Centre for Research and Technology – Hellas
Oil exploration is mainly based on large-scale geological surveys that yield huge amounts of data. The data collected by such surveys are organised as 3D arrays. A typical campaign may produce data that cover a volume of the Earth’s crust measuring 300x300x2km3, with 25x25m2 resolution parallel to the sea surface. The manual analysis of such data may take up to three years. It is therefore a major advantage to develop methodologies that will direct the attention of the expert to the most interesting parts of the data and accelerate their analysis.
There have been many attempts to automate the analysis of seismic data. Earlier studies concentrated on the analysis of volume data on a slice-by-slice basis. However, the 3D aspect of the data is not fully exploited by such methods. It is only recently that effort has been placed on the analysis of 3D data. Initially, 1D and 2D approaches were extended to 3D. The outcome of such studies has been to produce the so-called attribute cubes. Each voxel position in an attribute cube represents some property at or in the region around the corresponding voxel in the data under analysis.
We Are 3D-blind – The Trouble with Seeing Only Surfaces
Despite these efforts, the fully automatic analysis of seismic data remains elusive. Nevertheless, significant progress has been made and there are now many software tools available to the expert; some of the most impressive are concerned with data visualisation. The reason is very simple: humans cannot see in volumes, they can only see surfaces. If one cannot see along the third dimension, it is even more difficult to perceive a fourth dimension that corresponds to some variation of some data attribute.
So, although cave technology and clever semi-transparent projections allow the expert to ‘see through’ the data, allowing him/her to ‘see’ a fourth dimension, along with something other than the brightness or locality of the data changes, is beyond such technologies. Tools that make explicit the implicit data characteristics and present them in a way that the human visual system can cope with are required. Below, I will show how the 3D orientation histogram can help make explicit the fourth dimension in the data, namely that of data change.
Texture, Texture Everywhere, and Not a Universal Method to Use
One of the most striking characteristics of seismic data is their rich texture. Texture is a spatial property, as opposed to a point property. In other words, although we can characterise a particular point in space by the amplitude of the signal it returns, if we want to characterise the same point by its texture, we must consider several points around it too, i.e. we must consider a patch in space around it. Texture features may be used to discriminate different data regions, if we manage to capture with a few numbers the essence of the pattern represented by each texture. There are several methods that can be used to characterise
© TOUCH BRIEFINGS 2010
texture in both 2D and 3D data.1
For large-scale applications, we have
to use a texture characteristic that is mathematically simple, can be computed efficiently and quickly and captures the important characteristics of the textures we wish to discriminate.
If Only the World Were Black and White
Often, volume attribute values are used as features to segment the data. However, image segmentation forces each segment to belong to a particular texture class. As textures often are not fully distinct, a far better approach is to mine the data rather than segment them. Data mining is different from data segmentation as, instead of splitting the data into distinct classes, data mining answers the question: ‘Which parts of the data are compatible with this texture class?’ In theory, any of the attributes extracted may be used for data mining. However, when data are mined, it is very important to use attributes that capture as much as possible the structure and texture of the sought region that distinguish it from the other regions. So, not all features may be suitable for the job.
Know the Problem
Let us consider a slice of a seismic volume of data, as shown in Figure 1. We can see that, from the geological point of view, the textures we wish to discriminate are those that correspond to the horizon regions, the porous regions and the regions that are disrupted possibly by rising gas, on the right of this image. What captures the eye immediately in seeing this slice is the extreme anisotropy of the data at the stratified region: the brightness of the data varies very slightly along each horizon, but it varies significantly orthogonally to the horizons. In the chaotic region, the brightness of the data varies more or less in the same way along each direction. It is this property that mostly allows us to distinguish between these two types of texture. The 3D local orientation histogram is a rich texture descriptor that captures aspects of local data anisotropy, which is a distinct feature for seismic textures.
The 3D Orientation Histogram
Imagine that at each point inside the seismic volume we compute a vector that points in the direction of maximum data change. For
Maria Petrou is a Professor of Signal Processing at Imperial College London and Director of the Informatics and Telematics Institute of the Centre for Research and Technology – Hellas. She is a Fellow of the Royal Academy of Engineering, the Institution of Engineering and Technology and the Institute of Physics, a Senior Member of the Institute of Electrical & Electronics Engineers, Inc. and a Distinguished Fellow of the British Machine Vision Association. Professor Petrou studied physics at the
Aristotle University of Thessaloniki and applied mathematics in Cambridge, completed her PhD in astronomy in Cambridge and obtained her DSc in engineering from Cambridge.
E:
maria.petrou@imperial.ac.uk
139
ICT
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132 |
Page 133 |
Page 134 |
Page 135 |
Page 136 |
Page 137 |
Page 138 |
Page 139 |
Page 140 |
Page 141 |
Page 142 |
Page 143 |
Page 144 |
Page 145 |
Page 146 |
Page 147 |
Page 148