A new procedure to automatically delineate joints and lineations from digital imagery is presented in this paper. The spatial merging procedure is applied to the results of three different edge detectors on a digital aerial image. As a measure of accuracy, the automatically delineated structures, as well as the raw results of the edge detectors are compared to the lineations manually identified on the image. None of the techniques is highly successful in delineating joints and lineations on its own. Spatial merging, however, has the potential for becoming a useful automated technique in the future.
The identification and mapping of rock joints and other geologic lineations are necessary steps for both characterizing a rock mass and for input into mechanical and flow models. Mechanical modeling that includes a complete structural characterization of the rock mass can be used to accurately analyze the stability of slopes, underground openings, and foundations. Similarly, flow modeling can be used to more accurately track potential contamination from nuclear and toxic waste sites and to help locate productive groundwater sources. The mapping and measurement of linear geologic features and their characteristics in the field, however, are labor intensive and costly. There have been many suggestions over the past few years about using automated structural identification based on image analysis to increase the speed and volume at which mapping could be accomplished. A number of image processing procedures and algorithms have been proposed in the literature (Xu, et al., 1981; Moore and Waltz, 1983; Schowengerdt, 1983; Thiessen, et al., 1994; among others), but the results of these techniques are mixed. Most of these efforts have failed to detect all lineations present in the image, partly because the procedures are directional whereas lineations are omni-directional, but also because of confusion with linear cultural patterns, e.g. tree lines, fences, roads, and airfields.
A new procedure to automatically delineate joints and lineations from digital imagery is described in this paper. This technique, dubbed spatial merging, operates on a bilevel image, which is an image containing two graylevels, black and white. A bilevel image is first produced from the original 256 gray level images using an edge detection filter or other quantitative algorithm. The groupings of black (or white) pixels in the bilevel image are analyzed for shape and orientation and connected to each other based upon variable search parameters. These connected groupings represent the fractures and lineafions in the image, and are labeled for future analysis, such as trend and length, and identification.
Six comparisons are made with a digital image upon which the structural features were identified by a geologist with extensive image analysis experience. These features were drawn onto the image using ARC/INFO. The results of each method of edge detection before and after application of the spatial merging procedure was compared with this image. Comparisons are made of the number of partly detected features, fully detected features and imaginary detected features.
The digital image is of an area approximately 600m on a side in the Granite Mountains in central Wyoming. These mountains consist of flat to rounded bedrock surfaces with very well exposed joints. This image was chosen for its simplicity; it consists primarily of bare rock and the dark-toned joints stand out in marked contrast to the white granite surfaces. There is a small stream channel in the upper left corner of the digital image, one narrow dirt track in the same area, and some sparse vegetation, which includes grasses, sage