Reliability of Contrast-Based Automated Fracture Detection from Decimeter Resolution Aerial Imagery

Gianluca Amicarelli, Mark Ireland, Colin Davie

Aerial imagery that captures outcrop exposures of rocks enables the characterization of structural discontinuities such as faults and fractures across large and difficult-to-access areas. These datasets provide an opportunity to analyze the characteristics of fault and fracture networks as analogues for the subsurface. The application of automated interpretation methods to imagery has the potential to reduce subjectivity, increase efficiency, and improve reproducibility. While these methods are commonly used with high-resolution images, the widespread availability of decimeter-resolution imagery, means that demonstrating the reliability of automated methods with decimeter resolution data could significantly extend their application. This study investigates the reliability of two automated fracture detection methods using decimeter resolution imagery and specifically the methods’ sensitivity to contrast between rock mass and fractures.
The results indicate that: 1) at lower contrast thresholds, the average feature length increases while the number of feature intersections decreases; 2) extracted features demonstrate reduced robustness in discontinuous outcrop exposures. It is crucial to note that these findings presume a scenario where the rock mass exhibits high contrast, and fractures are characterized by low contrast. The findings demonstrate that automated feature detection is reproducible but currently produces very unreliable results. The use of these methods requires a detailed understanding of how image properties and the parameters defined in detection method impact the resultant features detected. The use of automated methods with decimeter resolution imagery requires a clearly defined criteria to quality check the results.