13 February 2023 in Geophysics
The mapping of seismic facies from seismic data is considered a multiclass image semantic segmentation problem. Despite the signification progress made by the deep learning methods in seismic prospecting, the dense prediction problem of seismic facies requires large amounts of annotated seismic facies data, which often are unavailable. These valuable labels are only helpful in one model and field due to geologic heterogeneity. To overcome these challenges, we have developed a few-shot seismic facies segmentation model. Few-shot learning has been designed to learn to perform with very few labels and we design reconstructing masked traces as a pretext task for self-supervised learning to obtain a good feature extractor. By these, this model can use all seismic data from different fields, which is different from image data as the texture-based data. With two different seismic data in turn as a meta-training set and a meta-testing set, our model works well in one- and five-shot settings, which means only one label and five labels, respectively.
19 January 2024 in IEEE Transactions on Geoscience and Remote Sensing
Automatic bedding extraction from borehole images of shale beddings and cross beddings is essential for reducing labor-intensive manual tasks. Three main types of algorithms, including Autodip, Random Hough Transform (RHT), and Pairing Algorithm Based Detection (PABD), have been proposed in the past decades but have never been directly compared. In this study, we first propose a novel method integrating a heuristic genetic algorithm (Autodip integrate with Genetic Algorithm, ADGA) to overcome the low efficiency of the Autodip algorithm. This method improves the efficiency by a factor of 3, and even by a factor of 21 with the addition of automatic constraints, while maintaining the advantages of high accuracy. Furthermore, to verify the effectiveness of the improved algorithm, all extraction algorithms are compared and analyzed systematically for the first time. The results demonstrate that three algorithms have presented their unique advantages and adaptability. The ADGA has the best stability, extraction accuracy, and image fitness for shale beddings. The PBAD algorithm shows its superiority in cross-bedding. The RHT has the best efficiency with moderate stability. In addition, the extraction results produced by RHT are of poor quality for images with high density and low amplitude bedding (less than about 10 pixels) due to the inadequate performance of the structure tensor image processing. This research will significantly promote the future development of automatic bedding extraction algorithms for industrial applications.
1 June 2022 in Journal of Physics: Conference Series
When the vehicle is only affected by road conditions and its own speed, in the low temperature rain and snow weather. In order to clarify the order of the importance of each influencing factor at the exit section of the tunnel on the driving stability, this article uses the analytic hierarchy process to sort them. The results show that the order of importance is: Friction coefficient μs
> driving speed ν (m/s)> circular curve radius R (m), longitudinal slope i (%). According to the sorting results, targeted safeguard measures can be provided for the tunnel design, operation and use process.