A novel feature fusion-based stratum image recognition method for drilling rig

Stratum identification is critical in numerous fields related to earth sciences, such as geotechnical engineering, mining engineering. This study presents and evaluates a feature-fusion image recognition model of stratum for rotary drilling rig foundation building, with the aim to identify stratum rapidly and accurately during real-time drilling of drilling rigs. To begin, the color domain illumination estimation method is employed in conjunction with the Bradford chromaticity adaption conversion formula to rectify the color bias of stratum images induced by complicated illumination in the field. Then, using the modified watershed method and color statistical features, an image concrete feature extractor is built, and the feature recursion algorithm selects the ideal feature combination. The image's abstract features are then extracted using a lightweight convolutional neural network. Finally, the depth fusion of concrete and abstract features is used to construct a stratum recognition model. The results reveal that the color correction method aids in the resolution of color bias produced by complicated illumination and that the accuracy of the recognition model trained using the corrected image features improves by 4% on average. The suggested recognition model achieves 91% accuracy with very high information quality in the fused features. The method has high efficiency and accuracy and is expected to contribute to the engineering applications of stratum identification.