Attention-enhanced RepBi-FPN YOLOv8 method for reservoir type identification in electrical imaging

At present, imaging logging image recognition mainly focuses on fracture recognition, while the research on the recognition of pore-like reservoir types is relatively lacking, and there has not been a systematic recognition method for layer-segment reservoir types. To this end, this paper proposes a reservoir type recognition method based on computer vision and deep learning, adopting YOLOv8-S as the model framework, and improving the recognition prediction ability of image reservoir types by introducing the attention mechanism and optimizing the Neck module. In addition, a sliding window method combined with edge detection is proposed to segment the layer segment images, which effectively reduces the loss of feature information and improves the accuracy of model recognition. Compared with the existing identification methods, this experimental research method significantly reduces the subjectivity and workload of manual operation, and is able to identify both fracture and hole reservoir types, with an identification accuracy of 83.3%, which has a certain reservoir type identification and prediction capability, providing a technical guarantee for the accurate evaluation of fracture and hole reservoirs