Method for calculating porosity in tight sandstone reservoir thin sections based on ICSO intelligent algorithm

Surface porosity is crucial for evaluating tight sandstone reservoirs' performance and resource potential. The current manual calculation and algorithm extraction methods have problems such as heavy workload, long time consumption, low accuracy in identifying complex pore morphologies, and weak learning ability for sparse samples. Drawing on the concept of hybrid intelligence, this paper proposes an intelligent calculation method for the surface porosity of tight sandstone reservoirs (ICSO) that combines the SOLOv2 algorithm and OpenCV. The SOLOv2 instance segmentation algorithm was used to segment and label pore regions in images. OpenCV was employed to extract pore distribution and proportions, thereby realizing the calculation of surface porosity. The performance comparison with similar algorithms demonstrates the advantages of this method in terms of accuracy, running speed, and generalization ability. It addresses the surface porosity calculation issue and provides a novel research approach for solving similar problems in related fields. The Graphical Abstract consists of three parts: establishment of a thin section dataset, identification of pore components, and calculation of thin section porosity. ICSO involves establishing a cast thin section image dataset through preprocessing, applying the instance segmentation SOLOv2 algorithm to identify pore regions in the images, and utilizing OpenCV to extract pore distribution and proportion for thin section porosity calculation. • The ICSO intelligent algorithm enables the identification of pores in tight sandstone reservoirs and the calculation of the surface-porosity ratio. • Image preprocessing and sample annotation augmentation improve the quality of cast thin-section images and solve the problem of small samples of thin-sections. • The Multi-scale Feature Enhancement and the parallel dual-attention mechanism can enhance SOLOv2's focus on the pore regions and improve its ability to extract pore features. • The adaptive pixel extraction mechanism based on OpenCV can achieve the display of pore distribution and the calculation of the porosity rate.