Algorithm for improving drill core sizing accuracy using image segmentation
摘要
Currently, image processing methods based on segmentation do not achieve sufficient accuracy to reliably capture the actual dimensions of drill core samples, limiting their ability to provide precise quantitative information on the extracted material’s true length. Although core trays are standardized in 50 cm segments, the core samples themselves often vary in size. The measurement of the Length of Intact Rock Core Pieces (LIRCP) is still carried out using empirical tools such as measuring tapes or rulers, and in some cases, unreliable algorithms, leading to operational delays and increased costs despite LIRCP being a key parameter for calculating the Rock Quality Designation (RQD). Furthermore, the core cutting process and LIRCP measurement are typically performed in separate locations. This project aimed to design and implement an integrated system capable of accurately calculating both the actual length and volume of core samples. The methodology was based on the YOLOv11 segmentation model, chosen for its high detection accuracy and speed, from which length and volume were estimated through 2D pixel analysis. This process was supported by experimental validation and statistical error analysis. The system achieved 98% accuracy in estimating both parameters. This study successfully developed and implemented an integrated system capable of accurately calculating the actual length and volume of drill core samples, overcoming the limitations of traditional image processing and manual methods. This innovation significantly improves operational efficiency by reducing time and costs, while providing precise real time data to support informed geological decision making.