abstract |
The invention discloses a QEMSCAN-based shale MAPS mineral identification method, comprising: S1, performing multi-stage, cross-resolution pixel-level alignment processing on the QEMSCAN picture and the MAPS picture; S2, constructing the fine-grained mineral category in the QEMSCAN picture and the The mapping relationship between the coarse-grained mineral identification tags in MAPS images, based on the aligned QEMSCAN images and MAPS images, to construct a supervised training data set; S3, based on the constructed supervised training data set to train a mineral identification model for MAPS mineral identification, Learn the mapping relationship between MAPS pictures and mineral categories; S4, input the MAPS pictures to be identified into the trained mineral identification model to obtain mineral category identification results. The method of the present invention does not require additional experimental steps, except that special equipment is required to obtain the QEMSCAN image and BSE image of the sample during model training, and the mineral composition of the corresponding position can be directly predicted with a separate MAPS image, which saves a lot of costs. With time, the identification effect can meet the identification needs of production scenarios. |