| The logging of core lithology is significant for the exploration of sandstone-type uranium deposit.Currently,manual logging is still the main means to obtain geological information such as lithology,which has the drawbacks of being time-consuming and laborious,strong subjectivity and incomplete information.As a new direction for obtaning and mining core geoscience information,imaging spectral scanning has the advantage of non-destructive detection,recording both spectral and textural data objectively which can reflect the characteristics of mineral composition,structure,rock type,etc.This article studied the lithology identification and core logging technology with the hyperspectral image of borehole ZKH3 searching for sandstone-type uranium deposit in the southwestern Ordos Basin.The fihished efforts included core imaging spectral scanning,data processing,color recognition,indicative mineral extraction,classification of rocks with different particle size,and construction of a comprehensive rock naming and automatic recognition scheme based on three elements.And the digitization and automation of rock naming and recognition of sandstone-type uranium deposit has been preliminarily achieved.The main achievements and understanding of this paper are as follows:(1)An automatic color recognition and nomination method has been developed to objectively express the rock color.Through experimental test and comparison,the full spectrum method was found more effective in line with reality than typical RGB band method and can meet the needs of core geological logging.(2)The spectral matching and statistical methods were used to realize the automatic extraction and digital logging of seven indicative minerals in ZKH3,and the formation and transformation environment of rocks in the borehole area was also analyzed.It was considered that the occurrence and association of dolomite,chlorite and sericite in red variegated continental clasts may indicate that the rocks were reworked by acidic hydrothermal fluids.(3)The imaging spectal based deep learning method was established to identify the grain size and logging lithology of clastic rocks in sandstone uranium deposit.The deep learning used1 DCNN model,attention mechanism and shortcut technology and reached an overall accuracy of94.6% in the validation set which is better than SVM and can well identify mudstone,siltstone,fine sandstone,medium sandstone,gritstone,glutenite and background.Moreover,the lithological logging procedure was designed and developed to optimize the outcome of grain size identification and meet the need of geological logging.(4)A comprehensive lithology coding method were preliminaryly designed and programmed on the identification result of color,grain size and indicative mineral,which can comprehensively identify and nominate the types of sedimentary rocks in sandstone uranium deposit exploration.The recognition results are more precise than the original geological logging data,and can directly reflect various information such as color,mineral composition,and grain size,which is helpful for subsequent research on alteration,oxidation-reduction zoning,mineralization fluid characteristics,and paleoclimate inversion.Main innovation points in this dissertation are as follows:(1)The imaging spectrometry was first used to study the color recognition and coding in the field of uranium exploration.And an innovative color optimization algorithm was proposed and programmed by comparing the relationship between the current segment and adjacent segments,which can effectively filter out the abnormal colors and enhance the continuity of color blocks.(2)The principle of selecting sample sets was established for the deep learning of lithological identification with core imaging spectra,which means that the selected samples should be distributed at the beginning or end of per attack and cover the whole borehole depth evenly.In addition,a filtering algorithm of abnormal sample was developed by comparing the reflectance relationship between the detected band and the adjacent bands,.(3)An optimization method and nomination scheme were designed and programmed to reach the concordance among three elements of lithology in sandstone-type uranium deposit,which can achieve automatic nomination of rocks with color,indicative minerals and grain size.In summary,this dissertation made accomplishment on the method study of lithology recognition for sandstone-type uranium deposit with imaging spectral data,and developed the programs for batch processing of scanning data and algorithm of automatiuc lithology identification,which were successfully used in borehole ZKH3 and lay a good foundation for engineering applications. |
PDF Full Text Request