The purpose of this study is to construct a computer-aided analysis framework for exploring the expression of natural concepts in the paintings of artists in southeast Guizhou under the guidance of ecological aesthetics. This method combines multi-modal image coding, graph structural relationship analysis and semantic mapping to quantify visual structures that rely on qualitative interpretation in the past. This study establishes a dataset of 1248 digitized works by 37 representative painters, and extracts relevant features around brush stroke, texture, composition and ecological imagery. Experimental results show that the proposed framework achieves 91.6% accuracy in natural concept recognition and 88.4% consistency in pen-ink expression measurement, and there is a stable coupling relationship between natural concept and pen-ink variables. The comparative analysis further finds that there are clear differences between different groups of artists in terms of ink density, line rhythm, white space organization and composition opening and closing. This study provides a computational path for interpreting regional painting language under repeatable conditions, and also provides a stable analysis reference for artistic image computing and digital humanities research.
