การใช้วิศวกรรมข้อความร่วมกับปัญญาประดิษฐ์เชิงสร้างสรรค์เพื่อการวิเคราะห์ข้อมูลภาพดาวเทียมเซนทิเนล-2 บนแพลตฟอร์มกูเกิลเอิร์ธเอนจิน
Year: 2026
Type: article
Abstract: This study aims to evaluate the potential of prompt engineering combined with generative AI to translate natural language into JavaScript code for Google Earth Engine (GEE), enabling land use classification of Sentinel-2 imagery in Mueang Khon Kaen District using a Support Vector Machine (SVM) classifier. The results demonstrated that code generated through prompt engineering and Generative AI could be effectively utilized for processing large-scale geospatial data on GEE, achieving 87 % overall accuracy and a Kappa coefficient of 0.84. The effectiveness of the results depended on two primary factors: (1) the quality of training data for the SVM classifier, and (2) the quality of natural language prompts used with Generative AI. Additionally, the user's geospatial expertise proved essential for optimizing command design, selecting appropriate indices, sampling code examples, and interpreting results. This research demonstrates that Generative AI combined with prompt engineering can serve as a valuable tool for supporting code generation for spatial data analysis on GEE, reducing technical barriers and expanding access to spatial data analysis technology for general users through effective natural language interaction.
Language: Thai
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7215276979