Graphical Recognition of Landscape in Luang Prabang Province Using AI-Based Technology

This paper aims to identify differences in the landscape of the historic preservation area (PSMV) and surrounding area (Buffer zone) through objective manipulation. This study utilizes machine learning technology and semantic segmentation analysis. As a first step, we collected 820 landscape images of two areas for use in machine learning and trained a machine-learning model with 700 pieces of teacher data. The remaining 120 test data will then be used in an experiment to determine if there are any objective differences between the two areas. In the second step, semantic segmentation analysis using deep learning is performed using 120 test data that were used in machine learning to analyze what segments on the landscape image are different between the two areas. The above results show that the correct response rate by machine learning in the first stage is as high as 79%. Next, the semantic segmentation analysis conducted in the second stage found 1% significant differences in the segment of fence, pole, vegetation, terrain, sky, person, and car. Thus, this study found that there are specific differences in the landscape images in the two areas of Luang Prabang through objective manipulation. This research shows the effectiveness of using machine learning and deep learning in recognizing and analyzing segments in the buffer zone and PSMV area. The study also served as a foundation for the future which will be useful in implementing AI technic in various fields. It also contributes to the study of landscape policy in Luang Prabang province.