<p>Understanding the historical evolution of spatial layout and architectural styles in a historic area is imperative for its preservation and rejuvenation. However, conventional methods such as document collation and field surveys are time- and resource-intensive. This research focuses on the historic area of Dujiangyan city (Dujiangyan Old Town) and uses computer vision techniques to increase the efficiency and accuracy of architectural information capture. Semantic segmentation techniques are employed to derive building footprints from remote sensing satellite images captured from three points in time over a period of nearly five decades, facilitating the comparison of spatial patterns and the identification of different architectural styles in panoramic street views. The distribution of architectural styles is analysed from urban planning and architectural perspectives. The experimental findings illustrate well-preserved spatial patterns and discernible development strategies across various periods. Image-processing methods have emerged as effective tools for analysing urban spatial dynamics and identifying architectural styles, thereby yielding quantifiable data essential for obtaining a nuanced understanding of historic areas. These insights can contribute to planning, management, and preservation efforts aimed at historic areas.</p>