Linear target change detection from a single image based on three-dimensional real scene

Change detection is a critical component in the field of remote sensing, with significant implications for resource management and land monitoring. Currently, most conventional methods for remote sensing change detection often rely on qualitative monitoring, which usually requires data collection from the entire scene over multiple time periods. In this paper, we propose a method that can be computationally intensive and lacks reusability, especially when dealing with large datasets. We use a novel methodology that leverages the texture features and geometric structure information derived from three-dimensional (3D) real scenes. By establishing a two-dimensional (2D)-3D geometric relationship between a single observational image and the corresponding 3D scene, we can obtain more accurate positional information for the image. This relationship allows us to transfer the depth information from the 3D model to the observational image, thereby facilitating precise geometric change measurements for specific planar targets. Experimental results indicate that our approach enables millimetre-level change detection of minuscule targets based on a single image. Compared with conventional methods, our technique offers enhanced efficiency and reusability, making it a valuable tool for the fine-grained change detection of small targets based on 3D real scene. By using the texture features and geometric structural information of the 3D model, we aim to establish a precise 2D-3D geometric mapping relationship between a single observation image and the corresponding 3D model based on the same scene. This mapping allows us to extract accurate positional data from the single image. Additionally, this geometric relationship effectively transfers depth information from the 3D model to the observation image, enabling us to detect changes in planar line targets during this period.image