Key points are not available for this paper at this time.
The detection and monitoring of key points on a retaining-wall surface using a vision camera is crucial for measuring displacement. Researchers have previously proposed adding several artificial makers to a retaining wall to monitor displacement. However, the use of artificial markers presents several disadvantages. Additionally, vision-based displacement monitoring under field conditions are typically affected by problems such as illumination variation and strong wind-induced nonstationary cameras. Therefore, this study proposes a natural-target-based feature-matching method to extract natural features from retaining walls, thus enabling displacement calculation without artificial targets. Histogram equalization and homography algorithms are used to enhance the performance of the feature-matching algorithms and improve the accuracy of displacement calculation. An image dataset from a monitoring camera installed on an actual retaining wall at Seobu Mountain, Busan, Korea, is used to evaluate the performance of natural-feature extraction and displacement calculation. The results show that the feature-matching algorithm performed well in extracting and matching the natural features of the retaining wall compared with artificial markers. Additionally, the calculated horizontal displacement, settlement, and facing displacement of the actual retaining wall based on the single-camera and natural-feature-based methods indicate a maximum overall error of 3.8 mm.
Song et al. (Sun,) studied this question.