Evaluation of Stereo Matching Algorithms for Temporary Structure Monitoring
Youyi Feng, Fei Dai · Computing in Civil and Building Engineering (2014) · 2014
Temporary structures, such as formwork support and scaffolding, play an important role in on-site production for construction projects. However, they are normally disregarded when considering the importance of qualifications and experience. In fact, most of the time, temporary structures won't be considered until it is actually constructed. As a result, most tragic failures during construction are usually the result of improperly designed, constructed and/or maintained temporary structures. A real-time monitoring system enabled by video cameras may facilitate the monitoring of integrity and safety of construction temporary structures. It entails applying close-range photogrammetry and digital image correlation to measure the deformation of structural material shapes. In this system, stereo image matching algorithms and accuracies are considered to be one of the most significant elements. It is to seek the corresponding relationships between left and right-view images of the same scene to obtain the disparity images. Thus, this paper attempts to evaluate the image matching algorithms that exist in order to identify the most appropriate one for the temporary structure monitoring system. First, the principle of stereo matching is introduced. This is followed by the comparison of three classical algorithms, semi-global matching (SGM), block matching (BM), and graph cuts (GC) method. The undertaking of the comparison includes statistical analyses of the time and disparity images quality estimation. The research concludes that the SGM algorithm possesses both relative high efficiency and disparity images with excellent qualities. The present work facilitates the investigation of the aforementioned monitoring system.