Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Binocular Vision Based Pavement Flatness Detection
Leveling detection belongs to an important part of pavement inspection, and its detection technology has been developing. In this paper, we propose to detect pavement flatness based on binocular vision, and obtain the point cloud data of the pavement through the study of the calibration, correction and matching algorithm of binocular camera. In the feature extraction of the pavement point cloud data, the self-attention mechanism module is introduced to capture the relationship between different points. By calculating the similarity between individual points, the model can focus on those points that are most helpful for feature extraction of the current point during feature extraction. For the collected pavement point cloud data firstly, voxel filtering is utilized for noise reduction and thinning, which preserves the geometric features of the overall point cloud. Then, the RANSAC algorithm is used to fit the point cloud to the plane. The height difference information between the fitted plane and the pavement point cloud is used. This information is used to calculate the pavement flatness. The experimental results of the levelness detection are based on binocular vision. The results are 0.086 and 0.06 higher than the two traditional detection methods. The two traditional methods are the precision level and the bump accumulator. This method reduces manual operation. It improves the working efficiency in the measurement work.
Binocular Vision Based Pavement Flatness Detection
Leveling detection belongs to an important part of pavement inspection, and its detection technology has been developing. In this paper, we propose to detect pavement flatness based on binocular vision, and obtain the point cloud data of the pavement through the study of the calibration, correction and matching algorithm of binocular camera. In the feature extraction of the pavement point cloud data, the self-attention mechanism module is introduced to capture the relationship between different points. By calculating the similarity between individual points, the model can focus on those points that are most helpful for feature extraction of the current point during feature extraction. For the collected pavement point cloud data firstly, voxel filtering is utilized for noise reduction and thinning, which preserves the geometric features of the overall point cloud. Then, the RANSAC algorithm is used to fit the point cloud to the plane. The height difference information between the fitted plane and the pavement point cloud is used. This information is used to calculate the pavement flatness. The experimental results of the levelness detection are based on binocular vision. The results are 0.086 and 0.06 higher than the two traditional detection methods. The two traditional methods are the precision level and the bump accumulator. This method reduces manual operation. It improves the working efficiency in the measurement work.
Binocular Vision Based Pavement Flatness Detection
Wang, Shuaipeng (Autor:in) / Yu, Chaogang (Autor:in) / Cao, Zhen (Autor:in) / Jin, Shengjie (Autor:in)
24.11.2024
470723 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch