Search

With the development of computer vision and high-precision 3D model reconstruction, used for the measurement and 3D reconstruction of the geological landslides, acquiring a high-precision relative orientation basing multiple images is crucial and the key point to ensuring and improving the accuracy of 3D model and space position. Currently, the conventional relative orientation model includes five independent parameters. For the linear relative orientation model, there are nine parameters to construct the linear space geometric relationship between the imaging and space point. To eliminate the impact of more parameterization and improve the accuracy and stability of solved parameters for the conventional direct relative orientation model, a new relative orientation model with seven constraints is proposed and validated in this paper. The additional constraints are derived from the orthogonal property of the rotation matrix of a stereo imaging pair and associated with the least squares adjustment to obtain a high-precision result of the relative orientation. Through the accuracy assessment using space position, it is revealed that the new proposed model is more advantage for the conventional type of direct relative orientation, especially at 3D model reconstruction and close range photogrammetric and applications for the geological landslides measurement. El modelo de orientación relativa directa con siete restricciones para la medida de deslizamientos de tierra y reconstrucción tridimensional ResumenCon el desarrollo del entorno computacional y la alta precisión del modelo de reconstrucción tridimensional, utilizados para la medida y reconstrucción de desprendimientos geológicos, es crucial la obtención de la orientación relativa de alta precisión basada en imágenes múltiples y es el punto clave para asegurar y mejorar la exactitud del modelo 3D y la posición espacial. Actualmente el modelo de orientación relativa incluye cinco parámetros independientes. En el modelo linear de orientación relativa hay nueve parámetros para construir la relación geométrica espacial linear entre el sondeo y la posición espacial. Para eliminar el impacto de más parametrización y mejorar la exactitud y la estabilidad de los parámetros resueltos el modelo de orientación relativa convencional, este artículo propone y valida un nuevo modelo de orientación relativa con siete restricciones. Las restricciones adicionales se derivan de la propiedad ortogonal de la matriz de rotación de la imagen estéreo y se asocian con el ajuste de los cuadrados mínimos para obtener un resultado de alta precisión de la orientación relativa. Al medir la exactitud con la posición espacial se revela que el nuevo modelo propuesto tiene más ventajas que aquel de orientación relativa directa, especialmente en el modelo de reconstrucción 3D y en las aplicaciones fotográmetricas de rango cercano para la evaluación de desprendimientos geológicos.

In recent years, space stereoscopic imaging system structured with two or more lens has been a lot of promotion and application. Especially in the three-dimensional measurement and spatial model reconstruction, it receives more and more extensive attention. Theoretically, 3D reconstruction from multi-image can be comparable with the laser scanner in accuracy. How exactly solving the spatial relationship between the spatial overlap image sequence, is an important part to ensure and improve the accuracy of the reconstructioual 3D model. This paper introduces an algorithm for generating point cloud based on the space relative orientation direct solver and additional constraints. To ensure the accuracy and efficiency of relative orientation, firstly, the rough values are obtained by direct relative orientation, then the values is refined by iterative solution in the error equation with additional constraints. The algorithm has higher precision solver on spatial relationship parameters between overlapping image sequence, can greatly improve the accuracy of the subsequent 3D point cloud, and ensure the accuracy of the 3D reconstruction.

High-accuracy image matching algorithms are a crucial step of 3D reconstruction from multiple images, which can be used to create the accurate 3D models of spatial entities. SURF, a conventional image matching algorithm, has the disadvantages of sparse matching, high incorrect rate. Thus, SURF matching algorithms are hard to meet the demand of high-accuracy 3D reconstruction. This paper introduces the MSURF, a new image matching algorithm which combines homography and SURF algorithm. Experiments are performed by comparing MSURF with conventional matching algorithms from the aspects of quantity and quality, and the experimental results verifies that MSURF can not only increase the quantity of match points, but also improve the accuracy of match points. In addition, the paper analyzes the experimental result, and acquired the application features of MSURF. Thereby, the new proposed matching algorithm, MSURF, extends and improves the existing matching algorithm, and ensures the high-accuracy 3D reconstruction.