Research On Key Technology Of Photogrammetry Methods In Engineering Applications

It is essential to monitor and measure the construction of buildings periodically for safety in engineering projects. The characteristics of photogrammetry, like no-contact, low cost, high efficiency and high accuracy make it very popular in engineering surveys. But it is still lack of study on photogrammetry conforming to the engineering characteristics, and nowadays researchers just copy the workflow and methods of aerial photogrammetry for engineering survey. Apparently, it limits the range of the application. Therefore, research on the theory and methods of engineering application photogrammetry is the eager requirement for the interdisciplinary development of photogrammetry and engineering survey. Considering the great demand of photogrammetry in engineering application and the limit of the traditional aerial photogrammetry, this paper focuses on solving the problem from2D image to3D shape, which improves the practicability of the traditional photogrammetry methods in engineering application. The main content in this paper includes:1) By investigating and analyzing the existing matching vector descriptors and matching measures, a summary and conclusion about them has been made, it is a fundamental in image matching to describe points and measure the similarity between them. By reviewing lots of existing approaches, we introduce several representative methods of matching vector descriptors and matching measures both from gray based and feature based images matching.2) By investigating the matching optimization problem of photogrammetry used in engineering application, a multiview matching optimization method considering the epipolar error has been proposed. In engineering application, considering the cost and flexibility, non-metric cameras like ordinary digital camera are usually used. Influenced by the camera distortion and spatial control point errors, epipolar errors between stereo images always still exist after aerotriangulation processing. So the images in engineering applications are very different from the aerial images or epipolar images obtained under laboratory condition. Noticing that, a multiview matching optimization method considering the epipolar error has been proposed. A theoretical derivation is made first. and then practical optimization procedure is introduced. At last our method is proved with practical engineering data and compared with the existing methods. Experiments show that the matching optimization method proposed by us can fit the engineering application well. So it will be very suitable for engineering applications, since it is not insensitive to the epipolar error and owns high accuracy.3) By investigating the matching strategy problem of photogrammetry used in engineering application, a hierarchical matching strategy based on region growing from the horizontal and vertical two directions has been proposed. Limited by the circumstance conditions, the stereo images in engineering application are always with more geometric deformations, more parallax shifts and more matching disturbance, which make the matching processing more difficult. Matching optimization focuses on solving the precision problem. While matching strategy focuses on taking account of all possible constraints, including the already existed or those generated during the matching process, and making full use of them to predict the next matching, in order to improve the rate of correct matching. Combining the traditional pyramid matching strategy with the region growing method, a new hierarchical matching strategy with region growing from the horizontal and vertical two directions has been proposed. Then the matching reliability measure, the hierarchical propagation flow and the horizontal and vertical propagation are also discussed in detail. Experiments show that such matching strategy is reasonable and effective, since the generated point cloud can describe the topography well.4) In some special cases of cylindrical buildings, the surface texture is too poor to get a good matching result, and a novel measurement method based on silhouette has been proposed. For the textureless objects, traditional procedure of finding corresponding points and forward intersection is impossible. By exploiting the model information and the silhouette constraint, this paper proposed a novel measurement method based on silhouette for cylindrical buildings. Firstly the related work of cylindrical building measurement is discussed and then the image station (IS) used in our paper is introduced. After that the principle of the projection model from the revolving solid to the silhouette is derived in detail. Based on the method, a practical iteration procedure is given. Lastly, the whole workflow is expressed according to the practical application. Experiments show that the proposed measurement method proposed can transform the laborious outdoor work into comfortable indoor work. Thus it is more efficient and economical compared with the traditional survey method. Moreover, it can realize a successive measurement on the cross sections of buildings, therefore it is more suitable for the periodical monitoring of the cylindrical buildings.