3D Reconstruction Algorithm Based On Image Sequence

People use the computer to achieve the visual function, usually called computer vision. People study in the field of computer is more and more advanced and mature, also makes it in the 3D world in the field of desire. More people in the world through 3D complete two-dimensional image recognition.Using the technique of 3D reconstruction of image sequences, the 3D model of the low cost, vivid, easy operation and reconstruction, has gradually become a hot research topic in the field of computer vision. Image sequences, based algorithm research and improvement of two images and multiple images in 3D reconstruction of the.Firstly, this paper uses an improved 2DPCA-SIFT features a large amount of data matching algorithm based on PCA-SIFT matching algorithm, and time-consuming problem. The gradient vector block improved 2DPCA-SIFT feature reduction. The method is compared with the PCA-SIFT, the space can be solved quickly, reduce the computational complexity, do not need more storage space. An improved method based improved 2DPCA-SIFT feature matching algorithm can eliminate the correlation between the image rows and columns, and the method is better than the PCA-SIFT algorithm has more accurate matching rate. The experimental results show that the improved 2DPCA-SIFT algorithm is stable, accurate and rapid details more prominent the reconstruction results can recover the 3D model.Secondly, as is evident in the 3D reconstruction of a characteristic point of the object in the context of the model improved the algorithm of 2DPCA-SIFT.But the characteristics of remote sensing image is not particularly evident in the image sequence, improved 2DPCA-SIFT reconstruction algorithm accuracy not high, 3D model realistic degree is not strong recovery. So due to the characteristics of remote sensing image is not obvious, the details of the reconstruction of irregular, multiple object could not be recovered and according to the recovery of sparse image sequences fusion effect and texture information is a problem difficult to solve. Therefore, in order to accurately restore the detail information of the remote sensing image. This paper uses a matching algorithm combining growth and CMVS-PMVS dense areas.Experimental results show that based on region growing CMVS-PMVS remote sensing image reconstruction of 3D reconstruction algorithm of point cloud density is large enough, the realistic reconstruction of the target object is extremely strong, can be fully restored the details of that objects in remote sensing image reconstruction with very strong practicality, and through the design of the clustering algorithm can effectively reduce the data. The improved algorithm can well eliminate the false matching points, the optimization speed can also speed up the reconstruction results.