Parallelization On Feature Extraction Algorithm Based On Heterogeneous Environment

About80%information which obtained in humans with the "five senses" from the visual. For humans, visual information is easy to understand and trust. The three-dimensional reconstruction is the focus and emphasis of computer vision technology, it is aimed at using the two-dimensional image reconstruction of three-dimensional scene. The Key of the three-dimensional reconstruction techniques is to obtain two-dimensional image which reflected in the collection of points of the surface characteristics and different image points of the surface features to match (ie, feature extraction and matching), and the establishment of real objects computer vision points data collection eventually. Therefore, the efficiency of the implementation of feature extraction has a great impact on the completion time of the three-dimensional reconstruction.This article extracted from the following aspects of the feature points in the three-dimensional reconstruction heterogeneous environment parallelization speed up the research:Harris operator and Dog Difference of Gaussian analysis of the classic feature point extraction, explore parallel processing processes and procedures. Combination of the harriers operator and Dog Difference of Gaussian parallel attempt, the experimental results in a heterogeneous environment with the ideal speedup.The above results combined with KLT feature points tracking algorithm for three-dimensional reconstruction of the image sequence in the three-dimensional reconstruction of the proposed heterogeneous environment.