| Image recognition are often easy to identify with the human eyes, but is not able to identify bycalculation with computer. This paper focuses on how to use the extension theory to analyze thecontradictions， and find a solution to deal with these problem with extension transform method.The purpose is to explore the image recognition methods to improve the accuracy of imagerecognition from a new perspective, and change can not be identified to can identify.This paper studies the method of the color histogram extension, proposed a colorhistogram-based image recognition method using the extenics divergence analysis and associationanalysis theory, by using the extension theory in the color histogram scaling transformation, makethe image can be recognized, the experiment shows that this method can greatly improve therecognition accuracy than not expand.Then, the paper points out the contradictions caused by orientation of the features, because thedescriptor of features calculated according to the orientation, if the orientation changes thedescriptor changes too. To cope with this, this paper proposed a novel method E-SURF, which baseon replacement principle and replacement transformation method. The proposed method add aextended orientation, if the descriptor calculated according to the primary orientation can’t findmatch pair, then use the extended orientation to get a new descriptor, after that we can use the newdescriptor to find the match pair. Experiments show that this orientation to expand by increasingthe feature points matching significantly increased to improve the possibility of identifying theimage.Finally, analysis the feasibility parallelization of various parts of the computing complexE-SURF algorithm, break down some of the tasks for multi-threaded processing to maximize theadvantages of mainstream multi-core computer, experimental results show that, through the use ofmulti-threaded calculation, greatly improving the efficiency of the implementation of thealgorithm. |
PDF Full Text Request