Research On The Fast Algorithm Of Digital Image Matching Based On Surf And Gray Projection

Digital image matching is a key technology in the field of digital imageprocessing. With the development of computer technology and digital imageprocessing technology, image matching technology plays an important role innumerous fields. Although there are many image matching algorithms at the presentstage, not an algorithm can fit every situation. The aim of image matching is to findan algorithm has faster speed and higher rate. For this reason, the research of the newalgorithm and improved algorithm is significant.In the first of this dissertation, history and basic concept of image matchingtechnique、 application fields are briefly introduced, several existing classicalalgorithms is analyzed, and traditional gray-based algorithms with feature-basedalgorithms is summarized. Then matching algorithm based on speed-up robustfeatures(SUFR) is studied. SURF shows its advantages in rotation and illuminationchanges, but SURF algorithm has a high probability of error matching points inpractical application. Therefore we propose a method for improving matchingprobability. First, a coarse points matching was performed by the nearest neighbormethod. Then error matching points were eliminated using the method of curve fitting.And finally the position of the components was found in the circuit board through thecoordinated relation of the matching points, using the optimized invariant momentfeatures method for product testing. The experimental results showed theeffectiveness of the proposed method. This method improves the accuracy of thefeature points matching, and realizes the accurately positioning and detection ofcircuit board components. Then Image matching based on pixel gray correlation isstudied in this dissertation, it combined with sequential similarity detecting algorithm,and used image pyramid from course to fine searching strategy. The experimentalresults proved that the proposed algorithm was faster than the conventional imagetemplate matching algorithm, and has practical application value. In this paper ofChapter IV, the image matching based on DSP is studied, the image matchingtechnology based on DSP is mainly divided into two parts: the hardware design andsoftware design, in the end of parts, the matching algorithm is implemented on DSP.In the end of this dissertation, a summary of the whole dissertation is given, and future research topics are suggested.