Research On Image Feature Matching For Mobile Applications

In recent years,with the development of electronic integration technology,the computing and storage capacity for mobile devices has a leap of improvement,which makes the image feature matching based of the image analysis has been widely and deeply studied.At present,many scholars have put forward suitable feature matching methods for mobile applications,and made some constructive researchs.However,current research still has several clearly insufficient key steps: Image point feature representation is not accurate,especially in the smoothing area of image,which has less number of feature points and the direction estimation is less characteristic,resulting in matching features difficultly;the feature matching are vulnerable to similar features descriptors interference in the merit-based selection,resulting in mismatching characteristic points.To solve these problems,the main research work of this paper is as follows:1.Propose a gradient operator based on fractional differential,improve the accuracy of image feature representation.For the mobile terminal,which is easy to cause blurred image and make image resolution not high,the paper studies the theory of fractional order differential,and uses it for estimating gradient operator to apply to the mobile terminal detection operator ORB,BRISK and FREAK characteristics,improving the accuracy of feature points;at the same time,by introducing the pyramid image to improve the ORB operator to be better adapted to the change of the image size.The experimental results show that the improved algorithm can effectively improve the number of feature points with an average increase of about15%.2.Present a graph matching method based on structure preserving,improve the image feature matching accuracy.Traditional feature matching method uses only the distance constraint to find the optimal matching point,when the feature is inaccurate,the error of matching rate is high.Therefore,this paper uses two layer matching strategy: firstly identify strong match point set by the distance constraint,and then introduce shape constraint methods,from both the angle and distance constraints to realize matching weak feature points.By comparing the experimental results,we can find that the proposed algorithm can greatly improve feature matching accuracy with an average increase of about 50%.3.Achieve a suitable image feature matching method for mobile applications on iOS-based system.Designed on iOS system architecture and MVC design patterns,fractional differential operators and graph matching are used on ORB,BRISK and FREAK algorithm and tested on i Phone 6 real machine.Reality shot through the viewfinder,the image feature matching accuracy rate is less than 90%.