Research On Rapid Workpiece Recognition Based On SIFT Algorithm

With more and more intelligent and anthropomorphic robot,machine vision system is becoming more and more important in the whole system.The robot vision system is approximately two-dimensional image acquisition environment through the camera,analyzed by visual elements,the overall deployment of the system in the machine,and then achieve with independent visual robot.This paper is supported by the national high technology research and development program(863 Program)support,mainly for the visual recognition system of satellite assembly manipulator is studied,using image feature recognition algorithm based on SIFT and improved,which can exist in the external assembly process interference case has strong robustness,can more quickly and more accurately identify the target object.The main contents and innovations of this paper are as follows:First of all,this paper is mainly related to technology of image feature recognition and feature matching based on feature matching technology to summarize and research at present,in order to find out and provide a theoretical basis for improved SIFT algorithm;image pretreatment experiment,find out the image pre-processing method is suitable for this,on the back of the algorithm to do the preparatory experiment.SIFT algorithm and PCA-SIFT algorithm,SURF algorithm for theoretical analysis,summed up the advantages and disadvantages of the original SIFT,reference PCA-SIFT and SURF algorithm theory analysis,to find out the direction of this improved algorithm.Secondly,In order to solve the high dimension of feature points SIFT algorithm and euclidean distance search point sets the problem of low efficiency,the use of LPP(local preserving projection theory)is improved based on the SIFT algorithm,a substantial reduction of 128 dimensional image features of the original SIFT operator;and using the improved KD tree(random KD tree)algorithm two feature points on the formation of SIFT algorithm in parallel comparison fast,greatly improve the operating speed of the algorithm.Finally,To build up the experimental platform on Visual Studio 2012 combination of OpenCV2.4.8,analysis of the proposed algorithm is carried out.It has been proved that the algorithm has good performance and efficiency.The external disturbance model is established to estimate the target transformation parameters and compared with the actual parameters,and the root mean square error is used as the evaluation index of the matching accuracy of the feature points,which verifies the feasibility and accuracy of the algorithm.