Study On Related Technologies Of Target Object Recognition Based On ASIFT Algorithem

Machine vision is one of the important parts of the robot system,and an important area of intelligent robot.The vision system of a robot makes it have the autonomous visual perception function.To get the two-dimensional images of the environment by a camera,and then analysis and translate through a visual processor,that make a robot possible to recognize the object and determine its position.Supported by the Hi-tech Research and Development Program of China,in this paper,we do some research on the visual recognition system of the satellite assembly robot arm.An image feature recognition algorithm base on ASIFT is proposed and optimized to make it possible to recognize the object more rapidly and more accurately,surmounting strong nonlinear difficulties like illumination,occlusion,clutter,and image affine transformation.The main research contents and innovative results are as follows:1.The research is based on technology of the image feature matching.The ASIFT algorithm is analyzed and compared with other mainstream image feature recognition algorithms.It is proved that in the same condition,the algorithm of recognition of target image has outstanding advantages such as affine invariance and scale invariance.But at the same time,because of the complexity,the real-time performance is poor.2.In order to solve the problem of the practical application of ASIFT algorithm,the PCA theory is used to accelerate the algorithm.It reduces the 128 dimensional feature descriptor to 36 and the use of multi thread parallel acceleration technology in this algorithms greatly improve the possibility of application in real time system.In the process of removing the feature points of the error matching,ORSA algorithm is used to balance the residual value and the number of false alarm,retaining more correct matching points.3.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 feasibility and accuracy of the proposed algorithm are verified through comparison and analysis of the estimated parameters and the actual parameters in evaluating indicator of root mean square error.