Research On Template Matching Algorithm Based On Translation,Rotation And Scale Change Image

Template matching algorithm is a way of image registration.The target is found in the scene image according to the given reference template by this algorithm.It has been widely used in many fields such as medical image analysis,remote sensing information processing,and industrial manufacture,and has become an indispensable technology in the development of industrial automation.In this paper,the template matching algorithm is studied,and the algorithm solution is studied when the scene target has translation,rotation,scale,or the above three conditions.Firstly,The template matching algorithm of translation invariance based on normalized cross-correlation and bounded partial correlation is analyzed.The effect of the initialization performance parameters of the bounded partial correlation algorithm on the matching efficiency of the algorithm is verified through experiments.The selection strategy of the initialization performance parameters is determined,and the lightness resistance of the algorithm is verified.On this basis,a two-stage dual-check bounded partial correlation template matching algorithm of translation invariance is proposed.The algorithm introduces a coarse-fine two-stage search strategy and dynamic sampling technology to solve the inefficient matching of the basic bounded partial correlation algorithm.Real-time template matching is realized.Secondly,a rotation-invariant template matching algorithm based on the circle projection technique of bounded partial correlation is proposed.This algorithm uses a frequency domain method to replace the look-up table method to extract the features of the circle projection,which greatly improves the efficiency of feature extraction and realizes the rotation-invariant template matching algorithm against arbitrary rotation angles.Experimental tests show that the algorithm can resist noise,brightness and contrast changes within a certain range,and the matching efficiency is 1.5 to 3 times that of the moment invariance algorithm.Then,a rotation scale-invariant template matching algorithm based on SIFT features is proposed.The threshold method and RANSAC algorithm are used to filter the correct matching point pairs to achieve more accurate matching.Experimental results show that the matching position error of the algorithm is within 1.5 pixels,the matching angle error is within 2 degrees,and the scale error is within 0.01.The algorithm not only can resist rotation and scale changes,but also maintain a certain degree of anti-interference to noise and lightness.Finally,the TDBPCcs algorithm is applied to the visual application of the circular optical lens automatic panning,and the matching position error is less than 1 pixel and the positioning error is less than 0.03 mm.The practicability of the proposed algorithm is proved by the visual application of the circular optical lens automatic panning.