The Application Of Swarm Intelligence Algorithm In Image SIFT Feature Matching

Image matching is the process of finding the same or close position in another image based on one image.Not only the characteristics of the image but also the performance of the matching algorithm need to be considered.With the development of the field of computer vision and imaging,more and more requirements have been put forward,mainly focusing on how to improve the speed or accuracy of matching and the stability of the algorithm.Generally,the focus will be on proposing new methods for measuring the similarity of features or The matching point pair is optimized during the search process.Image maps are classified from image information extraction based on gray-scale image matching [1,2] and feature-based image matching.The image matching based on the gray value has many pixels and a single layer.Although the principle is simple,the calculation is large.Feature-based image matching algorithms,such as SUFT algorithm [3],Harris algorithm [4],SIFT algorithm [5],etc,due to the sharp decrease in the number of feature points compared to the grayscale image matching,there will be a great speed The promotion.Moreover,the famous American scholar Mikolajczyk [6] compared the performance of these several image features in the experiment and concluded that the SIFT features have strong robustness,regardless of whether the image is a light conversion,a scale transformation or a spatial transformation.In the past decade or so,domestic and foreign researchers have used hundreds of methods to match images,but they can't reach a balanced state in matching speed,matching accuracy,versatility,and stability.This paper focuses on the matching search strategy of SIFT features and combines swarm intelligence optimization algorithms to improve the matching efficiency of images.The working ideas are as follows:(1)In order to avoid the complicated computational problems brought by the high dimension of SIFT features,principal component analysis(PCA)and kernel projection method(walsh-hadamard)are used to reduce the initial SIFT features.(2)Introduce swarm intelligence algorithm to optimize the search strategy in the matchingprocess.When the number of feature points is large,the matching method of exhaustive method takes more time,so the matching efficiency of feature points needs to be improved from the search method,so the swarm intelligence algorithm is introduced.The swarm intelligence algorithm has highly structured management and strong ability of collaborative work,and is widely used for target optimization.This paper mainly uses two kinds of algorithms,namely,the particle swarm algorithm and the ant colony algorithm.Finally,it elaborates in detail.The basic idea and implementation process of the improved algorithm are obtained.Finally,the conclusion is drawn from the experimental results.