Feature Point Detection Of Concrete Image Based On Self–similarity And Its Application

Concrete is commonly used in the construction of large buildings.The topography and feature characterization of concrete buildings are of great importance to the safety monitoring.To describe the three-dimensional topography and features of a building accurately,it is necessary to obtain a complete image.For larger concrete buildings,the method bases on image processing can be used to obtain the complete image.The process of obtaining a complete image by image processing generally includes image acquisition,image preprocessing,image registration,and image fusion.And the effective detection of feature points is the key of image registration.Due to the low contrast and little texture of some concrete images,the usual feature point detection algorithm is easy to fail.So,this paper focuses on the following tasks:1.A scheme for feature detection of large buildings and the related technologies of image mosaic are introduced.The local images of the building surface are collected by using the imaging device for image mosaic.2.The characteristics of image self-similarity is studied,and the image entropy is applied to select feature points,so that a self-similarity feature point detection algorithm based on local image entropy is designed.The results of experiment show that the algorithm has better robustness in detecting the feature point of concrete images.3.The method of image down-sampling is adopted,and the correlation window of feature points is rotated,which makes the improved method have some rotation and scale invariance compared with the traditional correlation matching method.At the same time,the scaling factor is introduced to improve the SURF feature descriptor,which improves the scale invariance of self-similarity feature point matching.4.Based on the results of feature point detection and matching,the overlapping area of the registration image is selected,and the weighted average method is used for image fusion.The results of experiment show that the method can realize seamless mosaic of concrete images.