一种基于改进ORB算法的显微图像实时拼接方法

A real-time stitching method for microscopic images based on improved ORB algorithm

工业显微镜与图像拼接技术的结合被广泛应用于微观检测领域,通过创建高分辨率、大场景的微观图像,可以对半导体材料或电子器件进行大范围的表面质量评估和缺陷检测。然而在拼接过程中,由于显微图像分辨率高、特征信息丰富以及存在大量相似区域,导致图像配准准确率低,拼接时间过长。针对这种情况,提出了一种基于改进ORB算法的显微图像实时拼接方法,以实现快速配准和融合。首先通过相位相关法获得重叠区域的初始估计,然后通过改进的FAST算法对重叠区域图像进行特征点检测,并以BEBLID描述子替换BRIEF描述子进行特征描述,之后采用GMS算法对特征点进行粗匹配,并采用改进的RANSAC算法进行精细匹配并计算变换矩阵完成图像配准,最后通过非线性权重融合完成图像拼接。实验结果表明,本文提出的方法在具有良好的实时性的同时也能保持较高水平的准确率,且在图像的不同变化场景中也有良好的稳健性,可以满足显微图像实时拼接的要求。

The combination of industrial microscopy and image stitching technology is widely used in the field of micro-inspection,where the creation of high-resolution,large-scene micro-images allows for a wide range of surface quality assessment and defect detection of semiconductor materials or electronic devices.However,in the splicing process,the high resolution of microscopic images,rich feature information and the existence of a large number of similar regions lead to low image alignment accuracy and long splicing time.To address this situation,this paper proposes a real-time splicing method for microscopic images based on the improved ORB algorithm to achieve fast alignment and fusion.Firstly,the initial estimation of the overlapping region is obtained by phase correlation method,then the feature points are detected in the overlapping region image by the improved FAST algorithm,and the BRIEF descriptor is replaced by the BEBLID descriptor for the characterization,after that,the GMS algorithm is used for the coarse matching of the feature points and the improved RANSAC algorithm is used for the fine matching and the calculation of the transformation matrix to complete the image alignment,and finally,the nonlinear weight fusion method is proposed to achieve the rapid alignment and fusion.Finally,the image stitching is completed by nonlinear weight fusion.The experimental results show that the method proposed in this paper can maintain a high level of accuracy while having good real-time performance,and has good robustness in different changing scenarios of images,which can meet the requirements of real-time splicing of microscopic images.