多摄像头侧视钻孔全景图像拼接  

Panoramic image mosaic by multiple side-view cameras in the borehole

作　　者：张炳文[1] 曹茂永[2] 张维[1] 滕升华[1] 赵猛[2] 

机构地区：[1]山东科技大学电子通信与物理学院,山东青岛266590 [2]山东科技大学电气与自动化工程学院,山东青岛266590

出　　处：《广西大学学报（自然科学版）》2017年第3期1092-1098,共7页Journal of Guangxi University（Natural Science Edition）

基　　金："十二五"国家科技支撑计划项目(2012BAB13B04);国际(中国南非)科技合作项目(CS06-L02)

摘　　要：目前,地质勘测主要采用数字式全景钻孔摄像获取具有真实感的孔壁全景图以进行地质分析,但是已有的全景图获取方法存在设备成本高或图像拼接过程复杂等不足,基此提出了一种水平环状均匀分布多个CCD摄像头的摄像模型,对钻孔孔壁四周360°进行拍摄获取图像,并对获取的图像进行畸变矫正。对钻孔孔壁的曲率造成的图像信息缺失,采用非均匀插值对其进行展开。通过提取图像SIFT特征点,利用基于(k-d)树的BBF算法确定图像的匹配点对,采用RANSAC算法二次消除错配求得变换矩阵。加权平均法在图像融合阶段的应用使得拼接图像光滑无缝,得到具有真实感的钻孔孔壁全景图。At present,geological survey mainly adopts the digital panoramic borehole camera to obtain realistic panorama image of hole wall in geological analysis,but the existing methods of getting panorama image have shortcomings,such as high cost of equipment,complex image stitching process,etc. A camera model is presented, which is made of horizontal circular uniformly distributed multiple CCD camera to capture the borehole wall around 360 degrees and corrects the distorted images. The image information which is lost caused by the curvature of the borehole wall is recovered by non-uniform interpolation. Through extracting the image SIFT feature point, thematching points are determined based on( k-d) tree BBF( best bin first) algorithm,and then the false matching points are removed and the transformation matrix are got by using RANSAC( random sample consensus) algorithm. In the image fusion stage,the weighted average method is used to obtain a smoothly seamless mosaic image,and finally the realistic pictures of the borehole wall are obtained.

关 键 词：钻孔孔壁全景图 畸变矫正 非均匀插值 拼接图像 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]