无人机的POS修正以及影像拼接研究

发布时间：2019-01-04 13:25

【摘要】：由于无人机有诸多优点,如重量轻,体积小,运输方便,成图周期短,时效性,成本低,并且能到达人员无法进入的危险区域,因而受到很多领域的青睐,尤其在大比例尺成图作业中更具优势。无人机不仅是卫星遥感的一种升华,更是大地测量手段的一种补充。随着无人机在测绘中的普遍应用,其影像快速有效的拼接及DEM生产越来越重要。但无人机也有缺陷,如受气流影响大、姿态不稳定、旋片角大、POS精度不高、像幅小致使照片数量多等问题,导致影像配准缓慢,甚至失败。尤其是在抗震救灾以及急需成图供野外调绘时,满足不了影像的速拼及各种附加产品生产的要求。基于无人机影像的特点,结合在生产中遇到的两个主要问题—POS精度不高以及像片的重叠度或者姿态角不能满足传统空中三角测量要求的情况,本论文主要针对以上两个问题做一个初探,并提出解决办法。针对第一个问题—因POS问题导致匹配失败,本文提出一种给定并修正POS线元素初值的办法。先是用SIFT算子对影像进行特征提取,而后基于欧式距离匹配,再依据同名点对应的模型点坐标大致相等的原理给出影像POS线元素的初值,之后通过连续法相对定向对初值修正,从而计算出新的POS线元素。再次应用于空三加密软件中,以辅助完成匹配及其后续工作。另外针对第二个问题,在无人机序列影像不满足三度以上重叠以及小角度等情况下,按照传统空三流程无法完成相对定向及影像拼接时,本文将对其基于计算机视觉的序列影像的拼接,给出一种利用SURF算子检测特征点,利用SIFT算子提取描述符,基于BBF匹配,利用RANSAC剔除误匹配,分别使用相似变换、仿射变换、射影变换建立几何变换模型,经试验发现射影变换的精度最高,更适合无人机影像的拼接,并利用LM局部优化算法对模型参数进行优化,无需POS和相机检校数据,只要满足有重叠的图像(大角度或者重叠度达不到摄影测量生产要求的)均可实现快速匹配以及拼接。最后可利用POS数据对拼接后的大影像进行简单的地理配准,再经过重采样后,影像便具有大致的坐标信息,对于应急灾害以及外业调绘指导性工作均有深远的意义。
[Abstract]:Because UAVs have many advantages, such as light weight, small volume, convenient transportation, short mapping period, low cost, and can reach dangerous areas that can not be accessed by people, they are favored in many fields. Especially in large scale mapping operations have more advantages. UAV is not only a sublimation of satellite remote sensing, but also a supplement to geodesy. With the widespread application of UAV in surveying and mapping, the rapid and effective stitching and DEM production of UAV images are becoming more and more important. However, UAVs also have some defects, such as large airflow, unstable attitude, large rotation angle, low precision of POS, small image, large number of photos and so on, which lead to slow or even failure of image registration. Especially in the earthquake relief and the urgent need of mapping for field painting, it can not meet the requirements of the rapid combination of images and the production of various additional products. Based on the characteristics of UAV image, combined with the two main problems encountered in production, the POS accuracy is not high and the overlap or attitude angle of the image can not meet the requirements of traditional aerial triangulation. This paper mainly aims at the above two problems to do a preliminary study, and put forward solutions. In view of the first problem, the matching failure caused by the POS problem, this paper proposes a method to set and correct the initial value of the POS line element. Firstly, the feature extraction of image is carried out by SIFT operator, then based on Euclidean distance matching, the initial value of image POS line element is given according to the principle that the coordinate of model point corresponding to the same name point is approximately equal, and then the initial value is corrected by relative orientation of continuous method. The new POS line element is calculated. It is again applied to the empty triple encryption software to assist in the matching and subsequent work. In addition, for the second problem, when the UAV sequence images do not meet three degrees of overlap and small angles, according to the traditional space three flow can not complete the relative orientation and image stitching, In this paper, we propose a method to detect feature points by using SURF operator, extract descriptor by SIFT operator, eliminate mismatch by RANSAC, and use similar transform and affine transform respectively. The projective transformation is used to establish the geometric transformation model. It is found that the projective transformation is the most accurate and is more suitable for the joint of UAV images. The parameters of the model are optimized by using the LM local optimization algorithm without the need of POS and camera calibration data. Fast matching and stitching can be achieved as long as the overlapped images (large angle or overlapped degree can not meet the requirements of photogrammetric production). Finally, the POS data can be used for the simple geographical registration of the large mosaic images, and then after resampling, the images will have approximate coordinate information, which is of far-reaching significance for emergency disasters and field mapping guidance work.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P237

【相似文献】