倾斜航空影像数据处理粗差探测方法研究

发布时间：2018-09-12 14:51

【摘要】：倾斜航空摄影技术是国际测绘领域近年发展十分迅速的一项高新技术，利用该技术获取的倾斜影像可为三维模型重建提供丰富的纹理信息。倾斜航空摄影后期数据处理中采用多视影像匹配技术自动化获取海量匹配点数据，，但因倾斜影像的摄影比例尺不一致、分辨率差异明显、地物遮挡严重等特点导致获取的数据中含有较多的粗差，严重影响后续的多视影像空三加密精度。因此，探测倾斜航空摄影海量数据中的多维粗差显得尤其重要。目前现有的粗差检测算法受其探测能力或处理效率的限制，无法准确、高效的探测海量数据中的多维粗差；又由于倾斜航空影像的姿态角类型多样，致使影像间的相对位置关系复杂，传统的连续相对定向模型不再适用。针对上述问题，本文分别从粗差检测算法和平差数学模型两方面进行研究解决。 1）本文基于相关分析粗差探测原则，提出了一种适合处理摄影测量领域海量数据的粗差检测算法——摄影测量中的多维粗差同时定位和定值法，简称LEGEP法。通过模拟粗差实验证明，LEGEP法能够准确地定位海量数据中的多维粗差，同时求得各个粗差的数值大小；将LEGEP法与其他几种具有代表性的粗差检测算法进行对比实验发现，LEGEP法能够利用更少的迭代计算量探测出更多的粗差，显著提高了平差精度，从而证明了该算法在探测能力和效率两方面的优越性。 2）本文提出采用直接解相对定向模型作为平差数学模型的方法。该模型的求解不需要任何参数的真值的近似值，即无需倾斜影像的姿态角初值，是一种适合倾斜航空影像匹配数据探测粗差的通用平差模型。实验证明，本文提出的基于直接解相对定向模型，结合应用LEGEP法的粗差探测方法，能够有效地探测倾斜航空摄影海量数据中的多维粗差，切实地提高观测数据的精度，在倾斜航空摄影实际数据处理中具有较高的应用价值。
[Abstract]:Tilt aerial photography is a new and high technology developed rapidly in the field of international surveying and mapping in recent years. The tilt image obtained by this technique can provide rich texture information for 3D model reconstruction. In the later data processing of tilted aerial photography, the multi-view image matching technique is used to automatically acquire the massive matching point data. However, because of the inconsistency of the photographic scale of inclined aerial photography, the resolution difference is obvious. Because of the serious feature of object occlusion, there are many gross errors in the acquired data, which seriously affect the accuracy of space triple encryption in the subsequent multi-view images. Therefore, it is very important to detect the multi-dimensional gross error in the massive data of tilted aerial photography. At present, the existing gross error detection algorithms are limited by their detection ability or processing efficiency, so they can not accurately and efficiently detect the multi-dimensional gross error in massive data, and because of the variety of attitude angle types of inclined aerial images, Because of the complexity of the relative position relationship between images, the traditional continuous relative orientation model is no longer applicable. Aiming at the above problems, this paper studies and solves the problem from two aspects: gross error detection algorithm and mathematical model of error. 1) based on correlation analysis, gross error detection principle is used in this paper. This paper presents a gross error detection algorithm suitable for dealing with massive data in photogrammetry field, which is the simultaneous location and determination of multi-dimensional gross error in photogrammetry, which is referred to as LEGEP method. The simulation results show that the LGEP method can accurately locate the multi-dimensional gross error in the massive data and obtain the numerical value of each gross error at the same time. By comparing the LEGEP method with other typical gross error detection algorithms, it is found that the LEGEP method can detect more gross errors with less iterative computation, and the adjustment accuracy is improved significantly. Therefore, the superiority of the algorithm in detecting ability and efficiency is proved. 2) in this paper, a direct solution relative orientation model is proposed as a mathematical model of adjustment. The solution of the model does not require the approximate value of the true value of any parameter, that is, the initial value of the attitude angle of the tilt image, so it is a general adjustment model suitable for detecting gross error of the tilted aerial image matching data. Experimental results show that based on the direct solution relative orientation model and combined with the gross error detection method of LEGEP method, the multi-dimensional gross error can be effectively detected in large amount of inclined aerial photography data, and the accuracy of observation data can be improved effectively. It has high application value in the practical data processing of tilted aerial photography.
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P231

【参考文献】