基于无人飞艇低空航测系统建筑物纹理获取与处理技术

发布时间：2018-08-14 12:43

【摘要】：自美国前副总统戈尔1998年提出“数字地球”概念以来,世界上许多国家都不约而同的以城市作为主要切入点之一,关于“数字城市”的理论与框架体系研究一直为地理信息科学领域的热点。随着当今物联网、云计算的飞速发展,作为新一代信息支持的“智慧城市”对高分辨率遥感数据需求与日俱增,最快最准确地获取基础地理数据是国家基本建设和信息化进程的首要问题。近几年低空遥感技术发展很快,它弥补了卫星遥感和中高空航测在及时性与精细度(影像精细度包括厘米级精细度和建筑物三维精细度)方面的不足。现有的低空航测多为轻小型固定翼无人机搭载单个数码相机,虽然具有光度学方面的光能量优势,但也有几何学方面的窄像幅劣势,这种劣势不仅使得作业效率低,更重要的是,它导致航测精度降低。为了解决此精度低下的问题,根本之路是使用宽角相机。而国内外成熟的组合宽幅面相机一方面缺乏无人机平台上应用的客观基础,另一反面由于受航高和空间摄影姿态限制(垂直或小角度摄影),仅能获得部分主体建筑的高度信息和顶部纹理信息,难以满足城市建筑物三维数据获取的要求。本文结合新近发展的多传感器集成技术,研究适应于无人机低空航测的组合宽幅面、宽视场角的航空相机(本文称纹理相机),通过快速安全地采集城市建筑物三维信息,建立镶嵌有影像纹理的真实感城市三维模型。其主要研究内容如下： (1)大比例尺低空倾斜航测飞行系统选择和传感器系统设计。首先,分析影响大比例尺低空倾斜航测获取多视、高清晰度和大比例尺测量精度影像的因素,作为大比例尺低空航测飞行系统选择和传感器系统设计的依据。分析可知,要想获取高分辨率、高清晰度影像,选择的飞行器必须飞得低飞得慢,从而选择无人飞艇低空航测系统作为城市三维建模主要数据源获取的遥感平台,分析了其优势所在,并简述系统的组成设计。在此基础上,研究和设计了适应于无人飞艇的轻小型组合宽角相机系统,包括纹理相机模型设计和稳定平台设计。 (2)基于纹理相机系统的低空航测精度分析及提高方法研究。首先,分析与宽角成像有关的精度问题,包括相对定向精度和测图精度。分析表明,低空航测的精度与所用相机的像场角有很大关系,因此低空航测应尽可能使用宽角相机。基于此结论,针对研制的适用于低空轻荷载无人机的纹理相机如何提高航测精度过程进行研究,包括有效像幅判断、自检校自稳定技术、采用大重叠度航空摄影的作业方法。最后,通过实验对比论证利用宽角相机进行大重叠度低空航测,可以有效提高成果精度。 (3)研究了基于纹理相机影像的纹理提取方法,包括建筑物轮廓线与影像配准的自动快速方法和半自动方法。首先,根据建筑物边界多以直线表达,因此如何精确获取影像上建筑物边缘的直线特征是解决纹理提取的关键所在。提出了利用Canny算子进行边缘提取,通过角度和距离限制,利用带权置信度获取最佳影像边缘线的自动快速方法。此外,考虑到在多幅影像中建立建筑物空间轮廓线与其影像直线特征映射关系的复杂性,对通过影像直线特征同名匹配来提高映射关系建立半自动化程度的可能性进行了探讨,提出了基于多视几何约束(核线约束、三视张量约束、相似性约束)的影像直线特征同名匹配策略。最后,对多纹理影像进行优选算法研究并进行几何纠正处理。 (4)研究了建筑物立面纹理颜色优化方法,包括纹理影像模糊消除和相邻影像色调一致性处理。首先,分析了纹理影像模糊现象产生的原因,包括噪声影响、影像像移以及几何纠正内插处理等。在对比现有图像复原算法优缺点基础上,提出了基于高频强调的MTF的图像复原算法,实现了利用刃边边缘图像数据估计MTF曲线,避免反复估计PSF过程,根据影像质量评价标准评定了该方法的有效性。其次,根据建筑物纹理构建中,存在单个建筑物各个面色彩斑驳,相邻建筑物色彩过渡不自然的情况,提出了基于纹理相机影像的主色调匹配方法,使偏暗或偏亮的影像得以纠正,处理后的纹理影像色调基本一致,达到三维城市视觉上的一致性和美观性效果。
[Abstract]:Since former US Vice President Gore put forward the concept of "Digital Earth" in 1998, many countries in the world have taken cities as one of the main entry points. The research on the theory and framework system of "Digital City" has been a hot topic in the field of geographic information science. The demand for high-resolution remote sensing data is increasing day by day in the new generation of "smart cities" supported by information. It is the most important problem for national capital construction and informatization process to obtain basic geographic data quickly and accurately.
In recent years, low-altitude remote sensing technology has developed rapidly, which makes up for the shortcomings of satellite remote sensing and mid-altitude aerial survey in timeliness and precision (image precision includes centimeter-level precision and building three-dimensional precision). In order to solve this problem, the fundamental way is to use wide-angle cameras. On the one hand, the mature combination wide-width cameras at home and abroad lack the application on UAV platforms. On the other hand, due to the limitation of altitude and space photographic attitude (vertical or small angle photography), only the height information and the top texture information of some main buildings can be obtained, which is difficult to meet the requirements of 3D data acquisition of urban buildings.
In this paper, combining with the recent development of multi-sensor integration technology, a combined wide-width, wide-field-of-view aerial camera (texture camera) suitable for low-altitude aerial survey of unmanned aerial vehicles (UAVs) is studied. By collecting the three-dimensional information of urban buildings quickly and safely, a realistic city three-dimensional model with image texture is built. :
(1) Selection of large-scale low-altitude tilt aerial survey flight system and design of sensor system. Firstly, the factors affecting the large-scale low-altitude tilt aerial survey to obtain multi-view, high-definition and large-scale measurement accuracy images are analyzed as the basis for large-scale low-altitude aerial survey flight system selection and sensor system design. Taking high-resolution and high-definition images, the selected aircraft must fly low and slow, so the unmanned airship low-altitude aerial survey system is selected as the remote sensing platform for urban three-dimensional modeling. Its advantages are analyzed and the composition and design of the system are briefly described. On this basis, the lightness suitable for unmanned airship is studied and designed. Small combined wide-angle camera system, including texture camera model design and stable platform design.
(2) The analysis and improvement of low altitude aerial survey accuracy based on texture camera system. Firstly, the precision problems related to wide angle imaging, including relative orientation accuracy and mapping accuracy, are analyzed. The analysis shows that the accuracy of low altitude aerial survey is closely related to the field angle of the camera used, so the wide angle camera should be used as far as possible in Low Altitude Aerial survey. The conclusion is that the texture camera suitable for low altitude light-load UAV can improve the aerial survey accuracy, including effective image amplitude judgment, self-calibration and self-stabilization technology, and the method of aerial photography with large overlap. Effectively improve the accuracy of the results.
(3) Texture extraction methods based on texture cameras are studied, including automatic, fast and semi-automatic methods for building contour and image registration. Firstly, according to the fact that building boundaries are mostly expressed in straight lines, how to accurately obtain the linear features of building edges on images is the key to texture extraction. Canny operator is used to extract the edge of the image, and the weighted confidence is used to obtain the best image edge line. In addition, considering the complexity of mapping between the building spatial contour and the image linear feature in multiple images, the mapping level is improved by matching the image linear feature with the same name. This paper discusses the possibility of establishing semi-automation degree, and proposes an image line feature homonymy matching strategy based on multi-view geometric constraints (kernel-line constraints, three-view tensor constraints, similarity constraints).
(4) The texture color optimization methods of building elevation are studied, including texture image blur elimination and adjacent image tone consistency processing. Firstly, the causes of texture image blur are analyzed, including noise, image motion and geometric correction interpolation. An image restoration algorithm based on high-frequency emphasized MTF is proposed, which uses edge image data to estimate MTF curve and avoids repeated estimation of PSF process. The effectiveness of this method is evaluated according to image quality evaluation criteria. Secondly, according to building texture construction, there are color mottles on each surface of a single building, and adjacent buildings have color overlap. In order to correct the dark or bright images, a method of main tone matching based on texture camera image is proposed. The color of texture image after processing is basically the same, which achieves the consistency and aesthetic effect of 3D city vision.
【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：P231

【参考文献】