基于航拍影像和LiDAR数据的山地滚石信息提取

发布时间：2018-04-05 00:10

本文选题：LiDAR　切入点：航拍影像　出处：《中国矿业大学》2017年硕士论文

【摘要】：近年来,滚石灾害时有发生,它具有分布范围广、发生突然、频率高等特点,严重威胁其周围的人类及各种构筑物,具有严重的危害性。通常,滚石防护需要通过野外实测获取滚石分布信息,再采取工程性防护措施,这种方法耗时耗力,且山区地形复杂,野外调绘具有一定的危险性。目前,国内外利用影像等数据进行滚石提取,存在提取精度低、误检率高、自动化程度低等问题。针对这些问题,本文拟提出一种综合利用航拍影像与Li DAR数据提取山地滚石信息新方法,主要包括:影像多尺度分割,对分割后的对象进行绿色植被信息和阴影信息提取,排除绿色植被和阴影对最终滚石信息提取的干扰,对去除植被信息和阴影信息的分割对象运用本文提出的基于阴影和高程的滚石提取算法,得到滚石信息。主要工作内容如下:(1)将绿色植被指数与面向对象思想相结合进行绿色植被信息提取。基于像元的绿色植被信息提取易出现噪声点多、提取结果不完整等问题。针对这些问题,本文将绿色植被指数运用到分割对象中能取得较好的提取效果。提取绿色植被信息,排除由于绿色植被形成的阴影对下文提取滚石信息的干扰。(2)改进一种基于彩色模型的阴影检测算法。针对航拍影像阴影检测中,与阴影区域具有相似光谱性质的绿地、亮度较暗的非阴影、亮度较高的阴影区域容易出现错检漏检等问题。在RGB和HSI色彩空间分别提出一种阴影指数进行阴影信息初步提取,根据阴影在RGB色彩空间中蓝光波段和绿光波段的大小关系对初步提取结果进一步过滤,再对两种色彩空间提取结果“与”运算作为最终阴影提取结果,最后通过实验验证本文的阴影提取算法。提取的阴影区域将作为的一个重要的上下文特征用于滚石信息提取。(3)提出一种基于阴影和高程的滚石信息提取算法。分析滚石在影像和LiDAR数据中的特征,综合利用滚石的光谱特征、上下文特征、Li DAR数据中的特征等,提出基于阴影和高程信息的滚石信息提取算法,并结合高程信息过滤得到最终滚石信息提取结果,最后进行精度评定和分析。本文提出的滚石提取算法识别精度高且提取轮廓准确,基本能够满足地政等部门的滚石防护需求。(4)采用面向对象的C++语言,基于Qt开发框架,使用开源的QGIS、GDAL、Laslib等类库进行二次开发,在微软Visual Studio 2015开发环境下实现本文提出的滚石信息提取流程,主要包括:影像与矢量数据处理、LiDAR数据处理、滚石信息提取等模块。
[Abstract]:In recent years, rolling stone disasters occur from time to time. They have the characteristics of wide distribution, sudden occurrence, high frequency and so on, which seriously threaten the human beings and all kinds of structures around them, and have serious harmfulness.Usually, the rolling stone protection needs to obtain the information of rolling stone distribution through field measurement, and then take engineering protection measures. This method is time-consuming and labor-consuming, and the terrain of mountain area is complex, so it is dangerous to adjust and map in the field.At present, there are some problems such as low precision, high rate of error detection and low degree of automation in the extraction of rolling stones with image and other data at home and abroad.In order to solve these problems, this paper proposes a new method for extracting rolling rock information from mountainous area by using aerial image and Li DAR data, which includes multi-scale segmentation of image, extraction of green vegetation information and shadow information from segmented objects.To eliminate the interference of green vegetation and shadow on the extraction of final rolling stone information, the rolling stone information is obtained by using the rolling stone extraction algorithm based on shadow and elevation proposed in this paper for the segmentation object which removes vegetation information and shadow information.The main work is as follows: (1) the green vegetation index is combined with the object oriented idea to extract the green vegetation information.The extraction of green vegetation information based on pixel is prone to many noise points and incomplete extraction results.In order to solve these problems, this paper applies green vegetation index to the segmentation object to obtain a better extraction effect.An improved shadow detection algorithm based on color model is improved by extracting green vegetation information and eliminating the interference of shadow caused by green vegetation to extract rolling stone information below.In the shadow detection of aerial image, the green space with similar spectral properties to the shadow region, the non-shadow with dark brightness and the shadow region with high brightness are prone to error detection and leak detection.A shadow index is proposed in RGB and HSI color space to extract shadow information. The preliminary results are filtered according to the relationship between the blue band and green band of shadow in RGB color space.Then the "and" operation of two color spaces is used as the final result of shadow extraction. Finally, the shadow extraction algorithm of this paper is verified by experiments.The extracted shadow region will be used as an important context feature for the extraction of rolling stone information. (3) A rolling stone information extraction algorithm based on shadow and elevation is proposed.This paper analyzes the features of rolling stones in image and LiDAR data, synthetically utilizes the spectral features of rolling stones and the features of Li DAR data, and puts forward a rolling stone information extraction algorithm based on shadow and elevation information.Combined with elevation information filtering, the final rolling stone information extraction results are obtained, and the accuracy evaluation and analysis are carried out at last.The rolling stone extraction algorithm proposed in this paper has high recognition accuracy and accurate contour, which can basically meet the rolling stone protection requirements of lands and other departments. It uses object oriented C language and is based on QT development framework.In this paper, we use the open source QGIS-GDALLLaslib library to develop the rolling stone information extraction flow in Microsoft Visual Studio 2015 development environment, which includes image and vector data processing module, rolling stone information extraction module and so on.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P694;P23

【参考文献】