CAJ | 학술논문

为验证地面大转角、小面幅序列数码影像直接相对定向质量.推导了一套基于 SIFT 特征匹配的计算机视觉分析方法,解算过程简洁明了.研究结果表明:基于计算机视觉方法能够有效解决大转角影像、特别是车载序列数码影像相对定向中的匹配难题,解算结果比较稳定;从目前的结果来看,直接相对定向质量还不高,但研究方法对室内移动机器人环境感知、无人驾驶等导航领域具有一定的借鉴价值;最后分析了制约其质量的主要原因,为今后分析和提高视觉量测质量提供了一条思路.
위험증지면대전각、소면폭서렬수마영상직접상대정향질량.추도료일투기우 SIFT 특정필배적계산궤시각분석방법,해산과정간길명료.연구결과표명:기우계산궤시각방법능구유효해결대전각영상、특별시차재서렬수마영상상대정향중적필배난제,해산결과비교은정;종목전적결과래간,직접상대정향질량환불고,단연구방법대실내이동궤기인배경감지、무인가사등도항영역구유일정적차감개치;최후분석료제약기질량적주요원인,위금후분석화제고시각량측질량제공료일조사로.
@@@@In order to verify the direct relative orientation quality of narrow frame sequential images with big rotation angels, this study derives a computer vision analysis method based on SIFT, which simplifies the calculation process. The comparison result between orientation parameters from POS and that based on computer vision shows that the method can overcome the matching problem of images with big rotation angles. In particular, for vehicle based sequential images, the result is quite stable. Currently the quality of direct relative orientation using this method is not very high, however, the method provides a useful reference for many surveying and mapping related fields, such as indoor robot mobile environment awareness and unmanned automobile. Furthermore, the study analyzes the factors that affect the quality of orientation based on computer vision and provides a solid foundation for future study.