| Summary: | 碩士 === 國立成功大學 === 測量工程學系碩博士班 === 91 === Automatic real-time surveying system is one of the developments in the field of surveying, while mobile mapping vehicle system based on close range photogrammetry becomes a primary research topic. Point transfer of the close-range photogrammetry requires the most manual operations. Thus, how to automate the point transfer becomes the foundation of the achieving real-time surveying. There are several automatic point transfer methods. However, most of them are used for matching high-level information in wide baseline applications. Recently, digital video becomes important in surveying because of its popularity, improved quality, and characteristics of gathering images rapidly and in quantity. These characteristics of digital video make short baseline image matching becoming one of the major research topics of Automatic Real-Time Surveying system. This research uses digital video as the tool of close-range photogrammetry, the environment of mobile mapping vehicle, such as streets and buildings, as targets. The effectiveness and quality of surveying optical flow method and NCC method and their results as well is to be analyzed and studied.
In this study, a new flowchart is designed for automatic extracting, measuring, tracking, and error detection of the feature points in digital video. This flowchart may be used with any point-based tracking method. The surveying environment is divided into two categories: (1) the fully 3-D environment and (2) the 2-D wall surface characteristics. For still targets, digital videos are took by moving the video camcorder from left to right, from fore to aft, and surrounding the targets. The tracking rate and the image matching accuracy of digital video images were analyzed. The results were then used for comparing the advantages and suitability of optical flow method and NCC method.
In addition, a new method is developed and presented to improve the function of traditional optical flow method. Test results show clearly that it significantly increases the maximum tracking distance and improves the reliability of the tracking results. The tracking range was increased from 3-4 pixels in traditional optical flow method to more than 30 pixels using the new method.
Also, the results indicate that the new flowchart is useful for automatic mosaic generation using digital video images. However, both DV-image matching approaches with sub-pixel accuracy and rules for adding new tracking points still ought to be further studied before the new method can be used in aerial triangulation applications.
|
|---|
