利用360度高解析環景影像和粒子群演算法的定位方法

Position Measurement Using Ultra-High Resolution 360-Degree Panoramic Images and Particle Swarm Optimization

王健安(Jian-An Wang)

指導教授 ： 陳偉堯

國立臺北科技大學/工程學院/土木與防災研究所/碩士(2013年)

https://doi.org/10.6841/NTUT.2013.00611

摘要

近年來數位相機功能日新月異，藉由數位相機即能拍攝出千萬畫素影像，然而受拍攝角度、範圍影響，並無法將眼前所看到的影像拍攝於一張照片之中，但使用 GigaPan 機器手臂搭配一般數位相機，即可進行超高解析度影像拍攝。
GigaPan 除可用於拍攝紀錄影像外，本研究也將其與 PSO 演算法結合，應用於定位空間座標；於目標物周圍任選 3 個拍攝點位，並以 GPS 量測拍攝點位空
間座標(當控制點使用)，各點位皆拍攝一組 360 度高解析度環景影像，利用
GigaPan 可拍攝記錄 360 度環景影像之原理，藉由環景影像之像素可計算任兩待測物與拍攝點位間之夾角，並藉此夾角模擬從 3 個拍攝點位對目標物射出虛擬射
線，並利用 PSO 演算法快速搜尋之特性與三方交會之原理，在 3 個拍攝點位分別以 0~360 度之隨機角度進行探討，在各種虛擬射線交集組合中朝最小交集區前進，其虛擬射線交集處位置即為目標物之空間座標。
除了可應用於定位目標物空間座標外，本研究亦藉此方法用於：(1)定位建築物四周，(2)觀測邊坡上樹木之傾斜角度，藉由不同時期拍攝之影像，觀測其樹木傾斜角度是否有增加趨勢，以判釋該邊坡是否處於穩定狀態，並將此角度與光達掃描結果相互比較，結果顯示使用 GigaPan 和 PSO 可以得到合理的成果，證明這項方法具有應用潛力。

關鍵字

GigaPan ； PSO ； 粒子群演算法 ； 高解析影像 ； 環景影像 ； 空間定位

並列摘要

The changing features of digital cameras in recent years enable the capturing of images in resolution as high as ten million pixels. Nevertheless, limited by camera angles and range, it is not always possible to capture what is in front of the eye in one picture.Now, the GigaPan Robot Arm combined with any common digital camera makes image capturing in ultra-high resolution possible.
In addition to its use in capturing and recording images, this study applied the combination of GigaPan and the PSO algorithm to locating spatial coordinates. Three capturing points were randomly selected around a target and their spatial coordinates were measured by GPS (and used as control points). At each of these points, one set of 360 degree high-resolution panoramic images were captured. Based on the theory that GigaPan can capture and record 360 degree panoramic images, it was possible to calculate the angle between any two objects and the capturing points with the pixels of the panoramic images. With this angle, virtual rays were simulated that streamed from the three capturing points to the target. With the rapid search feature of the PSO algorithm and based on the principle of triangulation, the three capturing points were examined randomly at 0~360 degrees. As different virtual rays moved towards minimum intersection areas among intersection unions, their intersections were the spatial coordinates of the target.
Besides locating the spatial coordinates of a target, this study also applied the same method to: (1) locating the surroundings of a building, (2) observing the
inclination angle of trees on a slope. The inclination angle was observed in images captured in different periods to see if it tended to increase, thus determining if the slope was in a stable state. Additionally, the angle was compared with the results from LiDAR scanning. The comparison shows that it is possible to obtain reasonable results using the combination of GigaPan and PSO, proving the application potential of this method.

並列關鍵字

GigaPan ； PSO ； Ultra-High Resolution Images ； Panoramic Images ； Position Measurement

參考文獻

[28] 陳泓錡 (2012)，猴山岳步道崩塌地的地形地貌記錄與分析，碩士論文，國立臺北科技大學土木與防災研究所。

[29] 溫仁志 (2010)，以粒子群演算法分析邊坡臨界滑動面，碩士論文，國立臺北科技大學土木與防災研究所。

[30] 劉正達 (2011)，人工智慧方法應用於聖火傳遞路徑最佳化，大專體育學刊，第13卷，第4期，第368頁-378頁。

[31] 鐘雅蘭 (2010)，猴山岳邊坡樹木生長方向的統計分析，碩士論文，國立臺北科技大學土木與防災研究所。

[1] Bertone, M. A., Blinn, R. L., Stanfield, T. M., Dew, K. J., Seltmann, K. C., and Deans, A. R. (2012). "Results and Insights from the NCSU Insect Museum GigaPan Project," ZooKeys, 209, pp. 115–132.