以粒子群最佳化微調相機校正之映射矩陣

Mapping Matrix Fine Tuning for Camera Calibration using Particle Swarm Optimization

陳姿帆(Tzu-Fan Chen)

指導教授 ： 鄭志宏

共同指導教授 ： 林義隆(Yih-Lon Lin)

義守大學/電機資訊學院/資訊工程學系/碩士(2017年)

摘要

隨著科技的發展，相機的類型及應用變的更多元，而為了拍攝較大的範圍，許多應用上會選擇使用有廣角類型鏡頭的相機，但其取得的影像往往有較嚴重的徑向形變，若要採用這些影像做進一步的應用，如計算物體距離等等，其結果可能就會受到形變的影響。所以影像校正是必要的，其技術近年也一直是熱門的研究議題，但目前的校正成果仍可做進一步改善，尤其是在影像四周部分。
本篇論文內容主要即是使用粒子群最佳化演算法(PSO)微調相機校正中的影像校正參數，以Python程式使用套件OpenCV裡的相機校正系列指令，就可計算出內部參數、外部參數、形變資料等等相機參數，而其中用來重建影像的位置映射矩陣(Mapx、Mapy)就是PSO的調整目標；而在考慮廣角鏡頭的特性下，我們假設影像僅受徑向形變影響，根據其特性及一些實驗顯示，我們可利用極座標轉換使PSO調整目標的維度降低，並且在這個方法下，我們可只對影像四周形變較為嚴重的地方進行調整，目前實驗結果皆可降低一般用來評估相機校正成果的Re-projection Error值。

關鍵字

相機校正 ； 粒子群最佳化演算法 ； 影像位置映射矩陣

並列摘要

With the development of science and technology, the type and application of cameras are becoming more complex and diverse. In order to acquire images with wider angle of view, many applications adopt cameras with wide-angle type lens. Due to lens properties, the acquired images usually suffer lots of radial distortion. For further applications using these images, camera calibration (or image calibration) is necessary and has been developed for many years. However, the current results of camera calibration can be further improved, especially in the surrounding part of images.
The purpose of this thesis is to improve the results of camera calibration based on traditional method by using particle swarm optimization (PSO) algorithm. Using OpenCV to calibrate camera under Python, we can find the camera intrinsic parameters, extrinsic parameters, and distortion parameters, etc. One of these parameters used to rectify images is the mapping matrix (Mapx and Mapy), which is the particle of PSO in this thesis. Consider the wide angle lens characteristic, the assumption is made that only the radial distortion is taken into account. According to some test results, we can use the transformation from Cartesian coordinates to polar coordinates to reduce the particle dimension. In this method, we can fine tune only the surrounding part of the image, which has most apparent distortion. The experimental results show that the Re-projection Error value can be reduced, which is used to assess calibration performance.

並列關鍵字

Camera Calibration ； Particle Swarm Optimization ； Mapping Matrix

參考文獻

[2] M. A. Penna, “Determining camera parameters from the perspective projection of a quadrilateral,” Pattern Recognit., vol. 24, no. 6, pp. 533–541, 1991.

[3] R. Tsai, “A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off-the-shelf TV cameras and lenses,” IEEE J. Robot. Autom., vol. 3, no. 4, pp. 323–344, 1987.

[4] R. Tsai, “An efficient and accurate camera calibration technique for 3D machine vision,” in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, 1986.

[6] J. Heikkila, “Geometric camera calibration using circular control points,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 10, pp. 1066–1077, 2000.

[7] Z. Zhang, “A flexible new technique for camera calibration,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 11, pp. 1330–1334, 2000.