透過您的圖書館登入
IP:18.189.14.219
  • 學位論文

基於基因演算法做攝影機參數自我校正做室外自動車導航

Outdoor Navigation of Autonomous Land Vehicle Using Camera Self-Calibration based on Genetic Algorithm

指導教授 : 駱榮欽

摘要


本論文中,我們提出一個應用於自然環境的攝影機參數自我校正的研究,搭配基因演算法完成攝影機的參數校正,並將其結果應用於雙眼立體視覺之無人自動導航車。 在本論文中所採用的自我校正,並非使用特定物體,而是以自然環境中經常可獲得的平面。首先,自雙攝影機擷取左右影像上共面的對應點,藉由基因演算法產生母代攝影機參數染色體,並以之將二維影像重新投影至三維空間,再利用共平面條件為適應函數用以檢驗參數的準確度。在我們的研究中採用實數型的基因演算法,除了節省編碼及解碼的運算時間,還可提高系統的精確度。對於染色體的複製與交配的過程,我們不僅使用啟發式交配、並加入類比於二元字串的單點交配,以及結合兩者的方式。而突變的策略則使用加入較大的隨機變量,以產生不同的物種以避免局部最小值。 相較於疊代的方式求解,使用基因演算法僅需合理的邊界範圍,不必給定初始值、或梯度值之類的資訊,如此可避免數學上複雜的運算,並能夠得到收斂良好的結果。在論文中也提供一些實驗結果證明所提方法的可用性。

並列摘要


In this paper, we have proposed a modified camera self-calibration approach apply to natural environment for autonomous land vehicle (ALV), based on genetic algorithm (GA). In this paper, the approach does not need. Firstly, the acquisition image points which are corresponding between the stereo images (left and right), are coplanar. The population created by GA consisting of the camera parameters. Those parameters are used to reproject the 2D image points onto 3D world, then, the points in 3D are regarded the coplanar condition as the objective function, which is to evaluate the parameters. In our research, we adopt the real–coded GA for saving the computation time in encoding and decoding the binary string, but also increasing the precision of the system. There are three ways for the process of selection and reproduction: heuristic crossover, single-point crossover, and combination of the above-mentioned both. The mutation operator applies some random arbitrary number to generate different individuals to avoid the local minimum. In contrast to the iterative method, GA needs only the reasonable boundary; so that the complex computation kept off since the initial value and the gradient are unnecessary for GA, and the solution is convergence. Some experimental results are also included to demonstrate the applicability of proposed method.

參考文獻


0] 盧振育,基於立體電腦視覺及人工智慧策略搭配雷射光作室外自動車導航之研究,碩士論文,國立臺北科技大學自動化科技研究所,台北,2004。
2] 葉哲昌,利用鬆弛相關性改進3D對應點匹配與重建應用於室外自動導航車,碩士論文,國立臺北科技大學自動化科技研究所,台北,2006。
3] 張煜青,以雙眼立體電腦視覺配合人工智慧策略作室外自動車導航之研究,碩士論文,國立臺北科技大學自動化科技研究所,台北,2003。
Y. I. Abdel-Aziz and H. M. Karara, “Direct linear transformation into object space coordinates in close-range photogrammetry,” in Proc. Symp. Close-Range Photogrammetry, Univ. Illinois at Urbana Champaign, Jan. 1971, 26-29, pp. 1-18.
R.Y. Tsai, “A Versatile Camera Calibration Technique for High-Accuracy 3D Machine Vision Metrology Using Off-the-Shelf TV Cameras and Lenses,” IEEE Journal of Robotics and Automation, Vol.RA-3, No.4, Aug. 1987, pp.323-344.

被引用紀錄


陳建成, C. C. C. (2013). 基於景象的平面限制運用基因演算法實現3D立體攝影機自我校正 [master's thesis, National Taipei University of Technology]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0006-0708201314143500

延伸閱讀