透過您的圖書館登入
IP:18.118.150.80
  • 學位論文

以即時影像辨識為基礎之輪型機器人間之距離估測之研究

The Research of Distance Estimation for Moving Car-Like Robots Based on Real-Time Image Recognition

指導教授 : 游文雄
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


在本篇論文裡, 我們提出一演算法來分析兩部輪型機器人的影像, 使得其中一部輪型機器人能追隨另一部輪型機器人的軌跡。我們使用兩部輪型機器人, 分別為X80及i90機器人。當輪型機器人(i90) 以自身位置當做座標中心點,以單眼攝影機取得X80機器人的即時影像, 做二值化處理, 經過膨脹及侵蝕的處理來消除背景的雜訊使得二值化影像中只存在X80機器人的白色像素區塊, 希望藉此準確估測出另一部輪型機器人(X80) 的實際所在位置之座標值, 並達成兩部機器人軌跡追隨。首先, 根據實驗數據, 即時影像中的X80機器人的背面與側面影像面積因有些微的比例差距, 所以我們以X80機器人影像的車寬做數條分割線並編號, 比對前後兩到三張影像中的分割線號碼差值即可獲得X80機器人當下的行動狀態(例如: 前進、左右移動、靜止), 並獲知X80機器人影像是否為其背面或側面。接著, 再以背面面積所測到的兩車間距離為基準, 根據兩車間不同距離的白色像素區塊面積計算出X80機器人影像的寬, 以最小平方誤差近似出一條兩車間距離對應於X80機器人白色像素影像的寬之曲線, 當判斷出影像為X80機器人側面時, 乘上一側面相對應於背面的面積比例,再利用此一近似曲線求得初步的估測距離, 並利用六十組影像的平均值去修正誤差, 如此即可獲得較準確的修正距離。另一方面, 由於i90攝影機擁有一固定視角, 當已知視角及兩車間修正距離, 可利用三角定理計算出X80機器人在x 軸方向(垂直於i90機器人行走方向) 的估測座標值。由於i90攝影機在每次重新開機後的初始位置不盡相同, 因此,i90攝影機初始角度位置對X80機器人在x 軸之估測座標值會產生誤差, 故需於每次i90機器人開始動作後便需估測攝影鏡頭初始角度對x軸座標值, 以此x 軸初始座標值做為往後每次估測X80機器人x 軸估測座標值的修正量, 如此可得到修正後之兩車間距離與X80機器人的x 軸座標值。最後本論文將以實作來驗證我們所提出方法的準確性與有效性。

並列摘要


In this thesis, an algorithm is proposed for analyzing the image of two carlike robots for one of them to track the other. We use X80 and i90 robots for experiments. The image of the X80 is captured by a single eye camera of the i90 robot, and then, it is binarized and eroded to get the simple binary image by filtering the noisy of the environment, and obtain the correct area of the X80 robot binary image. Because the area ratios of the back and side images of the X80 robot are different in the same distance, the theory of the divided lines is proposed to determine whether the captured image is back or side of the X80 robot. After several different distance experiments between these two robots, an approximation curve is obtained for estimating distances via least squares errors according to the image area widths of the X80 robot. On the other hand, when the estimated distance and the fixed visual angle of the i90 camera are known, the x-coordinate of the X80 robot can be obtained by the triangle theorem. Finally, the accuracy and validity of our proposed algorithm are verified by experiments.

參考文獻


[1] M.-C. Lu, W.-Y. Wang, and C.-Y. Chu, “Image-Based Distance and Area Measuring Systems, ” IEEE Sensors Journal, vol. 6, no. 2, pp. 495-503, April 2006.
[2] C.-C. Hsu, M.-C. Lu, W.-Y. Wang, and Y.-Y. Lu, “Three-dimensional measurement of distant objects based on laser-projected CCD images, ” IET Sci.
Method for the Measurement of Motion Parameters of IPMC Actuators, ”IEEE Transactions on Instrumentation and Measurement, vol. 58, no. 8, pp.2626-2633, August 2009.
[5] J. Yu and J. Amores, N. Sebe, P. Radeva, and Q. Tian, “Distance Learning for Similarity Estimation, ” IEEE Transactions on Pattern Analysis and Machine
[6] M.S. Stefanou and J.P. Kerekes, “Image-Derived Prediction of Spectral Image Utility for Target Detection Applications, ” IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 4, pp. 1827-1833, April 2010.

延伸閱讀