視覺障礙辨識之機器人平滑路徑規劃及控制

在本篇論文裡,我們提出一演算法使一部具單眼攝影機的輪型機器人完成障礙辨識、軌跡規劃及軌跡追蹤。此輪型機器人經由無線傳輸將週遭的圖像傳送到電腦端,藉由影像分析程式辨視障礙,進而規劃出避障路徑。首先,我們將利用圖像中的邊緣特徵辨識障礙物。因為所配的攝影機為單眼,故無法藉由視差算出物體的距離,我們利用當輪型機器人走近物體時物體放大的程度估算出約略的距離,進而求出物體約略的最大寬度。但是,當輪型機器人接近障礙物時,機器人並無法得知障礙物的深度，因此,我們提出一基於橢圓切線的避障路徑規劃。首先,依據障礙物寬度及與道路邊距建立一初始橢圓,並假設它可將障礙物包住,當機器人沿著此橢圓的切線走到障礙物側面時,再觀察障礙物的深度,此時可能只能觀察到障礙物一部分的深度,據此再判斷原建立之初始橢圓路徑是否能避過障礙物,若不能,則更新橢圓使其能包住此障礙物,並更新橢圓軌跡,接著再繼續往前移動並觀察障礙物深度,直到得到不與障礙物碰撞的安全路徑。最後我們將以實作來驗證我們所提出演算法的正確性及有效性。

關鍵字

輪型機器人；路徑規劃；避障；視覺

並列摘要

In this thesis, an algorithm for achieving obstacle recognition, motion planning, and trajectory tracking via a single-eye camera is proposed for the remote mobile robot via wireless connection. The single-eye camera is used to capture the environmental image and to recognize the obstacle image to construct the ellipse and to plan the collision-free path. Since it is hard to determine how far the distance from the mobile robot to the obstacle object using a single-eye camera by the stereo image technology, we use the ratio of the obstacle images from far to near to evaluate the distance to and width of the obstacle approximately. As an obstacle object is recognized by the active-vision of the mobile robot, a collision-free path planning is constructed based on the tangent line of an ellipse and edge feature of the obstacle image. Hence, an obstacle-covering ellipse is formed according to the largest width of the obstacle object and the distance to the lane edge from the part of the obstacle depth. Then, when the mobile robot drives along the tangent line of the ellipse to the side of the obstacle, it continues to judge if the path established by the initial ellipse is collision free. If not, a further obstacle-covering ellipse should be constructed until the path is collision free, and then the mobile robot keeps moving along the adapted tangent line of the ellipse. Finally, we demonstrate the validity of the proposed algorithm by experiments.

並列關鍵字

Mobile robot ； Path planning ； Collision free ； Visual

參考文獻

[1] V. Ferrari, L. Fevrier, F. Jurie, and C. Schmid, “Groups of adjacent contour segments for object detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 30, no. 1, pp. 36-51, 2008.

[2] A. Berg, T. Berg, and J. Malik, “Shape matching and object recognition using low distortion correspondence,” Proc. Conf. Computer Vision and Pattern

[3] D. Jacobs, “Robust and efficient detection of salient convex groups,” IEEE

Transactions on Pattern Analysis and Machine Intelligence, vol. 18, no, 1, pp. 23-37, 1996.

[4] B. Leibe and B. Schiele, “Scale-invariant object categorization using a scaleadaptive mean-shift search,” Proc. German Pattern Recognition Symp., 2004.

國際替代計量

視覺障礙辨識之機器人平滑路徑規劃及控制

未授權

主題瀏覽