θ-ρパラメータによる多角形の物体検出 [in Japanese] Object Detection Using Straight Line Matching in θ-ρ Space [in Japanese]
Access this Article
Search this Article
The contours of many industrial parts contain straight lines and positions of the lines are therefore useful information for object detection. This paper presents a matching technique of straight lines in <i>θ</i>-<i>ρ</i> space. Any lines in 2D space are represented with parameters <i>θ</i> and <i>ρ</i> by Hough transform. To find an object is to find combination of line parameters in <i>θ</i>-<i>ρ</i> space. Our matching method contains edge detection, line detection and matching processes. At first, we perform edge detection for model and scene image to detect contour of objects. Next, we extract the straight lines using Hough transform for provided edge image. Then in the matching process, we perform matching with parameters of straight lines. Matching process contains <i>θ</i> matching, <i>ρ</i> matching and pose estimation. In the <i>θ</i> matching, we use the relative <i>θ</i> values of corresponding lines. In the <i>ρ</i> matching, we compute parameters of transfer and deviations of <i>ρ</i> values. Finally, in the pose estimation we compute transfer parameters using corresponding intersection points of the straight lines. The experimental results show our method is robust for rotation, occlusion and scaling of objects.
- IEEJ Transactions on Electronics, Information and Systems
IEEJ Transactions on Electronics, Information and Systems 128(4), 600-606, 2008-04-01
The Institute of Electrical Engineers of Japan