• 한국컴퓨터그래픽스학회논문지
• /
• 제24권5호
• /
• pp.11-19
• /
• 2018

본 논문에서는 노이즈를 포함한 채 오버랩 영역이 적은 두 점군을 정합할 때 정확도와 수렴 속도를 향상시키는 알고리즘을 제시한다. 정확도를 높이기 위하여 점군의 기하학 정보를 최대한 활용하며, 정합 단계에서는 노이즈가 포함된 점군에서 오버랩 되는 영역을 적절히 선택하고, 개선된 가속 알고리즘을 사용하여 정합 속도를 향상시킨다. 정확도를 향상시키는 기존의 방법은 노이즈가 많은 점군에 적용할 수 없으므로, 본 논문에서는 정합에 사용되는 영역을 선택하는 것으로써 기존 방법의 문제를 해결하였다. 또한 똑같은 점군쌍에서만 적용되는 가속 알고리즘을 낮은 오버랩의 점군쌍에 적용하였다. 기존의 방법에 간단한 알고리즘을 추가함으로써 서너 배 더 빠른 수렴 속도를 낼 수 있도록 하였다. 결론적으로, 노이즈가 많고 오버랩이 적은 점군쌍의 정합에 있어서 본 논문에서 제시하는 알고리즘을 적용하면 속도와 정확도가 향상되었음을 알 수 있다.

• 자동차안전학회지
• /
• 제12권2호
• /
• pp.33-39
• /
• 2020

This paper presents an Around View Monitoring (AVM) stop-line detection based longitudinal position correction algorithm for automated driving on urban roads. Poor positioning accuracy of low-cost GPS has many problems for precise path tracking. Therefore, this study aims to improve the longitudinal positioning accuracy of low-cost GPS. The algorithm has three main processes. The first process is a stop-line detection. In this process, the stop-line is detected using Hough Transform from the AVM camera. The second process is a map matching. In the map matching process, to find the corrected vehicle position, the detected line is matched to the stop-line of the HD map using the Iterative Closest Point (ICP) method. Third, longitudinal position of low-cost GPS is updated using a corrected vehicle position with Kalman Filter. The proposed algorithm is implemented in the Robot Operating System (ROS) environment and verified on the actual urban road driving data. Compared to low-cost GPS only, Test results show the longitudinal localization performance was improved.

• Journal of Mechanical Science and Technology
• /
• 제19권12호
• /
• pp.2213-2223
• /
• 2005

In order to reconstruct a full 3D human model in reverse engineering (RE), a 3D scanner needs to be placed arbitrarily around the target model to capture all part of the scanned surface. Then, acquired multiple scans must be registered and merged since each scanned data set taken from different position is just given in its own local co-ordinate system. The goal of the registration is to create a single model by aligning all individual scans. It usually consists of two sub-steps: rough and fine registration. The fine registration process can only be performed after an initial position is approximated through the rough registration. Hence an automated rough registration process is crucial to realize a completely automatic RE system. In this paper an automated rough registration method for aligning multiple scans of complex human face is presented. The proposed method automatically aligns the meshes of different scans with the information of features that are extracted from the estimated principal curvatures of triangular meshes of the human face. Then the roughly aligned scanned data sets are further precisely enhanced with a fine registration step with the recently popular Iterative Closest Point (ICP) algorithm. Some typical examples are presented and discussed to validate the proposed system.

• 대한의용생체공학회:의공학회지
• /
• 제37권2호
• /
• pp.68-74
• /
• 2016

In this study, we investigated the rotational characteristics which were comprised of directionality and linearity of target registration error (TRE) as a study in advance to enhance the accuracy of contour-based registration in neuronavigation. For the experiment, two rigid head phantoms that have different faces with specially designed target frame fixed inside of the phantoms were used. Three-dimensional coordinates of facial surface point cloud and target point of the phantoms were acquired using computed tomography (CT) and 3D scanner. Iterative closest point (ICP) method was used for registration of two different point cloud and the directionality and linearity of TRE in overall head were calculated by using 3D position of targets after registration. As a result, it was represented that TRE had consistent direction in overall head region and was increased in linear fashion as distance from facial surface, but did not show high linearity. These results indicated that it is possible for decrease TRE by controlling orientation of facial surface point cloud acquired from scanner, and the prediction of TRE from surface registration error can decrease the registration accuracy in lesion. In the further studies, we have to develop the contour-based registration method for improvement of accuracy by considering rotational characteristics of TRE.