파노라마 영상의 정합오차 측정 기반 전방위 상황인식 시스템 유지보수 기술

Abstract

To apply omnidirectional situational awareness system in actual vehicle applications, it is essential to not only generate high quality panoramic images but also maintain the system in online manner. In this paper, the authors newly design a stitching error metric that measures image stitching errors used to generate panoramic images. Considering the high complexity of outdoor environments, the authors propose a robust but light stitching error measurement based on edge-histogram encoding. In addition, they propose an alarm function notifying the need for the vehicle maintenance with the stitching errors. The proposed method is verified in outdoor environments that contain complex scene information such as a few buildings, pedestrian, and trees. The experimental results show that the proposed method effectively and robustly monitor the omnidirectional situational awareness systems to environmental changes.