Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : May 10, 2017 - May 13, 2017
This paper proposes a Self-Localization method using traffic signs for automated vehicles. The proposed method aims for longitudinal Self-Localization improvement. In order to search on the highest probability position, map matching is performed using a predefined digital map data. Therefore, the position is moved virtually around the actual vehicle position. Then, the images (Predictive sign image, Ideal sign image) for matching are calculated for each position. A predictive sign image is extracted from the road forward camera image as well as an ideal sign image is created using the digital map data. The similarity between these two images is calculated using ZNCC (Zero-means Normalized Cross Correlation). The corresponding posterior probability of the similarity is updated by BBF (Binary Bayes Filter) and vehicle position is estimated accordingly. The real experiments have indicated that the localization accuracy in the longitudinal direction has been improved using the proposed method.
