航海環境時系列画像からのリアルタイム船舶抽出手法

概要

論文の詳細を見る
This paper proposes a method for detecting ships on real-time using navigational image sequence. As pointed out in the previous reports, our proposed method has problem in processing speed performance, although it is precise precision. First, this paper analyzes this problem, and proposes a new method for solving them. Our method distinguishes ships from the character of a moving vector of region. A moving vector is calculated on matching process of region. Matching process takes enormous-processing time. On the other hand, the Lucas-Kanade method (LKM) is popular version two-frame differential methods for optical flow estimation. The LKM can get a pixel flow first. We examined that applying the LKM in substitution for matching process, and we proposes a new method for detecting ships on real-time. We inspected an effect of prposed method by an experiment.
社団法人日本航海学会の論文
2007-09-25