【発表内容】
タイトル：撮影条件の変化に頑健な車載カメラ画像間の対応付け
一般的な車載カメラとGPSを利用した街並変化検出システムの実現を目指し、このシステムに必要な基盤技術の一つである「異なる時刻に撮影された画像間の対応付け」を提案する。提案手法では、撮影タイミングの大きく違う過去・現在画像間の比較を容易にするために、連続数フレームを繋げた仮想パノラマ画像間での対応付けを行なう。このパノラマ画像は、車載カメラ方位の回転補正を行なった撮影画像をパノラマ画像面に投影していくことにより生成できる。さらに過去・現在パノラマ画像間の比較時には、走行車線の違いに対応するための画像の拡大・縮小、日照の変化やノイズに対する頑健性向上のためのエッジ特徴量抽出を行なった。車載カメラから計算サーバへの無線通信を想定するために低解像度・低フレームレートであり、かつ撮影条件の異なる画像間の対応付け実験により、提案手法の有効性を確認した。

【会議の内容】
ICDSCで取り扱われるテーマは分散型smart cameraのアーキテクチャ、アルゴリズム、画像処理技術で特に産業用として実用化可能な技術に重点がおかれています。 そのため、発表はデモセッションや企業からの参加者も多く大変興味深いものでした。

【研究技術交流等】
ポスターセッションでは沢山の方々から、質問や今後の研究に参考となる意見を頂きました。自身の研究の弱点、アピールポイントが把握できました。
このような機会を与えて頂きました大学院教育改革支援プログラムに感謝いたします。

　
【発表内容】
タイトル：Dynamic Burst Ordering for Burst-Cluster Transmission to Improve Fairness in OBS Networks

Abstract: Hop-based burst-cluster transmission has been proposed to improve fairness in OBS networks. In this method, multiple bursts are assembled simultaneously and a burst-cluster is generated from the bursts. By arranging the bursts within the burst-cluster in order from the smallest number of hops to the largest one, the burst loss probability for a small (large) number of hops increases (decreases), improving fairness. However, depending on the amount of traffic on each link, the burst loss probability for a small number of hops becomes larger than that for a large number of hops for some source nodes. Therefore, all source nodes can not always improve fairness. In this paper, we propose dynamic burst ordering for the hop-based burst-cluster transmission. In the proposed method, each source node calculates the burst loss probability for each number of hops from the number of received ACK and NACK messages. Based on the calculated burst loss probabilities, the source node changes the order of bursts within a burst-cluster dynamically. It is expected that this dynamic burst ordering can improve local fairness for each source node. We evaluate the performance of the proposed method in tandem networks by simulation. Numerical examples show that the proposed method is effective for improving fairness for each source node regardless of the amount of traffic on each link. It is also shown that the proposed method increases the overall burst loss probability, however, each source node can decrease the largest burst loss probability among the ones for all destination nodes.

【会議の内容】
The World Congress on Engineering and Computer Science (WCECS) 2008 is organized by the International Association of Engineers (IAENG), a non-profit international association for the engineers and the computer scientists. The congress has the focus on the frontier topics in the theoretical and applied engineering and computer science subjects. It took place in San Francisco, USA, 22-24 October, 2008. WCECS 2008 is composed of 15 conferences. There were 376 presenters with an acceptance ratio of 56.71%. I attended the International Conference on Communication Systems and Technologies (ICCST). A variety of computer sciences areas has been presented such as Optical networking, Wireless networking and P2P networking. Optical networking includes fairness improvement in OBS networks as well as WDM PON architecture studies. Wireless networking includes topics such as Fast Channel Switching Technologies in IEEE 802.22 WRAN Systems, QoS Enabled MAC Protocol studies and Fault Tolerance and Network Depth in Hierarchical Wireless Sensor Networks. Finally, for P2P networking, an interesting paper about Message Routing Algorithm with Additional Node-Information for Capability-Aware Object Management was presented.

【研究技術交流等】
I had the opportunity to talk to many international scientists from India, France, Algeria, and Tunisia. It was an enriching experience, especially because we work in a similar research area which is computer networks. I found particularly interesting the presentation of Mr. Francois Armando, From LAAS/CNRS which is a research center in France. Mr. Francois Armando’s research is about QoS in the transport layer. The presentation title is "Introducing a Collaborative Congestion Control Based on TFRC". In this presentation Mr. Francois Armando shared with us his current simulation results and his future work.