We made fruitful progresses in the Digital Media Understanding Algorithm Evaluation Platform,Intelligent Surveil-lance Video Analysis System and Interactive IPTV Search System, with breakthroughs in key technologies that are related to these areas. Our team managed to publish 9 important papers, including 4 on top artificial intelligence conference International Joint Conferences on Artificial Intelligence (IJCAI), 1 on another top artificial intelligence conference Twenty-Seventh Conference on Artificial Intelligence (AAAI) and 1 on IEEE Transactions on Multimedia. These outstanding results build a solid ground for the following research work. In the Digital Media Understanding Algorithm Evaluation Platform part, we collected over 2000 hours of video data, making the total amount over 20TB. These data greatly supported our research work. We also participated in organizing several International Algorithm Evaluation Contests, which gained the project a wide-spread reputation. In Intelligent Surveillance Video Analysis System, we studied several topics to improve robustness of the system, including Vehicle Detection and Tracking under Single Camera, Video Fusion and Registration under Multi-Cameras and Spatial-Temporal Feature Extraction algorithms. In Interactive IPTV Search System, our research results are implemented in the backend of Oriental Cable Network (OCN), the Cable TV Service Provider in Shanghai. We are also developing search servers for IPTV systems of China Telecom, which will be brought online early next year as planned.%该研究在数字媒体理解算法验证平台、智能视频监控应用示范和网络互动电视搜索应用示范等方面已取得成果，并在相关核心技术领域取得突破。在数字媒体理解算法验证平台方面，该研究采集了超过2000 h的影视视频数据，主要为影视播出节目内容，累计多媒体数据已经超过20TB，为算法验证平台提供了充足的数据支持。该研究的团队还积极参与组织了国际算法评测活动，并发挥了重要作用。在智能视频监控应用示范方面，针对实际应用场景，该研究深入研究了单摄像机条件下车辆的检测与跟踪、多摄像机条件下融合、跨摄像机条件下对齐算法和时空联合特征提取等算法有效地提升了监控系统的性能。在网络互动电视应用示范方面，该研究成果已在东方有线后台试运营，并在与中国电信IPTV进行了视频搜索后台合作研发，预计明年初进入实际部署阶段。
展开▼
