在大数据时代，以图像、视频为代表的视觉数据不仅规模庞大，还具有高并发、高维度、大流量的显著特性，导致了对它们的实时处理和分析面临着巨大的挑战。鉴于此，本项目拟以面向大数据环境的高通量视觉计算为总体目标，围绕高通量视觉计算中任务内在关联性与计算结构并行性的多层次匹配、视觉数据分析中贯通式的高通量综合优化这两个关键科学问题，从高通量视觉计算理论、方法和应用系统三个方面开展研究，具体包括数据和任务的相关性分析、视觉计算的并行理论模型、高通量视觉特征提取算法、高通量聚类与学习算法、以及高通量视觉语义分析方法，并基于上述理论和方法搭建视觉大数据检索和网络特定视觉内容监管这两个高通量视觉计算的典型应用示范系统。

In the big data era, visual data including images and videos not only are huge, but also have the characteristics of high concurrency, high dimension and high throughput, which bring in huge challenges into the real-time analysis and processing of them. Towards this end, this proposal takes the high-throughput visual computing towards big data environments as the overall goal. We will study the high-throughput visual computing theory, the high-throughput visual data analysis methods and the high-throughput visual computing application systems, to address the two key problems, i.e., the multilevel matching between the association of computing tasks and the parallelism of computing architectures, and the high-throughput thorough and comprehensive optimization of the visual data analysis. Specifically, we will investigate the correlation analysis of data and tasks, build the parallel theoretical models of visual computing, and propose novel high-throughput visual feature extraction algorithms, high-throughput clustering and learning methods and high-throughput visual semantic analysis methods. Finally, we will develop the demonstration systems of the big visual data retrieval and the supervision of the specific content on the Internet based on the above theories and methods.