面向视频分析的类脑计算模型关键技术研究

This project draws on and simulates the human visual perception mechanism, and studies the key technologies of brain-inspired computational model for video analysis: including saliency detection, semi-supervised learning, transfer learning, spiking neural network, and demonstration of typical applications. A series of innovative algorithms are proposed， such as: saliency detection algorithm based on normalized Graph cut and adaptive region fusion; hierarchical model combined by low-level feature driven saliency detection and high level semantic feature driven target recognition; semi-supervised learning based on Teaching-to-Learn and Learning-to-Teach framework; transfer learning algorithm and semi-supervised spiking neural network suitable for complex video analysis. The results of this project can be applied in many fields, such as security.monitoring, intelligent traffic police, guardianship of the elderly, precision guided missile and so on. Therefore, the research of this project has important theoretical significance and application value. The project team has the good research foundation and working conditions, established close international cooperation and academic exchange. Detailed research contents and technical route have been made. The project team has laid a good foundation for the completion of this project.

本项目借鉴与模仿人脑视觉认知机理，研究面向视频分析的类脑计算模型的关键技术：包括显著性检测、半监督学习、迁移学习、脉冲神经网络、典型应用的演示验证。提出了一系列创新算法，如：面向视频分析的类脑计算模型的框架系统；基于归一化图分割与自适应区域融合的显著性检测算法；结合基于底层低级特征驱动的显著性检测和高层语义特征驱动的目标识别方法的层次化模型；结合导学和导教策略的半监督学习算法；适用于复杂视频分析的迁移学习算法和具有半监督学习能力的脉冲神经网络。本项目成果可应用于安全监控、智能交通的电子警察、老人监护，导弹精确制导等许多领域，因此本项目研究具有重要理论研究意义和应用价值。本项目组已具备很好的研究基础和工作条件，建立了密切的国际合作和学术交流关系，制定了详细的研究内容和技术路线，为高水平完成本项目打下了很好的基础。

本项目借鉴与模仿人脑视觉认知机理，研究面向视频分析的类脑计算模型的关键技术：包括显著性检测、半监督学习、迁移学习、脉冲神经网络、典型应用的演示验证。提出了一系列创新算法，如：面向视频分析的类脑计算模型的框架系统；基于归一化图分割与自适应区域融合的显著性检测算法；结合基于底层低级特征驱动的显著性检测和高层语义特征驱动的目标识别方法的层次化模型；结合导学和导教策略的半监督学习算法；适用于复杂视频分析的迁移学习算法和具有半监督学习能力的脉冲神经网络。本项目成果可应用于安全监控、智能交通的电子警察、老人监护，导弹精确制导等许多领域，因此本项目研究具有重要理论研究意义和应用价值。补充重要成果8888

内容获取失败，请点击重试