项目名称: 图像认知中的遮挡影响分析及建模
项目编号: No.61773231
项目类型: 面上项目
立项/批准年度: 2018
项目学科: 其他
项目作者: 马惠敏
作者单位: 清华大学
项目金额: 16万元
中文摘要: 复杂场景图像中的目标检测、识别是计算机视觉的一个难题,遮挡是复杂场景图像中一个普遍存在的现象,也是视觉导航、公共安全视频监控等实际应用无法回避的核心问题,遮挡对图像认知的影响规律及具有抗遮挡能力的认知模型直接关系到计算机视觉的实际应用,是一个迫切需要解决的科学问题。本研究针对图像认知中的遮挡现象展开深入的研究,包括两方面:(1)利用眼动仪对人类视觉的抗遮挡特性分析,提取遮挡对图像认知的影响规律,提出影响程度量化标准;(2)建立一个可定量遮挡属性的典型物体图像库。为复杂场景中的抗遮挡目标检测、识别研究提供基础支撑。
中文关键词: 遮挡规律;抗遮挡能力评估;遮挡图像库;图像认知模型
英文摘要: The task of object detection and recognition in image with complex scene is a difficult problem in computer vision. Occlusion is a widespread phenomenon in a complex scene image, meanwhile the core issues of autopilot practical visual navigation, public safety video surveillance and variety of other practical problem.The influence of occlusion on the image recognition and the cognitive model with anti-blocking ability directly related to the practical application of computer vision, so it's a scientific problem to be solved urgently. In this research, we focus on the image recognition of the phenomenon, including two aspects: 1. Analysing the effectiveness of occlusion to human vision, and extracting the occlusion influence on the image recognition to put forward quantitative criteria of the influence; 2. Proposing a quantitative occlusion image library of typical objects. This research will provide fundamental support to the research on object detection and recognition in complex scenes.
英文关键词: Influence of Occlusion;Assessment of Anti Occlusion Capability;Occlusion Image Library;Model of Image Cognition