项目名称: 图像统计建模基础理论及其在工业视觉系统中的应用研究
项目编号: No.61272337
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 唐朝晖
作者单位: 中南大学
项目金额: 82万元
中文摘要: 图像统计建模是通过分析图像统计特性来捕捉图像的关键特征,是图像分析、理解和识别的基础。本项目针对工业视觉监控系统中普遍存在的图像感知与识别的难题,通过模拟人类视觉的感知机理,提出基于字典学习和多尺度几何特征表达的图像最优表示方法;基于图像序列分析的多光照估计方法;图像区域几何特征混合统计分割建模方法;仿射不变图像统计描述方法,以解决图像检测、分析和识别中的图像先验模型不明确问题。将所形成的理论与方法应用到泡沫图像视觉监控系统中,实现泡沫图像最优化表示,解决泡沫图像多光照条件下色偏、混合粘连气泡尺寸测量、畸变坍塌严重泡沫图像识别的难题,最终实现浮选生产工况机器鉴别与智能决策,形成工业视觉监控系统中较完善的图像检测、测量、分析、理解和识别的理论和方法体系。本项目的成功实施将极大改进选厂的检测手段和自动化生产水平,为泡沫浮选的优化生产奠定基础,以达到充分利用有限矿产资源,提高企业经济技术水平的目
中文关键词: 泡沫浮选;机器视觉;在线检测;特征选择;工况识别
英文摘要: Image statistical modeling is used to capture the key features of the image by image statistical models,which is the foundation of image analysis,understanding and recognition. In order to get the explicit image statistical models to solve the commonly existed problems of image perception and recognition in vision based industrial monitoring and control system, we propose the optimal image representation method based on dictionary learning and multi-scale geometric features expression according to the human visual perception mechanism, the multi -illumination estimation method by image sequences, the image segmentation statistical modeling approach based on geometric features of mixture image regions, and the invariant image statistical description method. The formation of theories and methods is to be applied to the the visual monitoring and control system for froth flotation to achieve the optimized representation of bubble images , to solve the problem of color variance,to fix the problem of accurate measurement of mixed adhesion bubble size, and to identify the bubble images with serious distortion and collapse. Finally machine identification and intelligent decision on flotation operation conditions is achieved, and a mature architecture of theories and methods for image detection, measurement, analysis, un
英文关键词: Froth flotation;machine vision;online measuring;feature selection;conditions identiflcation