【作者】 黄健；

【导师】 葛树志；

【作者基本信息】 电子科技大学 ， 计算机技术（专业学位）， 2013， 硕士

【摘要】 计算机视觉在现代信息化产业中得到了迅猛发展，基于图像内容分析的技术已成功应用在交通车辆分析、工业产品检测、农业病虫害防治、交互式广告机、可视化对讲机和医疗病症检测等各领域各场合。在医疗康复领域，目前主要以机器人辅助康复和人工辅助康复两种方式。机器人辅助康复系统因其造价昂贵、维护成本高、智能性单一在实际的推广应用中市场惨淡；人工辅助康复是对患者康复最佳选择，但现实中待康复的病患对康复教练的需求远大于供方。提供一种具有康复效果同时满足康复病患购买能力和使用方便性的系统是康复训练领域的重要课题。本文针对上述问题，以计算机视觉为解决问题的核心，通过Kinect设备采集病患者的空间深度图，重建病患康复训练的动作流。系统设定两种运动角色:康复教练和康复患者，通过对其示范和训练的动作流的评估反馈给医生病患康复训练的进度报表。另外开发了系统的客户端和WEB版管理端提供给康复训练过程中各角色基于Kinect骨骼采集技术的康复训练模型和流程。具体的主要内容如下：1．研究康复训练的方法，提出了基于Kinect骨骼采集技术的康复方法。该方法通过对计算机视觉中深度图像运动的分析，采集了人体在康复训练时运动的动作流。通过对康复教练和康复病患两种角色的运动采集，建立康复训练的正负样本。2．研究康复训练的办公流，提出了三角色互动康复办公业务模式。该办业务模型将康复训练过程中的三种角色：医生、病患、康复教练融合到训练过程中。并针对其业务功能设计了基于角色权限的管理系统。3．研究康复训练采集时正负动作流相似度评估问题，提出了融合DWT算法和RANSAC算法模型的动作流相似度评估算法，有效的解决了病患训练效果的评估问题，给医生提供了置信度高的输出结果。4．研究康复训练系统中的三维建模和场景构建问题，针对视觉康复系统，动作流可在三维场景中通过动画人物重现，创新性的实现通过多终端异地同场景的康复训练效果。更多还原

【Abstract】 Computer vision in the modern information technology industry has been rapiddevelopment, technical analysis based on the image content analysis has beensuccessfully used in transportation vehicles, the detection of industrial products,agricultural pest control, interactive advertising machine, visual intercom and medicaldisease detection and other fields occasions. In the field of medical rehabilitationrobot-assisted rehabilitation and human support rehabilitation in two ways.Robot-assisted rehabilitation system because of its expensive maintenance and the highcost of a single intelligence in the actual popularization and application of market bleak;assisted rehabilitation is the best choice for the rehabilitation of patients, but in reality tobe rehabilitation patients need rehabilitation coach lofty on the supply side. Provide arehabilitation effect at the same time meet the rehabilitation of patients with the abilityto buy and easy to use system is an important topic in the field of rehabilitation.In this paper, the above problem, computer vision, the core of the problem-solvingand equipment acquisition through Kinect the spatial depth map of the patients, thereconstruction of patient rehabilitation action flow. The system set the two sports role: arehabilitation coach and rehabilitation patients, doctor patient rehabilitation progressreport feedback to the assessment of the stream through its demonstration and trainingactions. Client developed another system and WEB version management end to therehabilitation process, role the Kinect depth acquisition-based rehabilitation model andprocesses. The specific content is as follows:The method of rehabilitation, rehabilitation based on the depth of vision technology.The method through the analysis of the depth of image motion in computer vision,collected from the flow of human movement rehabilitation action. To establishrehabilitation of positive and negative sample collected by the movement of the tworoles of rehabilitation coaches and rehabilitation patients.Rehabilitation of office flow, three roles interact rehabilitation office businessmodel. The office business model to the integration of the three roles in therehabilitation process: doctors, patients, rehabilitation coach to the training process. And role-based permissions management system designed for business functions.Flow similarity assessment study rehabilitation collecting positive and negativeaction, action fusion DWT algorithm and hidden Markov model flow similarityassessment algorithm, an effective solution to the problem of assessment of the effectsof patient training, to provide confidence to the doctor The high output of results.3D modeling and scene study rehabilitation training system to build a system forvisual rehabilitation action flow through animated characters in a3D scene to reproducethe rehabilitation of the same scene, the realization of innovative multi-terminal offsitetraining effect.更多还原