【作者】 薛召军；

【导师】 万柏坤；

【作者基本信息】 天津大学 ， 生物医学工程， 2007， 博士

【摘要】 步态识别是生物特征识别技术中的一个新兴领域。它旨在根据人们的走路姿势实现对个人身份的识别或生理、病理及心理特征的检测,具有广阔应用前景,成为近年来生物医学信息检测领域备受关注的前沿方向。步态识别主要针对含有人体步行运动图像进行分析,其关键是寻找合适的步态特征及分类方法,融合了计算机视觉、模式识别以及视频/图像序列处理等多种技术。本研究主要采用如下处理流程:先提取目标人体的轮廓信息并将其规格化,经叠加处理后获取步态特征图,然后用小波变换将步态特征图分解,再依据骨架理论和图像空间不变矩提取两种人体模型的步态特征参数,输入至支持向量机(Support Vector Machine,SVM)进行步态识别。分别使用中国科学院自动化研究所(CASIA)和美国南佛罗里达大学(USF)的步态数据库进行了实验,分别取得了93.67%与83%-100%的良好识别率。考虑到不同应用环境,本论文还分析了红外热成像技术用于步态识别的可行性,并自建了天津大学红外步态数据库(Tianjin University Infrared Gait Database,TIGD)。实验表明该技术识别率受人体携带外物影响较小,而仅受衣着影响较大,值得深入研究与应用。作者还尝试设计了一套实用的智能步态识别门禁系统实验平台,具有步态图像实时采集、定位、特征提取与自动分类等功能,可实现对人体目标的安全监控。本研究中的创新性主要体现在:i.首次将步态识别领域中的两种人体模型进行了有机结合,弥补了单一模型存在的缺陷;并将骨架特征参数与不变矩矩参数同时运用于步态识别中,减弱了背景、光照、衣着、速度等因素变化的影响,提高了算法的实用性。ii.首次将小波变换与支持向量机相结合用于步态识别,提高了分类算法的精确性;并在步态特征参数提取中将小波变换与矩理论相结合,有效地提高了基于人体轮廓信息及区域特征的步态识别效果。iii.首次采用红外热成像技术获取步态图像数据,借助于红外成像可夜视、易定量的优势,提升了步态图像序列采集与识别的技术层次,并拓展了其应用领域。iv.构建了基于步态图像特征进行身份识别的门禁系统硬件平台,研究开发了智能门禁系统软件,为步态识别技术的实际应用进行了初步探索。更多还原

【Abstract】 Recognition by gait is a new field for the biometric recognition technology. Its aim is to recognize people or detect physiological, pathological and mental characters by their walk style. The future of this technique will be very good. Gait recognition, as one of the attractive research area of biomedical information detection, attracts more and more attention. Gait recognition mainly analyzes moment images including walk style. For this technique the key factors is to find out the gait characters and classification method. It contains many kinds of techniques, such as computer vision, pattern recognition, video and image sequences processing and etc.This study deals with the following process. Body silhouette sequences of gait were extracted and normalized in this study. The sequences were added together and gait character image could be obtained. The method of wavelet transform was used to decompose the image of gait character. In this paper two models were combined for the body object. At the same time moment invariants and skelecton theory were used in extracting gait character parameters. Then support vector machine (SVM) was applied for classification and recognition. This technique was applied to CASIA and USF gait data-set and achieves probability of correct recognition of 93.67% for the former and 83%-100% for the latter.Considering the different environment, the infrared thermal imaging technology was used to research gait recognition. Infrared imaging camera was presented for the TJU Infrared Gait Database (TIGD). The experiment was performed in this data-set. It is proved that recognition result was insensitive for the person with object such as backpack and ball. For the term of wearing down coat, recognition rate is affected apparently. So we should give attention to the study and application for infrared gait recognition. In this paper, the application of gait recognition was engaged in research. The experiment platform of automatic gait access control system was established. This system can be used to watch body objects. The function involves of real-time image collection, image orientation, feature extraction, recognition and etc.The originalities of this thesis were the followings:i. For the first time, two models of body were united in the filed of gait recognition. This method can repair the deficiency of using single model. Feature parameter of skelecton and moment invariants was applied together in gait recognition. This technique reduces the influence by noises such as background, light, clothes, speed and etc. At the same time it strengthens the application for practice.ii. For the first time, wavelet transform and support vector machine were combined to use recognize gait. It enforces the precision of classification. Moment invariants and wavelet transform were used to extract feature parameter of gait. This method can function as an efficient gait recognition based on body silhouette and area feature.iii. The new concept of using infrared imaging camera in gait recognition developed the attractive research areas. Infrared imaging camera can be used in night and transforms the temperature of body to the image. The level of research was improved in image collection of gait sequences and gait recognition.iv. This new access control system based on gait feature was built. The software of this system was compiled. Elementary work was researched in the application involving gait recognition.更多还原