【作者】 孙伟；

【导师】 卢光明；

【作者基本信息】 哈尔滨工业大学 ， 计算机科学与技术， 2013， 硕士

【摘要】 在现有的基于人手的人体生物特征识别系统与方法中，大多数需要把人手伸直放平并放在专用设备上采集图像以做识别。这些系统与方法难以与把手结合，无法进行自然姿态下的人手识别。本文设计的新型生物特征识别系统的采集端设备形状模拟门把手，可以在手握门把手的同时实现识别，采集速度更快，降低了造假可能性，增强了系统安全性，同时采集方式更舒适。系统的主要研发工作包括如下四部分：门把手式识别系统设备的设计与实现、人手图像采集及预处理、人手图像感兴趣区域提取、人手图像特征提取与识别。主要研究内容包括：（1）门把手式识别系统设备的设计与实现。该部分的主要任务如下：设计门把外形的图像采集设备，根据实际应用改进图像采集设备，通过计算选取合适的图像采集器材，根据光学原理设计图像采集方法，设计生物特征识别流程。（2）人手图像采集及预处理。该部分包括设计数据库采集流程及数据采集，根据数据采集时间因素进行图像标定，设计人手图像清晰度评价算法，依据图像采集时间提取人手环状图像。（3）人手图像感兴趣区域提取。该部分主要包括：利用K-means算法迭代求出人手图像提取所需的手部MASK图像，根据人手图像特征设计左右手分辨算法，根据手部MASK图像对采集图像提取感兴趣区域。（4）人手图像特征提取与识别。设计并实现基于竞争编码的人手图像特征提取及匹配方法，并对数据库图像进行各项试验，分析结果并得出结论。门把手式人手生物特征识别系统设计与开发的研究，对生物特征识别系统的发展和应用有一定的促进作用。本文通过初步研究，证明了该方案的可行性，但系统精度还有提升空间，可以从改进采集设备、图像处理和控制模块几个方面对系统做更深的优化和改进。其运行模式自然、方便，用户可在握住门把手时实现自动身份识别，以判定其是否具有门禁准入的权限，可广泛应用于各类门禁系统。更多还原

【Abstract】 Among all of existing systems and methods of biometric identification basedon the human hands, most of them are needed to put the hands on the specialequipment to capture images for identification. These systems are difficult tocombine with the handle equipment, which means that it cannot recognize thenatural state of the hands. The terminal equipment of the feature recognitionsystem that this article describes simulate the shape of the doorknob. The user canachieving recognition while grasping the doorknob. This system has the followingadvantages: saving time, reducing the possibility of fraud, and enhance systemsecurity. The system collects faster, and the ways it collecting is more comfortable.The system includes the following four parts: design and implementation ofthe doorknob identification system equipment, image acquisition andpre-processing, region of interest(ROI) extraction, feature extraction andrecognition. The main contents include:(1) Design and implementation of the doorknob identification systemequipment. The main tasks of this section are described as follows: Design imageacquisition device, improved the acquisition device according to practice, choosethe appropriate image acquisition equipment by calculating the resolution, designimage acquisition methods according to the principles of optics, design the processof the system.(2) Image acquisition and pre-processing. This section includes designdatabase collection system and collecting image data, calibrate the image dataaccording to the time it captured, design the image sharpness evaluation algorithm,extract manually ring images depending on the image acquisition time.(3) Region of interest(ROI) image extraction. This section includes: extractthe desired MASK image based on K-means, get the stable MASK by iterative,designed resolution algorithm according to the left and right hand image features,extract the region of interest area based on the MASK image.(4) Image feature extraction and recognition. Design and implement the imagefeature extraction by Competitive code and matching methods on the handimagines. With the hand imagines database, we do the experiments and analyze the results of the experiments and draw our conclusions.Doorknob hand recognition system described in this paper contribute to thebiometric identification system growth. In this paper, a preliminary studydemonstrates the feasibility of the system, but there is room to improve. Furtherstudy could be carried in improving image acquisition equipment, imagepre-processing and control part of the system. It is easy to use. It achieveautomatic identification when the user hold the doorknob. It then determinewhether the user has access privileges. It can be widely used in various types ofaccess control systems.更多还原