CN101520838A - Automatic-tracking and automatic-zooming method for acquiring iris images - Google Patents

Abstract

Translated fromChinese

本发明一种自动跟踪和自动变焦的虹膜图像获取方法。方法包括如下步骤：将一个宽视角摄像头和一个窄视角(长焦距)摄像头以及红外光源紧密安装在一个可控制云台上，云台带动三者同时旋转；系统通过宽视角摄像头连续检测人脸图像；当有人脸图像的时候，检测人眼图像，得到眼睛的准确位置；然后控制云台上下左右旋转，使窄视角摄像头和红外光源对准人的眼睛；并同时判断用户距离和图像质量，控制摄像头自动变焦；当采集到清晰的虹膜图像时进行图像处理和虹膜识别，判断用户身份。系统装置包括计算机、可控云台、宽视角摄像头、窄视角摄像机、红外光源和电源、图像采集卡等辅助设备。

The invention provides an iris image acquisition method with automatic tracking and automatic zooming. The method comprises the following steps: closely installing a wide viewing angle camera, a narrow viewing angle (long focal length) camera and an infrared light source on a controllable pan-tilt, and the pan-tilt drives the three to rotate simultaneously; the system continuously detects face images through the wide-angle viewing angle camera ;When there is a human face image, detect the human eye image to obtain the exact position of the eye; then control the pan-tilt to rotate up and down, left and right, so that the narrow-view camera and infrared light source are aimed at the human eye; at the same time, judge the user distance and image quality, and control The camera automatically zooms; when a clear iris image is collected, image processing and iris recognition are performed to determine the user's identity. The system device includes computer, controllable pan/tilt, wide-angle camera, narrow-angle camera, infrared light source and power supply, image acquisition card and other auxiliary equipment.

Description

Translated fromChinese

自动跟踪和自动变焦的虹膜图像获取方法Iris image acquisition method with automatic tracking and automatic zooming

技术领域technical field

本发明涉及一种利用自动跟踪和自动对焦的虹膜图像获取方法，是综合了计算机视觉、图像处理、模式识别和自动控制等技术，应用于生物特征识别领域的虹膜识别方法。The invention relates to an iris image acquisition method using automatic tracking and automatic focusing, which is an iris recognition method applied in the field of biological feature recognition by integrating technologies such as computer vision, image processing, pattern recognition and automatic control.

背景技术Background technique

在目前生物特征识别技术中，虹膜识别技术是准确度最高、防伪性最好的技术之一，具有广泛的市场应用前景。虹膜识别的步骤一般包括虹膜图像获取、活体检测、预处理、特征提取和特征匹配等步骤。其中，虹膜图像获取是非常重要的一个环节。然而，由于虹膜的直径非常小，而采集的虹膜图像又要求必须有足够多的像素，所以虹膜图像获取也是非常困难的一个环节。Among the current biometric identification technologies, iris recognition technology is one of the most accurate and anti-counterfeiting technologies, and has broad market application prospects. The steps of iris recognition generally include steps such as iris image acquisition, liveness detection, preprocessing, feature extraction and feature matching. Among them, iris image acquisition is a very important link. However, since the diameter of the iris is very small, and the collected iris image must have enough pixels, the acquisition of the iris image is also a very difficult link.

最早的虹膜图像采集装置是接触式的和固定的，即需要把眼睛紧紧地贴近镜头来采集图像，这样得到的图像质量很好，但只能用于实验室的图像采集，如果用于实际生活，会引起很多用户的抵触心理。The earliest iris image acquisition device is contact and fixed, that is, the eyes need to be tightly close to the lens to collect images. The image quality obtained in this way is very good, but it can only be used for laboratory image acquisition. Life will cause many users to resist.

后来的虹膜图像获取采用了长焦距的摄像头，使人不用贴近镜头就可以进行图像采集，但是由于这种摄像头的视角和景深都比较小，人站的位置靠前靠后靠左靠右都会造成人眼对不准摄像头。必须用户积极的配合才能采集到合适的图像，使用很不方便。The later iris image acquisition adopted a camera with a long focal length, so that people can collect images without getting close to the lens. The human eye cannot focus on the camera. Only the active cooperation of the user can be used to collect suitable images, which is very inconvenient to use.

为了解决这个问题，人们做了很多改进，一种是利用距离传感器测得人的距离，然后利用声音或指示灯提示人前进或后退，一种方法是利用镜子或显示器让人看到自己的图像从而自动调整位置。但这些设备仍然存在很多问题：第一，在有些情况下，用户不愿意去主动配合设备，或者用户不方便靠近仪器。第二，即使用户愿意去主动配合，无论是利用声音提示或者是人自己对准镜子，都必须要人去配合仪器移动，造成了很大的不方便；甚至有些人花很长时间都对不准镜头或掌握不好距离。In order to solve this problem, people have made many improvements. One is to use distance sensors to measure the distance of people, and then use sound or indicator lights to prompt people to move forward or back. One method is to use mirrors or monitors to let people see their own images. This automatically adjusts the position. But there are still many problems in these devices: First, in some cases, users are unwilling to actively cooperate with the devices, or users are inconvenient to approach the instruments. Second, even if the user is willing to take the initiative to cooperate, whether it is to use the voice prompt or to align the mirror by the person himself, the person must cooperate with the movement of the instrument, which causes a lot of inconvenience; some people even spend a long time. The camera is not accurate or the distance is not well grasped.

所以，我们提出了从“人配合机器”到“机器配合人”的主动式虹膜采集方法，即让仪器去主动寻找人的眼睛的位置，无论人站在哪个位置，我们都利用云台的旋转去对准眼睛，并自动变焦来捕捉人的眼睛的图像，这样就扩大了范围，增大了景深，方便了用户。Therefore, we proposed an active iris collection method from "human cooperation with machine" to "machine cooperation with human", that is, let the instrument actively find the position of the human eye, no matter where the person stands, we use the rotation of the pan/tilt To align the eyes, and automatically zoom to capture images of people's eyes, thus expanding the scope and increasing the depth of field, which is convenient for users.

模式识别技术使得在图像中检测人脸成为可能，而市场现有的长焦距(大于75mm)镜头使大范围(0.3m-1m米)拍摄虹膜(虹膜直径大于150像素)成为可能，可控云台使得快速跟踪成为可能，这些都为主动式虹膜采集方法提供了条件。Pattern recognition technology makes it possible to detect human faces in images, and the existing long focal length (greater than 75mm) lens on the market makes it possible to shoot irises (iris diameter greater than 150 pixels) in a wide range (0.3m-1m meters), controllable cloud The platform makes fast tracking possible, which provides the conditions for the active iris collection method.

发明内容Contents of the invention

本发明的目的是公开一种自动跟踪和自动变焦的虹膜图像获取方法，以实现主动寻找用户的眼睛，进行虹膜识别。The purpose of the present invention is to disclose an iris image acquisition method with automatic tracking and automatic zooming, so as to realize actively searching for the user's eyes and performing iris recognition.

为达到上述目的，本发明的技术解决方案是：For achieving the above object, technical solution of the present invention is:

一种基于自动跟踪和自动变焦的虹膜图像获取方法，其特征在于，包括如下步骤：A kind of iris image acquisition method based on automatic tracking and automatic zooming, it is characterized in that, comprises the steps:

1)将一个宽视角摄像头、一个窄视角摄像头和一红外光源安装在可控云台上；1) Install a wide viewing angle camera, a narrow viewing angle camera and an infrared light source on the controllable pan/tilt;

2)利用宽视角摄像头连续拍摄图像，检测是否有人脸图像，当检测到人脸图像时，在人脸的上半区域进行人眼检测，获得眼睛的准确位置；2) Use the wide-angle camera to continuously shoot images to detect whether there is a human face image. When the human face image is detected, perform human eye detection in the upper half of the face to obtain the exact position of the eye;

3)根据眼睛在宽视角摄像头图像中的位置，旋转云台直到窄视角摄像头和红外光源对准人的眼睛，同时使窄视角摄像机中有人的眼睛图像；3) According to the position of the eyes in the wide-angle camera image, rotate the pan-tilt until the narrow-angle camera and the infrared light source are aimed at the eyes of the person, and at the same time make the person's eye image in the narrow-angle camera;

4)判断用户距离和图像质量，控制窄视角摄像头自动变焦和自动对焦；4) Judging the user distance and image quality, controlling the automatic zoom and automatic focus of the narrow viewing angle camera;

5)窄视角摄像机中有人的眼睛图像并对焦后，窄视角摄像头连续拍摄虹膜图像；5) After the person's eye image in the narrow-view camera is focused, the narrow-view camera continuously captures iris images;

6)判断是否采集到清晰的虹膜图像，如果有清晰的图像则进行图像处理和虹膜识别，判断用户身份。6) Judging whether a clear iris image is collected, if there is a clear image, image processing and iris recognition are performed to determine the user's identity.

所述的方法，其所述第1)步中，将一个宽视角摄像头、一个窄视角摄像头和一红外光源紧密安装在可控云台上，两摄像头角度和光源平行安装，但不共线；云台带动三者共同旋转。Described method, in its described 1st) step, a camera with a wide angle of view, a camera with a narrow angle of view and an infrared light source are closely installed on the controllable platform, and the angles of the two cameras are installed in parallel with the light source, but not collinear; The gimbal drives the three to rotate together.

所述的方法，其所述第2)步中，利用宽视角摄像头检测人脸，其步骤包括：Described method, its described 2nd) in the step, utilize wide viewing angle camera to detect people's face, its step comprises:

a)在系统运行时，利用宽视角摄像头连续拍摄图像；a) When the system is running, use the wide-view camera to continuously capture images;

b)利用已有方法训练好的分类器检测图像中是否有人脸图像，有人脸图像时，获取人脸的坐标和大小。b) Use the classifier trained by the existing method to detect whether there is a human face image in the image, and when there is a human face image, obtain the coordinates and size of the human face.

所述的方法，其所述第2)步中，在人脸图像中进行人眼检测，其步骤包括：Described method, its described 2) in step, carry out human eye detection in face image, and its steps include:

a)根据已检测到的人脸位置，估算人的眼睛的大体位置，a) Estimate the general position of the eyes of the person according to the detected face position,

b)利用已有方法训练好的人眼分类器在这个区域检测是否有人的眼睛，有人眼时，获取人眼的坐标和大小。b) Use the human eye classifier trained by the existing method to detect whether there are human eyes in this area, and if there are human eyes, obtain the coordinates and size of the human eyes.

所述的方法，其所述第3)步中，根据眼睛在宽视角摄像头的图像位置来转动云台，其具体方法如下：Described method, its described 3rd) in the step, rotate the cloud platform according to the image position of eyes at wide viewing angle camera, its concrete method is as follows:

a)计算眼睛是否出现在宽视角摄像头的一个标定位置：如果在标定位置，则说明窄视角摄像头已对准眼睛，如果不在，说明需要旋转云台；a) Calculate whether the eyes appear at a calibrated position of the wide-view camera: if it is at the calibrated position, it means that the narrow-view camera is aligned with the eyes; if not, it means that the gimbal needs to be rotated;

b)当眼睛位置偏离标定位置时，转动云台使之回到标定位置：当眼睛位置偏左或偏右时，计算旋转时间，然后则旋转云台向右或向左旋转此时间；当眼睛位置偏上或偏下时，计算旋转时间，然后则旋转云台向下或向上旋转此时间；旋转直到眼睛出现在标定点。b) When the eye position deviates from the calibrated position, turn the gimbal to return to the calibrated position: when the eye position is left or right, calculate the rotation time, and then rotate the gimbal to the right or left for this time; When the position is up or down, calculate the rotation time, and then rotate the gimbal down or up for this time; rotate until the eye appears at the calibration point.

所述的方法，其所述标定位置，是根据用户的大致距离来线性计算，具体方法如下：In the method, the calibration position is linearly calculated according to the approximate distance of the user, and the specific method is as follows:

a)标定位置大约在宽摄像头中心点的下方Δh处，和人距离摄像头的距离D成线性关系，同时和检测到的人脸的大致长度H成线性关系，即Δh＝aH+b，其中a、b为待计算参数；a) The calibration position is about Δh below the center point of the wide camera, which is linearly related to the distance D between the person and the camera, and is linearly related to the approximate length H of the detected face, that is, Δh=aH+b, where a , b is the parameter to be calculated;

b)在准备工作中，分别在用户距离摄像头的不同距离得出若干不同的标定位置Δh1，Δh2，…Δhn，和相对应的不同人脸大致长度值H1，H2，…Hn，令，$H=\left[\begin{array}{c}{H}_1,1\\ H_2,1\\ ...\\ H_n,1\end{array}\right]$，Δh＝[Δh₁，Δh₂，...，Δh_n]^T，然后做线性回归计算，即得出$\left[\begin{array}{c}a\\ b\end{array}\right]={\left(H^{T}H\right)}^{-1}\left(H^T\mathrm{\Δh}\right);$b) In the preparation work, a number of different calibration positions Δh1, Δh2, ... Δhn, and corresponding approximate face length values H1, H2, ... Hn are obtained at different distances from the user to the camera. Let,$h=\left[\begin{array}{c}{h}_1,1\\ h_2,1\\ ...\\ h_{\mathrm{no}},1\end{array}\right]$ , Δh=[Δh₁ , Δh₂ ,..., Δh_n ]^T , and then do linear regression calculation, that is,$\left[\begin{array}{c}a\\ b\end{array}\right]={\left(h^{T}h\right)}^{-1}\left(h^T\mathrm{\Δh}\right);$

c)在系统运行时，通过每次检测人脸得到的人脸大致长度H，计算标定位置，然后对比此位置的坐标和检测到人眼的坐标，判断眼睛是否出现在标定位置。c) When the system is running, the calibration position is calculated by the approximate length H of the human face obtained by detecting the human face each time, and then the coordinates of this position are compared with the coordinates of the detected human eyes to determine whether the eyes appear at the calibration position.

所述的方法，其所述旋转时间的计算，是通过经验判断，具体方法如下：Described method, the calculation of its described rotation time is judged by experience, and specific method is as follows:

a)在准备工作中，分别设定不同的旋转时间，让云台向水平和垂直方向分别旋转，得到每次旋转眼睛位置的变化距离；a) In the preparation work, set different rotation times respectively, let the gimbal rotate horizontally and vertically respectively, and obtain the change distance of the eye position for each rotation;

b)将这些经验数据列成表格，记录到计算机程序中；b) tabulate these empirical data and record them in the computer program;

c)在系统运行中，查询应该水平和垂直旋转的距离，程序查询此表格，既得出需要向水平和垂直旋转的时间。c) When the system is running, query the distance that should be rotated horizontally and vertically, and the program queries this table to obtain the time required to rotate horizontally and vertically.

所述的方法，其所述第4)步中，控制窄视角摄像头自动变焦和自动对焦，其具体方式包括：Described method, its described 4th) in the step, control narrow viewing angle camera auto-zoom and auto-focus, its specific method comprises:

1)根据人脸的大小粗估用户的距离；1) Roughly estimate the distance of the user according to the size of the face;

2)根据大致估计的距离调整变焦到预先设置的焦距，距离远则变到长焦距，距离近则变到短焦距；2) Adjust the zoom to the pre-set focal length according to the roughly estimated distance. If the distance is far away, it will be changed to a long focal length, and if the distance is short, it will be changed to a short focal length;

3)判断虹膜图像的高频分量，然后控制镜头对焦变远或变近，如果变焦高频分量变大，则继续原方式对焦，反之，如果变焦高频分量变小，向反方向对焦；3) Determine the high-frequency component of the iris image, and then control the lens to focus farther or closer. If the high-frequency component of the zoom becomes larger, continue to focus in the original way; otherwise, if the high-frequency component of the zoom becomes smaller, focus in the opposite direction;

4)直到采集到清晰的虹膜图像为止，其中图像清晰的判断是通过系统试验中得到的高频分量经验值确定。4) Until a clear iris image is collected, the clear judgment of the image is determined by the empirical value of the high-frequency component obtained in the system test.

所述的方法，其所述第6)步中采集到清晰虹膜图像后需要进行虹膜识别，虹膜识别方法，包括步骤如下：Described method needs to carry out iris recognition after collecting clear iris image in its described 6th) step, and iris recognition method comprises steps as follows:

a)进行图像预处理和归一化；a) Perform image preprocessing and normalization;

b)进行虹膜定位和虹膜特征提取；b) performing iris location and iris feature extraction;

c)将采集的虹膜特征与数据库中的虹膜特征进行比对，看是否与数据库中的虹膜图像吻合；c) comparing the collected iris feature with the iris feature in the database to see if it matches the iris image in the database;

d)如果与数据库中的虹膜图像吻合，则自动读出该人的姓名，保存识别成功的图像。d) If it matches the iris image in the database, the person's name is automatically read out, and the image with successful recognition is saved.

所述的方法，其当第5)步虹膜识别完成后进行声音提示、灯光提示、或依据比对结果进行开或不开门的操作。Said method, it performs voice prompt, light prompt, or performs the operation of opening or not opening the door according to the comparison result after the iris recognition in step 5).

一种所述的方法使用的系统，其包括可控云台，宽视角摄像头，窄视角摄像头，红外光源，计算机，电源，音箱；一宽视角摄像头和一窄视角摄像头以及红外光源顺序叠置，安装在可控云台的垂直旋转轴上；A system used in the described method comprises a controllable cloud platform, a wide viewing angle camera, a narrow viewing angle camera, an infrared light source, a computer, a power supply, and a sound box; a wide viewing angle camera and a narrow viewing angle camera and an infrared light source are sequentially stacked, Installed on the vertical rotation axis of the controllable head;

其中，各个设备的选型和连接：Among them, the selection and connection of each device:

a)宽视角摄像头选用普通的会议摄像头，通过USB接口与计算机连接；a) The wide viewing angle camera is an ordinary conference camera, connected to the computer through the USB interface;

b)窄视角摄像头选用改装的一体化摄像机，其视频数据通过数据采集卡和计算机连接，一体化摄像头以通过RS-232总线控制对焦和变焦；b) The narrow viewing angle camera is a refitted integrated camera, whose video data is connected to the computer through a data acquisition card, and the integrated camera controls focus and zoom through the RS-232 bus;

c)可控云台选用100度/秒的快速数字化云台，通过RS-485接口和计算机连接，由标准化的Pelco-D协议控制；c) The controllable pan-tilt selects a fast digital pan-tilt with a speed of 100 degrees per second, connects with the computer through the RS-485 interface, and is controlled by the standardized Pelco-D protocol;

d)红外光源选用LED灯阵列，由多个发射角度为15度的红外LED构成。d) The infrared light source is an LED lamp array, which is composed of multiple infrared LEDs with an emission angle of 15 degrees.

所述的系统，其所述计算机硬盘中拷有软件：Described system, software is copied in its described computer hard disk:

a)软件开启两个线程，分别处理宽视角摄像头和窄视角摄像头的图像数据，并在后台进行图像处理、运动控制和数据库检索功能；a) The software starts two threads to process the image data of the wide-view camera and the narrow-view camera respectively, and performs image processing, motion control and database retrieval functions in the background;

b)软件的主窗口有两个画面，分别显示人脸图像和虹膜图像；用户识别成功后，显示用户信息，并发出用户姓名的声音；主窗口上有针对用户管理的注册、识别按钮。b) The main window of the software has two screens, displaying the face image and the iris image respectively; after the user is successfully identified, the user information is displayed, and the voice of the user's name is emitted; there are registration and identification buttons for user management on the main window.

本发明应用了计算机视觉技术、模式识别技术、和运动控制技术来辅助虹膜图像采集，使虹膜识别可以在很大的范围内进行，并且不需要人主动配合去对准镜头，从而大大提高了效率，扩展了虹膜识别的应用范围。The invention uses computer vision technology, pattern recognition technology, and motion control technology to assist iris image acquisition, so that iris recognition can be carried out in a large range, and no one needs to actively cooperate to align the lens, thereby greatly improving the efficiency , expanding the application range of iris recognition.

附图说明Description of drawings

图1为本发明方法的基本步骤示意框图；Fig. 1 is a schematic block diagram of the basic steps of the inventive method;

图2为本发明方法所使用的系统硬件设备示意图；Fig. 2 is a schematic diagram of system hardware equipment used by the inventive method;

图3为两个摄像头的定标示意图；Figure 3 is a schematic diagram of the calibration of two cameras;

图4为控制云台转动到理想位置的实例示意图；Fig. 4 is a schematic diagram of an example of controlling the pan tilt to rotate to an ideal position;

图5为本发明方法所使用的系统软件流程示意图。Fig. 5 is a schematic flowchart of the system software used in the method of the present invention.

具体实施方式Detailed ways

本发明的一种自动跟踪和自动变焦的虹膜图像获取方法，包括如下步骤：将一个宽视角摄像头和一个窄视角摄像头以及红外光源共同安装在可控云台上；系统通过一个宽视角摄像头连续检测人脸图像，当检测到人脸图像的时候，在人脸的大致部位进行人眼检测，获得眼睛的准确位置；控制云台旋转窄视角摄像头和红外光源，对准人的眼睛；同时判断用户位置和图像质量，并进行变焦和对焦；当采集到清晰的虹膜图像时进行图像处理和虹膜识别，判断用户身份。从而实现主动寻找用户的眼睛，进行虹膜识别。An iris image acquisition method with automatic tracking and automatic zooming of the present invention comprises the following steps: a wide viewing angle camera, a narrow viewing angle camera and an infrared light source are jointly installed on a controllable pan/tilt; the system continuously detects Face image, when the face image is detected, human eye detection is performed on the approximate part of the face to obtain the exact position of the eyes; the pan-tilt is controlled to rotate the narrow-view camera and the infrared light source to aim at the human eyes; at the same time, the user is judged Position and image quality, and zoom and focus; when a clear iris image is collected, image processing and iris recognition are performed to determine the user's identity. In this way, the user's eyes can be actively searched for iris recognition.

本发明方法使用的是利用云台自动跟踪的虹膜图像采集系统，该系统由以下设备构成：计算机，可控云台，宽视角摄像头，窄视角摄像机，红外光源和电源、图像采集卡等辅助设备。What the inventive method used is to utilize the iris image acquisition system of pan-tilt automatic tracking, and this system is made of following equipment: computer, controllable pan-tilt, wide viewing angle camera, narrow viewing angle camera, auxiliary equipments such as infrared light source and power supply, image acquisition card .

(一)系统运行的基本步骤如图1：(1) The basic steps of system operation are shown in Figure 1:

S1：将一个宽视角摄像头和一个窄视角摄像头以及红外光源共同安装在可控云台上；S1: Install a wide viewing angle camera, a narrow viewing angle camera and an infrared light source on the controllable pan/tilt;

S2：系统利用宽视角摄像头连续拍摄图像，检测是否有人脸图像；当检测到人脸图像时，在人脸的大致部位进行人眼检测，获得人眼睛的准确位置；S2: The system uses the wide-angle camera to continuously shoot images to detect whether there is a human face image; when the human face image is detected, the human eye detection is performed on the approximate part of the human face to obtain the accurate position of the human eye;

S3：根据眼睛在宽视角摄像头图像中的位置，旋转云台，使窄视角摄像机和红外光源都对准人的眼睛；S3: According to the position of the eyes in the image of the wide-angle camera, rotate the gimbal so that both the narrow-angle camera and the infrared light source are aimed at the eyes of the person;

S4：根据人脸的大小判断用户的距离，控制自动变焦；根据图像的高频分量大小估计图像质量，控制自动对焦；S4: Judge the distance of the user according to the size of the face, and control the automatic zoom; estimate the image quality according to the size of the high-frequency component of the image, and control the automatic focus;

S5：判断是否采集到清晰的虹膜图像，如果有清晰的图像则进行图像处理和虹膜识别，判断用户身份。S5: Determine whether a clear iris image is collected, and if there is a clear image, perform image processing and iris recognition to determine the identity of the user.

(二)系统硬件设备的组成和连接如图2，其中，可控云台1，宽视角摄像头2，窄视角摄像头3，红外光源4，计算机5，执行装置6；设备捕捉到的图像如下：用户7；宽视角摄像头拍摄的图像8；利用人脸检测算法检测出的人脸区域9；窄视角摄像头拍摄的人眼图像10。(2) The composition and connection of the system hardware equipment is shown in Figure 2, wherein, the controllable pan/tilt 1, the wide viewing angle camera 2, the narrowviewing angle camera 3, the infraredlight source 4, thecomputer 5, and theexecution device 6; the images captured by the equipment are as follows: Auser 7; animage 8 captured by a wide-angle camera; a human face area 9 detected by a face detection algorithm; aneye image 10 captured by a narrow-angle camera.

设备的选型如下：可控云台为水平垂直旋转速度大于100度/秒的云台；宽视角摄像头为普通的USB会议摄像头；窄视角摄像头选取改装后的一体化摄像机，最大焦距大于75mm，可以自动对焦、自动变焦；红外光源为红外LED阵列，波长850nm，发射角15度。The selection of equipment is as follows: the controllable pan-tilt is a pan-tilt with a horizontal and vertical rotation speed greater than 100 degrees per second; the wide-angle camera is an ordinary USB conference camera; the narrow-angle camera is a modified integrated camera with a maximum focal length greater than 75mm. It can auto-focus and auto-zoom; the infrared light source is an infrared LED array with a wavelength of 850nm and an emission angle of 15 degrees.

设备的连接方式如下：可控云台通过RS-485总线和计算机连接；宽视角摄像头为通过USB和计算机连接；一体化摄像机通过数字图像采集卡和计算机连接；两个摄像头紧紧靠到一起，基本上共同安装；可控云台、摄像头和红外光源都有电源供电。The connection method of the equipment is as follows: the controllable pan-tilt is connected to the computer through the RS-485 bus; the wide-angle camera is connected to the computer through USB; the integrated camera is connected to the computer through a digital image acquisition card; the two cameras are closely connected together, Basically, they are installed together; the controllable pan/tilt, camera and infrared light source all have power supply.

(三)系统运行中的具体方法(3) Specific methods in system operation

1，人脸检测和人眼检测的方法1. Methods of face detection and human eye detection

人脸检测的分类有多种方法，我们采用haar-like特征，并用boosting的算法进行训练，一个320*240的图像的人脸检测的平均时间在几十毫秒，可以满足时间要求。人眼检测和人脸检测的方法基本相似，但人眼检测是在人脸图像的上半部分进行查找，所以查找范围很小，检测时间也相对很少。There are many ways to classify face detection. We use haar-like features and use boosting algorithm for training. The average time for face detection of a 320*240 image is tens of milliseconds, which can meet the time requirements. The methods of human eye detection and face detection are basically similar, but the human eye detection is searched in the upper half of the face image, so the search range is small, and the detection time is relatively small.

2，两个摄像头的定标方法如图3：2. The calibration method of the two cameras is shown in Figure 3:

可以看到，由于两个摄像头安装在一起，所以窄视角摄像头的图像和宽视角摄像头的图像会有对应的关系：当目标点(眼睛)图像出现在宽视摄像头图像中心点(11)垂直偏离△h的位置(12)，就会出现在窄视角摄像头的图像区域(10)内。It can be seen that since the two cameras are installed together, there will be a corresponding relationship between the image of the narrow-view camera and the image of the wide-view camera: when the image of the target point (eye) appears at the center point (11) of the image of the wide-view camera and deviates vertically The position (12) of Δh will appear in the image area (10) of the narrow viewing angle camera.

系统运行时，即按照如上原理，当在宽视角摄像头找到了眼镜后，转动云台，使得眼睛的位置出现在宽视角摄像头标定位置(12)，就可以使眼睛出现在窄视角摄像头的图像(10)中。When the system is running, according to the above principle, when the glasses are found by the wide-angle camera, the pan-tilt is rotated so that the position of the eyes appears at the calibration position (12) of the wide-view camera, and the eyes can appear on the image of the narrow-view camera ( 10).

其中，△h会随着人和相机的距离D变化，人越远△h会越小，人越近△h会越大。Among them, △h will change with the distance D between the person and the camera. The farther the person is, the smaller △h will be, and the closer the person will be, the larger △h will be.

假设标定位置Δh和人距离摄像头的位置D成线性关系，而D又与检测到的人脸高度H成线性关系。所以Δh和H之间也假设成线性关系，即Δh＝aH+b。我们在实验中分别在人距离摄像头不同位置D1，D2，…Dn时，得出不同的标定位置Δh1，Δh2，…Δhn。然后由多组实验值做线性回归，$\left[\begin{array}{c}a\\ b\end{array}\right]={\left(H^{T}H\right)}^{-1}\left(H^T\mathrm{\Δh}\right),$其中，$H=\left[\begin{array}{c}{H}_1,1\\ H_2,1\\ ...\\ H_n,1\end{array}\right],$Δh＝[Δh₁，Δh₂，...，Δh_n]^T。即可得到a，b的经验直。Assume that the calibration position Δh is linearly related to the position D of the person from the camera, and D is linearly related to the detected face height H. Therefore, a linear relationship is also assumed between Δh and H, that is, Δh=aH+b. In the experiment, we obtained different calibration positions Δh1, Δh2, ... Δhn when the distance between the person and the camera is different at positions D1, D2, ... Dn. Then do linear regression from multiple sets of experimental values,$\left[\begin{array}{c}a\\ b\end{array}\right]={\left(h^{T}h\right)}^{-1}\left(h^T\mathrm{\Δh}\right),$ in,$h=\left[\begin{array}{c}{h}_1,1\\ h_2,1\\ ...\\ h_{\mathrm{no}},1\end{array}\right],$ Δh=[Δh₁ , Δh₂ , . . . , Δh_n ]^T . You can get a, b empirical straight.

3，控制云台转动摄像头的方法如图4：3. The method of controlling the pan-tilt to rotate the camera is shown in Figure 4:

在宽视角摄像头的图像中，通过人脸检测和眼睛检测，得出眼睛的实际位置和标定位置的偏差，然后根据此偏差来旋转云台。图4是旋转的一个实例，描述了一个过程，大的图像为宽视角摄像头的图像，左上角小的图像为窄视角摄像头的图像，其中(12)为标定位置，(11)为中心点，当眼睛的位置移动到宽视角摄像头的标定点，也同时出现在窄视角摄像头区域(10)中。In the image of the wide-angle camera, through face detection and eye detection, the deviation between the actual position of the eyes and the calibrated position is obtained, and then the gimbal is rotated according to this deviation. Figure 4 is an example of rotation, which describes a process. The large image is the image of the wide-angle camera, and the small image in the upper left corner is the image of the narrow-angle camera, where (12) is the calibration position, (11) is the center point, When the position of the eye moves to the calibration point of the wide viewing angle camera, it also appears in the narrow viewing angle camera area (10) at the same time.

为了让云台旋转更快，我们采用了查找经验表的方法：眼睛位置偏差标定点越大，需要设置旋转速度和旋转时间越大；偏差越小，需要设置的旋转时间越小。偏差值和旋转时间之间的关系并不是线性关系，但可以通过实验来确定一个经验的查找表。例如，偏差值为5个像素，通过实验确定，设定旋转时间为300ms；偏差值为10个像素，通过实验确定，设定旋转时间为400ms。In order to make the gimbal rotate faster, we use the method of looking up the experience table: the larger the eye position deviation calibration point, the greater the rotation speed and rotation time that need to be set; the smaller the deviation, the smaller the rotation time that needs to be set. The relationship between offset and spin time is not linear, but an empirical look-up table can be determined experimentally. For example, if the deviation value is 5 pixels, it is determined through experiments, and the rotation time is set to 300ms; if the deviation value is 10 pixels, it is determined through experiments, and the rotation time is set to 400ms.

例如，人的眼睛位于宽视角摄像头标定点向左5个像素，则控制云台向右旋转300ms；人的眼睛位于标定点向上10个像素，则控制云台向下旋转400ms。如果一次没有旋转到预定位置，则在下一个循环中再次旋转，直到眼睛的位置到达标定点。For example, if the human eye is located 5 pixels to the left of the calibration point of the wide-angle camera, the gimbal is controlled to rotate to the right for 300ms; if the human eye is located 10 pixels above the calibration point, the gimbal is controlled to rotate downward for 400ms. If it does not rotate to the predetermined position once, it will rotate again in the next cycle until the position of the eyes reaches the calibration point.

4，控制变焦和对焦的方法：4. How to control zoom and focus:

设置几个变焦的档，当用户在较远距离时，调到长的焦距，当用户在近距离时，调到短的焦距，当用户在中间位置时，调到中间的焦距。这样可以保证无论用户无论在远处或近处，都可以使其虹膜的图像足够大。Set several zoom levels, when the user is at a long distance, adjust to a long focal length, when the user is at a short distance, adjust to a short focal length, and when the user is at a middle position, adjust to an intermediate focal length. This can ensure that no matter whether the user is far or near, the image of the iris can be made large enough.

其中用户的距离可以根据宽视角图像中人脸的大小来估计。因为检测到的人脸框的大小是和用户距离成反比例的，所以可以粗略估计出用户距离。The user distance can be estimated according to the size of the face in the wide-view image. Because the size of the detected face frame is inversely proportional to the user distance, the user distance can be roughly estimated.

变焦是为了保证虹膜在图像中的大小能够满足模式识别的要求。除了变焦之外，必须同时进行自动对焦，以保证能够得到清晰的虹膜图像。Zooming is to ensure that the size of the iris in the image can meet the requirements of pattern recognition. In addition to zooming, autofocus must be performed at the same time to ensure a clear image of the iris.

判断图像的高频分量，然后控制镜头变焦远或变近。如果变焦高频分量变大，则继续变焦，如果高频分量变小，则向反方向控制变焦，直到图像的高频分量最大的时候。Judge the high-frequency components of the image, and then control the lens to zoom far or close. If the high-frequency component of the zoom becomes larger, continue zooming, and if the high-frequency component becomes smaller, control the zoom in the opposite direction until the high-frequency component of the image is at its maximum.

摄像头(包括镜头)通过RS-232总线和计算机连接，可以通过Pelco-D协议控制变焦和对焦，同时，摄像头本身也带有利用视频分量驱动自动对焦的功能。The camera (including the lens) is connected to the computer through the RS-232 bus, and the zoom and focus can be controlled through the Pelco-D protocol. At the same time, the camera itself also has the function of using the video component to drive the auto focus.

(四)系统的软件(4) System software

软件有良好的交互界面：主窗口内有两个区域，一个区域显示宽视角摄像头的图像，一个区域显示窄视角摄像头的图像；当跟踪到人的眼睛之后，前者显示人脸的图像，后者显示红外光下人眼(虹膜)的图像；识别成功后，软件可以显示识别出识别用户的姓名；窗口有“注册”、“识别”等功能键，方便管理The software has a good interactive interface: there are two areas in the main window, one area displays the image of the wide-angle camera, and the other area displays the image of the narrow-angle camera; when the person's eyes are tracked, the former displays the image of the face, and the latter Display the image of the human eye (iris) under infrared light; after the recognition is successful, the software can display the name of the recognized user; the window has function keys such as "registration" and "recognition", which are convenient for management

图5为系统软件流程示意图。在计算机的程序内，初始化后开启两个线程，一个线程是用宽视角摄像头连续拍摄图像，寻找人脸，控制云台旋转；另一个线程是用窄视角摄像头连续拍摄图像，进行虹膜识别。Figure 5 is a schematic diagram of the system software flow. In the computer program, two threads are opened after initialization. One thread is to continuously capture images with a wide-angle camera, find faces, and control the rotation of the pan/tilt; the other thread is to continuously capture images with a narrow-angle camera for iris recognition.

P0：系统初始化。系统初始化时有许多工作要做，包括：装载已经训练好的人脸分类器、人眼分类器；打开虹膜特征数据库；打开云台控制接口；打开红外光源等。系统初始化完毕后，开启两个线程。P0: System initialization. There is a lot of work to be done when the system is initialized, including: loading the trained face classifier and human eye classifier; opening the iris feature database; opening the PTZ control interface; opening the infrared light source, etc. After the system is initialized, start two threads.

P1-1：宽视角摄像头拍摄一桢图像，并在窗口中显示。P1-1: The wide-angle camera captures a frame of image and displays it in the window.

P1-2：在图像中利用已训练好的分类器来检测是否有人脸的图像。P1-2: Use the trained classifier in the image to detect whether there is a human face in the image.

P1-3：如果没有人脸，则继续检测，如果有，则获得人脸在图像中的位置和大小的参数信息P1-3: If there is no face, continue to detect, if there is, obtain the parameter information of the position and size of the face in the image

P1-4：在人脸的上半部分区域进行人眼检测，人眼检测仍然依靠已训练好的分类器，得到两只眼睛中心位置的准确坐标值。P1-4: Human eye detection is performed on the upper half of the face. Human eye detection still relies on a trained classifier to obtain accurate coordinates of the center positions of the two eyes.

P1-5：根据人眼的坐标和中心位置的偏差，控制云台旋转到标定位置。(当偏离人脸左边时，则对准左眼；当偏离人脸右边时，则对准右眼。如果偏差在允许范围内，则停止旋转。P1-5: Control the rotation of the gimbal to the calibration position according to the deviation between the coordinates of the human eye and the center position. (When it deviates from the left side of the face, it aligns with the left eye; when it deviates from the right side of the face, it aligns with the right eye. If the deviation is within the allowable range, stop the rotation.

P1-6：根据人脸的大小粗略估计用户的距离，然后根据此距离改变摄像头的焦距。并同时打开摄像头的自动对焦功能。完毕后返回P1-1，继续跟踪人脸图像。P1-6: Roughly estimate the distance of the user according to the size of the face, and then change the focal length of the camera according to this distance. And at the same time turn on the auto focus function of the camera. Return to P1-1 after completion, and continue to track the face image.

P2-1：窄视角摄像头拍摄一桢图像，并在窗口中显示。P2-1: The narrow view camera takes a frame of image and displays it in the window.

P2-2：检测是否有清晰的虹膜图像。如果没有，说明云台旋转还没有结束，继续采集图像。P2-2: Detect whether there is a clear iris image. If not, it means that the rotation of the gimbal has not ended yet, continue to collect images.

P2-3：如果采集到图像。虹膜识别，虹膜识别算法采用中国科学院自动化所的虹膜识别SDK软件包irismatching.dll，进行虹膜图像的预处理、虹膜定位、特征提取、特征比对等步骤。P2-3: If the image is collected. Iris recognition, the iris recognition algorithm adopts the iris recognition SDK software package irismatching.dll of the Institute of Automation, Chinese Academy of Sciences, and performs steps such as iris image preprocessing, iris positioning, feature extraction, and feature comparison.

P2-4：如果虹膜识别成功，同时声音提示成功，执行开关通道、打开指示灯等动作；如果不成功则继续识别，连续三次不成功，则进行报警提示。P2-4: If the iris recognition is successful, at the same time the sound prompts successfully, perform actions such as switching the channel and turning on the indicator light; if it fails, continue to recognize, and if it fails three times in a row, it will give an alarm prompt.

本发明设计了了一种自动跟踪和自动变焦的虹膜图像获取方法。经过实验，一般情况下，对于距离0.4米到1米，水平30度角内，高度在1.55至1.90米用户，平均都可以在5秒钟内完成图像采集和识别。The invention designs an iris image acquisition method with automatic tracking and automatic zooming. After experiments, in general, for users with a distance of 0.4 meters to 1 meter, a horizontal angle of 30 degrees, and a height of 1.55 to 1.90 meters, image acquisition and recognition can be completed within 5 seconds on average.

本方法的优点在于以下几个方面：The advantages of this method lie in the following aspects:

1.本方法可以在较远距离进行图像采集，适应性更广。1. This method can carry out image acquisition at a relatively long distance, and has wider adaptability.

2.本方法可以主动跟踪人的眼睛，扩大了采集图像的范围。2. The method can actively track people's eyes and expand the range of image collection.

3.本方法可以自动变焦和对焦，扩大了采集图像的景深。3. The method can automatically zoom and focus, expanding the depth of field of the collected images.

4.本方法可以适应各种不同类型的人群，任何人都在几秒钟内自动完成识别，不会出现无法对准或无法对焦所带来的问题，大大方便了用户。4. This method can be adapted to various types of people. Anyone can automatically complete the identification within a few seconds, and there will be no problems caused by inability to align or focus, which greatly facilitates users.

综上所述，本发明对于目前的虹膜识别系统是一个很大的改进，能够在更远的距离进行图像采集，并且不需要人的配合就可以主动跟踪人的眼睛并自动变焦，从而大大地提高了虹膜识别系统的应用范围，方便了用户的使用，具有很大的市场潜力。In summary, the present invention is a great improvement to the current iris recognition system, which can carry out image acquisition at a farther distance, and can actively track people's eyes and automatically zoom without the cooperation of people, thereby greatly improving the iris recognition system. The application range of the iris recognition system is improved, the use of the user is facilitated, and the invention has great market potential.

实施例1：应用于自动化和智能化的虹膜识别系统Embodiment 1: the iris recognition system applied to automation and intelligence

本发明可用于自动化和智能化的虹膜识别系统。使用过程如下：在某个会馆入口，需要对来访人员进行虹膜识别验证身份，但由于客人都比较尊贵，不方便让机器指挥客人去上下左右移动来完成识别，所以会馆入口采用了自动化的虹膜图像识别装置。当客人进入会馆后，面对智能虹膜识别系统逗留3秒钟时间，设备会自动旋转窄视角摄像头，捕捉到了虹膜图像，识别出客人身份，并发出该“某某先生，欢迎光临”的声音。客人感觉到会馆安全措施很好，并且使用方便、智能化高，结果非常满意。The invention can be used in automatic and intelligent iris recognition system. The process of use is as follows: At the entrance of a clubhouse, visitors need to perform iris recognition to verify their identity. However, since the guests are more distinguished, it is not convenient for the machine to direct the guests to move up, down, left, and right to complete the identification. Therefore, an automated iris image is used at the entrance of the clubhouse. Identify the device. When a guest enters the clubhouse and stays in front of the intelligent iris recognition system for 3 seconds, the device will automatically rotate the narrow-angle camera, capture the iris image, identify the guest, and issue the voice of "Mr. So-and-so, welcome." The guests feel that the security measures of the hotel are very good, and it is convenient to use and highly intelligent, and they are very satisfied with the result.

实施例2：应用于大规模用户的虹膜识别系统Embodiment 2: The iris recognition system applied to large-scale users

本发明可以提高虹膜识别系统的实用性，尤其对于大规模用户和特殊人群。实例如下：The invention can improve the practicability of the iris recognition system, especially for large-scale users and special groups. Examples are as follows:

某个通关口岸需要利用虹膜识别系统验证身份，过往的人需要排队进行验证，其中有三个人：张三，身高190，李四，身高150，王五，从未使用过该仪器。三个人都曾注册了自己的虹膜图像，他们来到关口进行识别。关口以前使用的虹膜识别仪器高150，张三需要蹲下身子并将脸抬起来才能进行识别，而李四掂起脚尖才能看到镜头，王五由于不熟悉仪器，虽然对准了眼睛，但是不能把握好和摄像头的距离。结果三个人都很长时间不能将自己的眼睛对准镜头，后来在工作人员的帮助下才完成了识别，以至于在后面等待是别的人排队拥堵，浪费了大量时间。A certain customs clearance port needs to use the iris recognition system to verify the identity, and the passers-by need to queue up for verification. Among them are three people: Zhang San, who is 190 tall, Li Si, who is 150 tall, and Wang Wu, who have never used the device. All three had registered images of their irises, and they came to the gate for identification. The iris recognition instrument used by Guankou was 150 meters high. Zhang San had to squat down and raise his face to recognize it, while Li Si had to lift his toes to see the lens. Wang Wu was not familiar with the instrument, so although he aligned his eyes, Can't grasp the distance from the camera. As a result, the three of them couldn't keep their eyes on the camera for a long time, and later they completed the identification with the help of the staff, so that waiting behind was other people queuing up and congesting, wasting a lot of time.

使用本发明的虹膜识别系统后，张三站在仪器约几秒钟，摄像头迅速向上旋转了一定角度，对准张三的眼睛，虹膜识别成功；李四站在仪器前约几秒钟，摄像头迅速向下旋转了一定角度，对准李四的眼睛，虹膜识别成功；王五不熟悉仪器，站在仪器较远的地方，约几秒钟的时间，摄像头转动对准王五的眼睛并自动对焦，虹膜识别成功。后面的整个识别队伍都顺利地完成了识别，平均每个人用时只有几秒钟，节省了大量时间。After using the iris recognition system of the present invention, Zhang San stood in front of the instrument for about a few seconds, and the camera quickly rotated upwards at a certain angle, aiming at Zhang San's eyes, and the iris recognition was successful; Li Si stood in front of the instrument for about a few seconds, and the camera Quickly rotated downwards at a certain angle, aimed at Li Si's eyes, and the iris recognition was successful; Wang Wu was not familiar with the instrument, so he stood far away from the instrument, and within a few seconds, the camera turned to align with Wang Wu's eyes and automatically Focus, iris recognition is successful. The entire identification team behind successfully completed the identification, and the average time for each person was only a few seconds, which saved a lot of time.

实施例3：用于隐蔽的虹膜识别系统Embodiment 3: For concealed iris recognition system

本发明可以用于在用户不知情的情况下隐蔽进行虹膜识别。实施方法如下：在某个安全中心的必经地带(例如电梯)，安装一台自动跟踪的虹膜识别系统。并在设备前，安装一面可以透红外光、反射可见光的镜子，使得用户可以在镜子中看到自己，却看不到里面的设备。某个犯罪嫌疑分子凭借假的证件进入此安全中心，进入电梯，在电梯里无意的照镜子整理一下头发。镜子后边的虹膜采集仪器自动捕捉到了此人的虹膜图像，并识别出为犯罪嫌疑分子，于是发出报警声，公安人员赶来将其抓获。The invention can be used for concealed iris recognition without the user's knowledge. The implementation method is as follows: an automatic tracking iris recognition system is installed in a necessary zone (such as an elevator) in a certain security center. And in front of the device, install a mirror that can transmit infrared light and reflect visible light, so that users can see themselves in the mirror, but they cannot see the equipment inside. A certain criminal suspect entered the security center with a false certificate, entered the elevator, and unintentionally checked his hair in the mirror in the elevator. The iris collection device behind the mirror automatically captured the person's iris image and identified him as a criminal suspect, so the alarm sounded, and the public security personnel came to arrest him.

Claims (12)

1. An iris image acquisition method based on automatic tracking and automatic zooming is characterized by comprising the following steps:

1) installing a wide visual angle camera, a narrow visual angle camera and an infrared light source on a controllable holder;

2) continuously shooting images by using a wide-view-angle camera, detecting whether a face image exists, and when the face image is detected, carrying out human eye detection on the upper half area of the face to obtain the accurate position of eyes;

3) rotating the holder until the narrow-view camera and the infrared light source are aligned with the eyes of the person according to the position of the eyes in the wide-view camera image, and simultaneously enabling the narrow-view camera to obtain the eye image of the person;

4) judging the distance of a user and the image quality, and controlling the automatic zooming and the automatic focusing of the narrow-view-angle camera;

5) after the eyes of people in the narrow-view-angle camera are focused, the narrow-view-angle camera continuously shoots iris images;

6) and judging whether clear iris images are acquired or not, and if clear images exist, performing image processing and iris recognition to judge the identity of the user.

2. The method according to claim 1, wherein in step 1), a wide-angle camera, a narrow-angle camera and an infrared light source are closely arranged on the controllable holder, and the two cameras and the light source are arranged in parallel but not collinear; the tripod head drives the three to rotate together.

3. The method of claim 1, wherein in the step 2), the human face is detected by using a wide-angle camera, and the method comprises the following steps:

1) when the system runs, continuously shooting images by using a wide-view-angle camera;

2) the classifier trained by the existing method is used for detecting whether a face image exists in the image or not, and when the face image exists, the coordinates and the size of the face are obtained.

4. The method according to claim 1, wherein in step 2), the human eye detection is performed in the human face image, and the method comprises the following steps:

1) based on the detected face position, estimating the general position of the eyes of the person,

2) the human eye classifier trained by the existing method is used for detecting whether human eyes exist in the region, and when the human eyes exist, the coordinates and the size of the human eyes are obtained.

5. The method according to claim 1, wherein in the step 3), the pan-tilt head is rotated according to the image position of the eye on the wide-angle camera, and the specific method is as follows:

1) calculating whether the eyes appear at a calibration position of the wide-view camera: if the camera is at the calibration position, the camera with the narrow visual angle is aligned with the eyes, and if the camera is not at the calibration position, the camera needs to rotate the holder;

2) when the eye position deviates from the calibration position, the cloud platform is rotated to return to the calibration position: when the eye position deviates to the left or the right, calculating the rotation time, and then rotating the cradle head to the right or the left for the time; when the position of the eye is on the upper side or on the lower side, calculating the rotation time, and then rotating the holder downwards or upwards for the time; rotate until the eye appears at the index point.

6. The method of claim 5, wherein the calibration position is linearly calculated according to the approximate distance of the user by:

a) the calibration position is approximately at a position delta H below the center point of the wide camera, the linear relation is formed between the calibration position and the distance D between a person and the camera, and the linear relation is formed between the calibration position and the approximate length H of the detected face, namely delta H is aH + b, wherein a and b are parameters to be calculated;

b) in the preparation work, a plurality of different calibration positions delta H1, delta H2 and … delta Hn and corresponding different human face approximate length values H1, H2 and … Hn are respectively obtained at different distances from the user to the camera, so that, <math> <mrow> <mi>H</mi> <mo>=</mo> <mrow> <mfenced open='[' close=']' separators=','> <mtable> <mtr> <mtd> <msub> <mi>H</mi> <mn>1</mn> </msub> <mo>,</mo> <mn>1</mn> </mtd> </mtr> <mtr> <mtd> <msub> <mi>H</mi> <mn>2</mn> </msub> <mo>,</mo> <mn>1</mn> </mtd> </mtr> <mtr> <mtd> <mo>&CenterDot;</mo> <mo>&CenterDot;</mo> <mo>&CenterDot;</mo> </mtd> </mtr> <mtr> <mtd> <msub> <mi>H</mi> <mi>n</mi> </msub> <mo>,</mo> <mn>1</mn> </mtd> </mtr> </mtable> </mfenced> </mrow> <mo>,</mo> </mrow></math>Δh＝[Δh₁，Δh₂，...，Δh_n]^Tthen linear regression calculation is carried out to obtain <math> <mrow> <mrow> <mfenced open='[' close=']' separators=' '> <mtable> <mtr> <mtd> <mi>a</mi> </mtd> </mtr> <mtr> <mtd> <mi>b</mi> </mtd> </mtr> </mtable> </mfenced> </mrow> <mo>=</mo> <msup> <mrow> <mo>(</mo> <msup> <mi>H</mi> <mi>T</mi> </msup> <mi>H</mi> <mo>)</mo> </mrow> <mrow> <mo>-</mo> <mn>1</mn> </mrow> </msup> <mrow> <mo>(</mo> <msup> <mi>H</mi> <mi>T</mi> </msup> <mi>&Delta;h</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow></math>

c) When the system runs, the calibration position is calculated by detecting the approximate length H of the human face obtained each time, and then the coordinates of the position and the coordinates of the detected human eyes are compared to judge whether the eyes appear at the calibration position.

7. The method of claim 5, wherein the rotation time is calculated by empirical determination, and the method comprises:

a) in the preparation work, different rotation time is respectively set, and the holder is respectively rotated towards the horizontal direction and the vertical direction to obtain the change distance of the eye position in each rotation;

b) tabulating the empirical data and recording the tabulated empirical data into a computer program;

c) in the system operation, the program looks up the distance that should be rotated horizontally and vertically, and looks up this table to find the time needed to rotate horizontally and vertically.

8. The method according to claim 1, wherein in the step 4), controlling the narrow-view-angle camera to automatically zoom and automatically focus comprises:

1) roughly estimating the distance of the user according to the size of the face;

2) adjusting the zoom to a preset focal length according to the approximate estimated distance, wherein the focal length is changed from a long distance to a long focal length when the distance is far, and the focal length is changed from a short distance to a short focal length when the distance is close;

3) judging the high-frequency component of the iris image, then controlling the lens to focus to be far away or close, if the zooming high-frequency component is increased, continuing the original mode of focusing, otherwise, if the zooming high-frequency component is decreased, focusing in the opposite direction;

4) until a clear iris image is acquired, wherein the clear judgment of the iris image is determined by high-frequency component empirical values obtained in a system test.

9. The method as claimed in claim 1, wherein iris recognition is required after the clear iris image is acquired in the step 6), and the iris recognition method comprises the following steps:

a) carrying out image preprocessing and normalization;

b) iris positioning and iris feature extraction are carried out;

c) comparing the collected iris features with iris features in a database to see whether the iris features are matched with iris images in the database or not;

d) if the iris image is matched with the iris image in the database, the name of the person is automatically read out, and the successfully recognized image is stored.

10. The method according to claim 1, wherein voice prompt, light prompt or operation of opening or not opening the door is performed according to the comparison result after the iris recognition of the step 5) is completed.

11. A system for use in the method of claim 1, comprising a controllable pan/tilt head, a wide view camera, a narrow view camera, an infrared light source, a computer, a power source, a speaker; a wide visual angle camera, a narrow visual angle camera and an infrared light source are sequentially superposed and arranged on a vertical rotating shaft of the controllable holder;

the selection and connection of each device are as follows:

a) the wide-view camera is a common conference camera and is connected with the computer through a USB interface;

b) the narrow-view camera adopts a modified integrated camera, video data of the camera is connected with a computer through a data acquisition card, and the integrated camera controls focusing and zooming through an RS-232 bus;

c) the controllable tripod head is a fast digital tripod head with 100 degrees/second, is connected with a computer through an RS-485 interface and is controlled by a standardized Pelco-D protocol;

d) the infrared light source is an LED lamp array and is composed of a plurality of infrared LEDs with 15-degree emission angles.

12. The system of claim 11, wherein the computer hard disk has software copied therein:

a) the software starts two threads, respectively processes the image data of the wide-view camera and the narrow-view camera, and performs image processing, motion control and database retrieval functions in the background;

b) the main window of the software has two pictures which respectively display a face image and an iris image; after the user identification is successful, displaying user information and making a sound of the user name; the main window is provided with a registration and identification button aiming at user management.

Images