基於嵌入式系統之電腦視覺設計與實現

Design and Implementation of an Embedded System based on Computer Vision

楊曜坤(Yao-Kun Yang)

指導教授 ： 曾遠威

義守大學/電機資訊學院/電機工程學系/碩士(2014年)

https://doi.org/10.6343/ISU.2014.00341

摘要

人臉具有很多生物特徵，因此人臉識別安全系統是比那些只需要通行碼更為安全。在這篇論文中，嵌入式的人臉識別安全系統的設計。是當嵌入式系統識別一個特定的人出現在鏡頭前，關閉的門將會被允許解鎖。換言之，如果該使用者是不被系統識別的人，門會保持鎖定狀態。在嵌入式系統上運行的人臉識別程式是基於Haar-like特徵訓練和AdaBoost分類器開發的。建立一個基於ARM架構的嵌入式系統，如Bootloader，Linux內核的遷移的細節，創建根文件系統和安裝外圍驅動設備的移植已呈現。對於嵌入式系統，人臉辨識過程通常只約要2秒鐘。因此，嵌入式系統的即時性能是有保證的。

關鍵字

嵌入式系統 ； 人臉識別 ； Haar-like特徵 ； AdaBoost分類器 ； 即時性

並列摘要

Human faces consist of many biological features. Therefore, face identification security systems are much safer than the traditional ones that only take passcodes. In this thesis, an embedded face identification security system is designed. When a particular person who is recognized by the embedded system appears in front of the camera, a closed door will be unlocked. On the other hand, if the person is not recognized by the system, the door will keep locked. The face identification program running on the embedded system is developed based on Haar-like feature training and AdaBoost classifier. The details of building up an ARM based embedded system such as migration of Bootloader, migration of Linux kernel, fabrication and migration of root document system and setup of peripheral driving devices have been presented. For the embedded system, face identification process usually only takes about 2 seconds. Therefore, the real-time performance of the embedded system is assured.

並列關鍵字

embedded system ； face identification ； Haar-like ； AdaBoost ； real-time ； ARM ； Lunix

參考文獻

[1]王崇吉， “嵌入式系統設計在人臉辨識應用”，義守大學電機工程學系碩士學位論文，2013。

[2]沈啟亮， “即時的人臉檢測系統”，淡江大學電機工程學系碩士在職專班碩士論文，2003。

[3]Rainer Lienhart, JochenMaydt,” An Extended Set of Haar-like Features for Rapid Object Detection”.

[4]Paul Viola, Michael Jones. “Rapid Object Detection Usinga Boosted Cascade of Simple Features,＂IEEE, 2001.

[5]HannesKruppa,ModestoCastrillonSantana,BerntSchiele.“Fast and Robust Face Findi-ng via Local Context,” Joint IEEE International Conference, 2003.