眼睛檢測演算法的比較

Abstract

本論文最主要是研究及探討，人的臉部器官中眼睛的辨識。為什麼要辨識人臉器官中的眼睛呢？因為人的眼睛可以傳達很多的訊息，當人累的快睡著的時候，眼睛會不自覺的闔上；而當你精神飽滿的時候眼睛就會比平常時還要大。如果我們能把眼睛辨識的方法應用在駕駛人身上，我們就可以做一個自動提醒駕駛人的安全系統；或者應用在辨識人的身份。因為只要我們能夠建立一個公司員工的臉部影像的資料庫，那我們在辨識一個人的身份時，臉部的五官就非常重要了；尤其眼睛是靈魂之窗，每個人的眼睛在臉的什麼位置、什麼形狀跟大小，我們都可以依據這些資訊來判斷這個人的身份。 然而做眼睛檢測的方法有許多，比如：1、粗略輪廓預測副程式RCER(Rough Contour Estimation Routine)、數學型態學(mathematical morphology)配合變形樣本模型(deformable template model) 這些影像處理的技術。2、邊緣搜尋的技術。然後利用這些技術來找出眼睛的位置跟形狀，並比較彼此之間的準確度。在本篇論文中，最主要是評估本論文所提及檢測眼睛的形狀及大小的方法，並作之間的優劣比較與說明。

This thesis is towards the research of detecting people's eyes in the face. People's eyes contain transmit and rich information. When we are tired and sleepy, eyes will not be getting conscientious. On the other hand, Eyes will be also large when you are energetic. If we can watch driver's eyes, We can design a driving security system reminding the driver automatically. Eyes are also important to distinguish people's identity. If we can set up the face image database of the staff of a company, Then we can distinguish a person's identity by one's facial features. Especially eyes are the window of the soul, how big and where in the face, and iris are crucial to determine the identity of a person. There are many methods to detect eyes in a face, For example, 1, The rough eye outline prediction by RCER (Rough Contour Estimation Routine), mathematical morphology and deformable template model; 2, edge detection technologies. We make use of these technologies to find the position and shape of eyes, and then the accuracy of each method is computed. This thesis aims at finding out an efficient method to detect the eyes and their shapes. Comparison among these methods are made. Furthermore, advantages and disadvantages of these methods are finally shown and noted.