頁籤選單縮合
題 名 | 基於不變量的單訓練樣本人臉識別=Face Recognition Using a Set of Similarity Invariants for Single Example Image Per Person |
---|---|
作 者 | 江仁宏; 陳以明; 林俊達; 王世昌; | 書刊名 | 前瞻科技與管理 |
卷 期 | 2:1 2012.05[民101.05] |
頁 次 | 頁107-122 |
分類號 | 312.1 |
關鍵詞 | 人臉識別; 單人單訓練樣本; 解析傅立葉梅林變換; 泰勒變換; 小波變換; Face recognition; Single example image per person; Analytical Fourier-Mellin Transform; AFMT; Taylor transform; Wavelet transform; |
語 文 | 中文(Chinese) |
中文摘要 | 單人正面人臉識別之困難,除了表情、光照、遮擋和姿態等因素外,另一是來自訓練樣本的獲得。當對每個人只能得到一幅影像作人臉識別時,由於訓練樣本不足,所提取的特徵向量不足以支撐整個人臉樣本子空間,識別效能因而大幅下降。這類問題稱之為單訓練樣本人臉識別問題。為解決此一問題,近年來多位學者已提出許多方法,諸如影像增強法、樣本擴張法和通用學習框架法等,概以利用各種技術合成多個虛擬影像,擴張訓練樣本數,使得變成一般的人臉識別問題。這樣子的解決方案,帶來更大的計算量以及儲存空間,且當有新的人員加入識別時,則需重新訓練,不利大範圍推廣。有鑒於此,本研究提出不需訓練樣本,直接萃取混合解析泰勒梅林頻譜臉之不變量,經投影至小波子空間降低維度後,可以有效減少表情、光照和遮掩條件變化所帶來的識別誤差,並以貝葉斯決策分類,可取得很好的效果。經分別應用於YALE和ORL人臉資料庫上,實驗顯示較之一系列之PCA方法等有更好的成效。 |
英文摘要 | Despite the constantly change of human face pose, illumination, expression, and occultation, one major problem of the face recognition technique arises from the difficulties of collecting samples. Under this limited condition, the database at hand is too weak to offer sufficient distinctive features to the system, and of course, the performance drop is expected. This problem is called the single example image per person problem. Single example image per person problem has received significant attention during the past years. Researchers have proposed sample-strengthen method, sample-expansion method, and generic learning framework, etc, which mostly aim to expand each person's face image sample by using computer technique to create several combining images based on the original one. Therefore, the single example image per person problem simply becomes multiple images per person recognition problem. However, these methods result in enlarging the calculation volume and requiring bigger storage space. It also needs to be retrained once a new person is put into system. These problems make it extremely difficult to popularize these methods. In this paper, we try to exclude training and to extract features directly from the hybrid Taylor-ATMT , which has constructed a set of similarity invariants. Human face expression change and partial occultation could be reduced by projecting it to wavelet space to lower the dimension, and then to compare the poster probability of those and classify categories, and which results positive recognition rate. Experiments are implemented on YALE and ORL face databases to demonstrate the efficient of proposed approach. The experimental results show that the average recognition accuracy rates of our proposed method which are higher than those of previous methods. |
本系統中英文摘要資訊取自各篇刊載內容。