| Abstract: | 現在影像辨識技術的研究已經研究許久，在許多領域上皆有所應用，如車牌辨識、人臉辨識。而本論文主要的研究是另一領域的應用，目前尚未有人做過的“與面相學結合的人臉特徵與個性分析系統”，而且此研究剛好也結合了中國的傳統文化。本系統主要提供一個與面相學結合的個人大頭照自動化人臉特徵分析系統，結合面相特徵來分析個性特質，經由對映工作職務的個性特質，給予適任的職務的建議。
系統首先找出臉部的位置，第二步驟、對各臉部子區域（眼睛、嘴巴、鼻子）做定位動作。第三步驟、針對臉部子區域做特徵擷取，得到位置和大小的特徵值。第四步驟、定義眼睛、嘴巴、鼻子的面相特徵，如“大眼睛”。第五步驟、使用臉部子區域特徵的特徵值來與對映的面相特徵的特徵值，得到個人的個性特質。第六步驟、經由個人的個性特質，與工作職務的個性特質對映，給予適任的職務的建議。
使用OPENCV既有的一些Function來達到基於特徵的人臉偵測系統，在人臉和眼睛、鼻子、嘴巴的偵測是使用Paul Viola的Haar特徵原理來定位。在人臉定位中，雖然原本的正確率已經達到90％以上，本系統加入判斷人臉區域是否有眼睛特徵的條件，有眼睛特徵才是真正的人臉物件，使正確率更達接近100％。
在眼睛、鼻子、嘴巴特徵的定位因為背景複雜的因素變多，致使正確率降低，本系統加入五官比例定位判斷的方案，實驗後確實可以提高辨識率。臉部子區域特徵擷取則經由各種影像處理技術，如平滑、腐蝕、膨脹、直方圖均衡化，並且加上二值化與影像映射而得到相關的座標特徵值，再經由座標特徵值計算而得到大小特徵值。本系統將面相特徵給予定義和量化，在使用座標特徵值和大小特徵值當參數來對映，得到個人的個性特質。最後由個人的個性特質與工作職務的個性特質對映，而給予適任職務的建議。
The Face recognition can be applied to many different fields. For example, the building entrance guard control, the criminal verification, the security verification of finance, and the identity verification of the network trade, etc. But, I don’t want to do these things that someone already do. I thing I can find an interesting application and nobody did it ever. This is my purpose.
The system is different with the above applications. It provides a new application and interesting applications. Combination of Physiognomy characteristics and facial recognition do the personality characteristics analysis. And, combination of the social psychology and the personality characteristics of work duties do suitable job recommendations by matching the personality characteristics.
The object of my research is the Analysis of Face Feature and Personality. I combine the Chinese’s Physiognomy and Face recognition to do a system. The system automatically analyses personality from personal photo of big head. Some image technical, for example, location and analysis I use some function from OPENCV that did from INTEL. In this paper, an approach to face detection based on Boosted Cascade arithmetic is introduced. This arithmetic adopts the image representation of integral image and cascade classificatory structure, so that the arithmetic can detect faces quickly and accurately. This paper also presents solutions of improving Boosted Cascade arithmetic. The first solution is using the model of eyes feature to verify face detection. The result is better than original. And, the second solution is to use the proportion and position of sub-block of face to verify the result of detection from eye, nose and mouth. The result is also better than original.
When we location the eye, nose and mouth, then we can get the position and size feature by some image process. Then, using these features to map the Chinese’s Physiognomy features, the Chinese’s Physiognomy features have to define and to quantify. We can get the personality of personal. Finally, we can use the personality of personal to map the personality of job. Then, get the right job of personal. |
