Eye corners detection using Haar cascade classifiers in controlled environment

Salina Mohd Asi, Nor Hidayah Ismail, Roshahida Ahmad, E. Ramlan, Zainal Arif Abdul Rahman

Research output: Contribution to journalArticlepeer-review


Facial landmarks detection is undoubtedly important in many applications in computer vision for example face detection and recognition. This article demonstrated the use of Haar Cascade Classifiers to automatically locate the eye corners. We acquired our 3D face image data by Vectra 3D camera in a controlled environment. We use two data set of 300 eye images to train en and ex cascade classifiers regardless of the left and the right eye. These classifiers were then used to detect and locate the inner (en) and outer (ex) eye landmarks. To train HAAR cascade classifier we usually use huge amounts of data. But in this study, about 300 positive images used to train each classifier. Due to this we observed quite an amount of false positive detection. We developed a simple algorithm to predict the eye corners by first eliminate the false detection and geometrically modeled the eye. Our classifiers able to detect and locate en on 53 out of 60 test images and the ability to detect ex in 59 out of 60 test images. In craniofacial anthropometry, it is very important to locate the facial landmarks as per the standard definition of the landmarks. Our results demonstrated accurate detection of ex and en facial landmarks as per standard definition. In conclusion, our trained enHaar and exHaar cascade classifiers are able to automatically detect the en and ex craniofacial landmarks in a controlled environment.

Original languageEnglish
Article number598
Pages (from-to)69-75
Number of pages6
JournalJournal of Engineering and Technology
Issue number1
Publication statusPublished (in print/issue) - 1 Jul 2014


  • Craniofacial anthropometry
  • Facial landmarks
  • Haar cascade classifier


Dive into the research topics of 'Eye corners detection using Haar cascade classifiers in controlled environment'. Together they form a unique fingerprint.

Cite this