基于约束局部模型的人眼特征点定位

Location of Eye Feature Points Based on Constrained Local Model

在眼动追踪技术中需要提取有效的眼动参数,针对各类人脸人眼特征点定位过程中不涉及瞳孔定位的问题,提出了一种有效的人眼特征点定位方法。首先用约束局部模型训练方法对人眼训练好模型,然后利用AdaBoost算法检测到人脸和人眼图像,最后把人眼的位置和图像信息传递给约束局部模型,经过搜索,精确定位出人眼特征点。实验表明,提出的方法不仅对光照和头部姿态有很好的鲁棒性,而且检测速度快,定位准确率高,能适应实际场景中应用。

In eye tracking technology,effective eye movement parameters need to be extracted.In order to solve the problem of pupil positioning in the face feature points of all kinds of faces,this paper proposes an effective method for eye feature point positioning.First,use the constrained local model training method to train the human eye model,then use the AdaBoost algorithm to detect the face and human eye image,and finally pass the human eye position and image information to the constrained local model.After searching,the human eye is accurately located Feature points.Experiments show that the method proposed in this paper not only has good robustness to illumination and head posture,but also has fast detection speed and high localization accuracy,and can be adapted to practical scenarios.