Camera Calibration Based on Face Key Point Detection

Abstract

基于使用二维精密转台和平行光管来标定相机的复杂性，提出了一种基于人脸关键点检测的相机标定方法，这种方法只需要对几张带有相同人脸的图片进行图像处理，识别出人脸的关键点，利用人脸关键点的坐标计算出二维射影变换矩阵，就可以得到相机的内方位元素和外方位元素。这种方法操作简单、速度快，容易由实验室内推到室外。由实验可知，相较于张正友的相机标定方法，该方法用人脸的图像作为标定模板进行标定，能达到更高的精度，验证了实验方法的简便性和有效性。 Based on the complexity of using a two-dimensional precision turntable and collimator to calibrate the camera, a camera calibration method based on face key point detection is proposed. This method only requires image processing on several pictures with the same face. Recognizing the key points of the face, using the coordinates of the key points of the face to calculate the two-dimensional projective transformation matrix, you can get the camera’s inner and outer orientation elements. This method is simple, fast, and easy to push from the laboratory to the outdoors. It can be seen from the experiment that compared with Z. Zhang’s camera calibration method, this method uses the image of the face as a calibration template for calibration, which can achieve higher accuracy, which verifies the simplicity and effectiveness of the experimental method.