Reserch On High Resolution 3D Surface Measurement Of Human Face

Three-dimensional (3D) non-contact profilometry in the machine vision field, with the excellence in high speed and high accuracy, has been widely applied in Industrial Automation, Converse Engineering, and Biomedical Engineering etc.With the support of National Science Funds of China, we have developed an efficient and precision method for 3D surface acquisition of human face based on the optical measurement technique, aiming at supplying enough space information with high quality for 3D infrared image.A binocular 3D measurement system has been constructed based on multi–stripe, and a self-developed instrument and attached software with intellectual property right have also been developed. A projector is used to project digitally created fringe patterns onto the object, and CCD cameras are selected to capture the deformed fringe patterns. In order to obtain 3D surface data of human face, we have discussed all the related techniques completely and integrally from the point of view of the basic theoretical analysis, construction of the system, calibration and Image processing.Shadow effect is an unavoidable problem existed in the process of human face imaging. The binocular 3D measurement system contains of two cameras, and every camera having a corresponding image acquisition card and both of them construct a subsystem together with the projector which located between of them in space. Then the system effectively decreases the shadow influence by widening the view of measurement, and enhanced the veracity of measurement. In addition, combination of the projector and moving plane-stripe, made the system scan the inspected object by the lights with optional parameters such as the width, number, space distance and scanning speed optional. And contour maps are formed by the lights that modulated by the inspected object. In the map,match between the lights and image points becomes complex as the stripes are always curved or broken. The method based on stripe gray code presented in the paper, has helped each stripe image point easily find its corresponding structured light.In order to acquire 3D surface data with high precision, we have also developed a more practical 3D measurement model in the paper and analyzed the factors that affect the measurement precision.It’s well known errors existed in the system component would reduce the measurement precision. This paper has discussed the basic arithmetic of camera calibration, and finished the calibration of the intrinsic and extrinsic parameters, as well as the space location of each camera respectively with the system coordinates. Furthermore, single camera method and binocular method that based on multi-point-fitting have been presented in detail and an interpolation method for calibrating all dynamic scanning lights based on geometry has also been introduced.The key factor of stripe image processing is drawing out the center of the projected stripes. The paper has supplied a simple three-step method which can eliminate white noise in experiment to detect center points with high speed. It is validated that the measurement precision can reach to 0.5mm in the 3D reconstruction of a plaster head.