Research On Calibration Method For Line-Structured Light Vision Sensor

With the rapid development of computer vision technology and photovoltaic technology, three dimensional vision measurement as the non-contact measuring method has been applied more and more widely. And, sensor calibration technique, which is closely related to precision, is one of key techniques of the method. For the relaxation of the calibration field environmental conditions, and decoupling calibration parameters, the research about image feature extraction and vision sensor calibration are done. Also, the effectiveness of research results is validated. The main work of the thesis is as follows:1. The corner feature data on single circular target is extracted by the improved Harris algorithm, which can be reached sub-pixel level.2. Combined variable orientation template with grayscale barycenter method, the center of the laser stripe is extracted effectively. The creativity reflected in the method is that using of variable orientation template to determine the normal direction of each point on the skeleton. The efficiency and precision of extraction for laser stripe center data are improved by this method.3. A new method for camera calibration is proposed based on the cross-ratio invariance theory in projective transformation. This method using the linear solution to gain the distortion parameters of the lens is able to prevent the parameters coupling in the nonlinear iterative optimization process in traditional methods. In addition, the real-time for calibration can be insured based on the simple operation of experimental equipment. And, reliability and stability of the scheme is verified by the experiments. Experimental data show that:the calibration accuracy of system in the context of field operation is satisfactory, re-projection error is0.3589.4. Three-point perspective principle is adopted to realize the rapid light plane calibration. The method is suitable for the field calibration, with several advantages in the simplicity and flexibility of calibration, the low cost of calibration equipment, and no need to calculate the transformation matrix among different coordinate systems. The experimental data show that:the relative accuracy of sensing system can achieve about0.2953%.