The Research On Optimization For Line-Structured Light Sensor And Camera Calibration

With the development of 3D digital technology, three-dimensional detection techniques are playing an increasingly important role on the surface quality evaluation system for the modern industrial products. However, some problems in the traditional contact detection method, such as low accuracy and efficiency, severely restricted the improvement of product quality. In recent years, the optical 3D surface measurement technique, structured-light vision inspection technology as the representative, has attracted a great deal of attention in industrial environment for its some good characters, such as its high precision, non-contact, easy to control by computer and so on. Thus, this technique is too important to be replaced by others in the fields of 3D auto-measurement and object recognition.In this paper, the flexible robot 3D visual inspection system was used as the experimental platform, and some deeply researches and many related experiments about Line-structured light have been done, including design and optimization of line-structured light visual sensor, lens distortion correction for filter and camera calibration. These works laid a foundation for the improvement of detection precision.The main work and important contributions in the paper includes the following aspects:1.According to the mathematical model of line-structured light vision sensor, the imaging characteristics of the sensor is analyzed in detail. And on this basis, the constraint equations that the structural parameters of the visual sensor should satisfy are deduced with the combination of measurement error theory. Then the best structural parameters were obtained through nonlinear optimization methods, which meet the performance requirements of error and resolution.2.In order to improve the noise immunity of detection system, filter is installed on the camera. The distortion caused by filter is analyzed and verified. And then the division model (DM) is used to correct the distortion. With the Comparison of the polynomial correction model (PM), the experiments shows that this method is simple and effectively restrain the distortion to ensure the system's detection accuracy.3. A new camera calibration method using non-metric distortion correction is proposed, which is based on the same geometric type between straight line and its projection image. Single parameter division model is used to correct lens distortion. First, the Levenberg-Marquardt scheme is applied to fit circle using directly the distortion points, and the distortion parameter and the center of image can be calibrated robust. Second, using feature points after distortion correction, the remaining camera parameters can be calibrated linearly, based on the two basic constraint equations about camera intrinsic parameters. Different from existing methods, this method can be calibrated using only single image with fast efficiency and high precision, flexible and practical.