| Recent years, modern industry puts forward higher requirements on the measurement techniques. However, the traditional measurement methods are unable to satisfy the requirements and severely restrict the development of related industries. Under this background, the vision measurement technique which has advantage of non-contact, fast speed, high precision,100%online and easy to computer integrated control, arises at the historic moment and is now playing an increasingly important role.Considering motion body characteristics of existing structured light vision measuring systems, a line structured light surface three-dimensional (3D) measuring system is developed which can realize high precision measurement of object surface in multi-posture. In this dissertation, the self-designed measuring system is used as experimental platform, and related calibration techniques are studied intensively. The main work and research results are summaried as follows:1. A line structured light surface3D measuring system is established. In the aspect of hardware, mechanical motion body, electronic control unit and sensor data acquisition module are designed. On the software side, the overall software architecture and module division are introduced, then the software interface of the system is developed.2. A novel two-step calibration approach for nonlinear camera model is proposed. First, the distortion parameters are calibrated by means of non-metric method, and a more robust correction scheme is selected by comparison. Then, the remaining camera parameters can be calibrated according to the pinhole camera model. The fact that the distortion center and principle point are different is demonstrated through experiments, and calibration accuracy is further improved. In addition, the non-metric correction process avoids the possibility of coupling between distortion parameters and remaining intrinsic camera parameters, and the dimensions of optimization problem are reduced.3. Geometric calibration approach for line structured light sensor is proposed. According to the geometric meaning of light plane parameters, the calibration work can be completed in two steps. First, the normal vector of light plane is calibrated, and then we calibrate the distance information. Because of the flexible characteristics of the motion body, two different implement ways are designed. One is calibration method based on active vision, and another one calibrates the sensor using projection characteristics of single circle target. In summary, the geometric method is simple with no complex computation, and it can achieve rapid calibration on site.4. A novel kinematics calibration method based on flexible splicing target is proposed. The flexible splicing target is composed of fixed target and moving target, the fixed target’s position is unchanged in the calibration process, but the moving target can be freely placed. The advantage of flexible splicing target is that a large target can be formed after splicing feature points on the fixed target and the moving target, and then a wide range of joint motion trajectory can be computed. The experiment results show that the method has high accuracy, and the cost of calibration is reduced.5. A hand-eye calibration method based data clustering and selection is proposed. The fact that different distribution of calibration data will affect final calibration accuracy is pointed out, and the corresponding data selection mechanisms are established. Besides that, a novel objective function is defined considering error presence of calibration data, and then hand-eye relationship can be calibrated through nonlinear optimization algorithm. The experiments result show that the proposed method has higher accuracy.Finally, future work is provided based on the conclusion of whole work. |
PDF Full Text Request