Research On The Key Technologies Of3D Coordinate Measurement With The Virtual Controlling Network Of Single Camera

The optical-probe single-camera system is a typical3-D coordinate visionmeasuring system. With one camera’s single image acquisition the results can beobtained, since its low costs, no on-site calibration and flexibility, it has wideapplication prospect in the fields of modern machining and manufacturing industry.The key problem of the system is to design and construct the assisted constraints ofspace geometry, and then combine the camera’s perspective transformation from3Dto2D to realize the measurements of the spatial3D coordinates. The geometrystructure of the probe directly affects the measurement precision, and for the limitednumber of controlling points on the probe, their relative distance constraints areweakened and some individual image points with worse positioning errors cannot beexcluded, these are becoming the main restraining factors to improve measurementprecision.3-D coordinate vision measurement is a positioning measurement method basedon the principle of angular intersection, so it’s the fundamental way to add theconstraints of space geometry and strengthen the structure to improve themeasurement precision. The idea of bundle adjustment has been introduced into theoptical-probe single-camera system. We add redundant constraints by measuring theprobe positioned at various orientations, and then apply bundle adjustment to theseconstraints in order to improve the uncertainty of measurement. The main workaccomplished includes:1. A single-camera measuring method based on virtual controlling net isproposed. The idea of bundle adjustment usually applied in movable vision has beenintroduced into the fixed single-camera system. Before measurement, a initial virtualcontrolling net is constructed by moving the probe in measuring space, and imagingthe probe at different stations for a dozen times.2. A method on the precision correction of the camera’s distortion based onimage plane’s subzone is proposed. At first, the controlling space, constructed byseveral CFRP plates which were marked with GSI code target, is used as thecalibration target, and the camera takes photograph of it several times at differentdistance to acquire the camera’s internal parameters. After getting the global camera internal parameters, we implement the part calibration by dividing the image planeinto several parts, and in every part we do a part calibration to acquire the localdistortion parameters. The deviations from the global parameters are used as theimage coordinates’ corrections.3. The influence to the measurements brought by the positions and postures ofthe probe when establishing initial virtual controlling net is studied. We model thetwo-dimension and three-dimension measuring errors, and apply the numericalmethod, Monte Carlo simulation, to analyze the influence to the results by numbers ofmeasuring stations, and give the optimal measuring stations. On the basis ofstereoscopic vision’s non-planar intersection model, we apply the analytic method togive the optimal range of the angle of the optical axis intersection. By calculating thecoefficient matrix’s condition number of the over-determined equation systemestablished by co-linearity equations, and analyzing the relation between the measurepoints’ calculated errors and the condition number, we give the optimal arrangementof multiple measuring stations.4. The experimental platform was established, and measured point’s measuringrepeatability uncertainty, multi-posture measuring repeatability uncertainty and spacelength’s measuring uncertainty were obtained by traditional single-camera method andalso the proposed method, and the validity of the proposed method was verified bycontrasting the results obtained by the two methods.