الفهرس 
INTRODUCTIUONOnly 14 pages are availabe for public view
 |  | from | 109 |  |  |
| | | from | 109 | | |
Abstract
Structured light is an efficient optical technique for three-dimensional (3D) reconstruction of complicated and fragile objects. Many researchers focused more on studying different factors related to 3D reconstruction of complex objects, but less on the measurement error in the evaluated dimensions. Errors in products dimensions directly affect their function and performance. Product dimensions are evaluated through measurement processes consequently. Measurement errors directly affect the function of the product. In metrological applications, measurement errors are crucial and should be accurately evaluated. In the current study, the factors affecting measurement errors in the dimensions of a 3D reconstructed test object are studied. The test object is a cuboid. The considered factors are namely; the center distance between the camera and projector lenses (baseline), normal distance between baseline and calibration board, wavelength of the projected pattern (spacing between corresponding points of two black or white stripes), and the structured light coding technique. The camera and projector are calibrated using a calibration pattern to obtain the relationship between the 3D world coordinates and the captured image (2D pixel) coordinates. A full factorial experimental design is used to investigate the effect of the considered factors (parameters) on the measurement errors resulting in the three main dimensions of the test object. An analysis of variance (ANOVA) test is carried out to determine the most significant factors (parameters) effecting measurement errors of the 3D reconstructed object dimensions. The obtained results suggest that the center distance between the camera and projector lenses has the most significant effect on the resulting error in the length of the test object. However, the interaction between wavelength of the projected pattern and the coding technique has the most significant effect on the resulting error in width of the test object. While, the normal distance between baseline and calibration board has the most significant effect on the resulting error in height of the test object. The factors (parameters) levels that result in relatively smaller error values are found to be 270 mm for baseline, 900 mm for the normal distance between baseline and calibration board, 10 pixels for the wavelength of the projected pattern. Shifting and sub-pixeling coding technique also result in relatively smaller error values compared with shifting and gray coding technique. In this study, a two-step calibration procedure is suggested resulting in more accurate estimation of object dimensions. The first calibration step is implemented to relate the captured image coordinates to the real world coordinates using a calibration pattern. This calibration step is obligatory in structured light systems (SLSs) where the camera and projector are calibrated by capturing images of the calibration pattern consisting of black circular targets located at predefined geometrical positions. Before 3D reconstruction of the test object, the captured images point clouds must be filtered to exclude outlier points which impair the accuracy of estimated dimensions of the reconstructed object. In this study, the filtering process is carried out in two steps. The first filtering step is applied using the HP 3D scan software. The filtering factors (parameters) levels resulting in smaller error values are determined using design of experiments (DOE) and an ANOVA test. The optimum values of the considered factors (parameters) are later used in all experiments. To further improve system performance, a second filtering step is implemented where the Inter Quartile Range (IQR) and Sigma limits outlier filtering methods are used. The outlier filtering methods are applied to the point cloud of each plane applying the obtained optimum conditions. Outliers filtering methods are applied on both the full plane of the test object surface and part of the plane selected manually. The experimental results suggest that the 2-Sigma limits method applied to the plane point cloud, as whole, results in minimizing the measurement errors in the evaluated dimensions of the test object. In the second calibration step, calibrated block gages are used as calibrators. Stepped block gages with known sizes (from 0.100 mm to 5.000 mm) are reconstructed and their dimensions evaluated and compared to their nominal values. The two-step calibration method proposed in this study opens the prospects for the application of SLSs in several metrological applications. The measurement error in each step height is calculated. A relationship between the average value of the evaluated step heights and their corresponding reference values is obtained. The experimental results show the effectiveness of the two-step calibration method. The measurement error is reduced from 0.330 mm after applying the first calibration step to 0.095 mm after applying the proposed calibration step.