Research On Three-dimensional Modeling And Elementary Measurement In Micro Stereovision

Micro stereovision system based on stereo light microscope (SLM) has been applied in some micro-domains widely because highly accurate measurement and location are implemented by the vision feedback. The micro stereovision system contains six steps, which refer to capturing images, system calibration, stereo matching, 3D reconstruction, 3D modeling and 3D measurement. To satisfy requirement in measurement, image preprocessing, stereo matching, 3D modeling and 3D measurement are studied in this paper.The quality of the image captured takes effect of results in stereo matching directly. However, the defects of vision system always impact the quality of image captured in fact. On the basis of analyzing the hardware's defect for vision system, image normalizing is implemented to reduce the difference between left image and right image, and SUSAN filtering is applied to eliminate bad effect of abnormal noise in image. Besides, the obvious features in images are detected by the SUSAN algorithm and edge detection method based on the signal processing.In order to make rapid matching come true, after the mathematic model of micro stereovision based on SLM is analyzed, single scanning strategy is implemented to improve the matching speed in according to the imaging disciplinarian between image pairs. In experiment, the computing time is less than half of original strategy's time. At the same time, a robust and multi-feature matching method is put forward by combining color value in RGB channels, vertical grads and geometrical comparability in image pairs, which is proved an effective algorithm in experiments.Effect of error matching points is a very intractable problem that must be resolved in 3D modeling. In terms of the characteristic of matching points, double linear interpolation is implemented to make the matching points thick, and the filtering for 3D matching points eliminates the bad effect of error matching points in 3D modeling. In order to obtain smooth 3D mesh, a method based on trilateral filter is applied to smooth mesh while preserving the feature of 3D graphic. Besides, Delaunay triangulation, mesh optimizing, B-B triangle interpolation, and texture mapping are implemented to reconstruct the real objects.At last, the model of micro stereo vision is analyzed to evaluate the system accuracy in quantity. Relative error and maximal error are evaluated both in quality and in quantity through statistical method for vision system. The results of analyzing show the relative quantization error of CCD is less than 2% when the height of obeject measured is less than depth of field. In experiments, the relative depth error is 3.57%, and absolute position error k 0.063mm for measuring a resistance. The relative depth error is 18.10%, and absolute positionerror is 0.016mm for measuring a micro chip. The experiments prove that the micro stereovision system based on SLM may reach to high precision.