Precise Vectorization Of Industrial CT Image And Software Development

Industrial Computed Tomography(CT) is a non-destructive testing technique. It can test the objects without contact or damage. Industrial CT images are of grayscale.When we apply these images to Reverse Engineering(RE), the vectorization procedure is necessary. Industrial CT images contain geometry information of the corresponding slices. With the vectorization procedure, it is possible to restore the geometric information for redesign. In order to convert an industrial CT image to a two-dimensional CAD drawing, we use Visual Studio 2010 to develop the software system.Firstly, the image preprocessing methods are studied, including image enhancement, thresholding, image point operations and edge detection, etc. We mainly use the median filtering for image enhancement, the adaptive thresholding for thresholding, improved edge detection and improved Canny algorithm for edge detection. By selecting the threshold, the improved edge detection algorithm achieves the effect of intuitive judgments to the edge detection. The improved Canny edge detection uses the weighted value of the heuristic to detect image edge information, and finally use the Canny algorithm to enhance edges.Then we study the vectorization procedure based on geometry recognition and the calculation functions of model registration and error analysis. The geometry recognition methods include line vectorization, the improved circle detection algorithm, 3-point ellipse detection of the Hough transform algorithm, the vertical bisector method for arc detection. Based on the comparison of existing algorithms, we proposed an improved arc detection method: determining the seeds of the arc firstly, then set up the equations and detect the useful sampling point. The improved arc detection method can reduce the processing time, but its accuracy largely depends on the accuracy seeds of the arc.Moreover, we match the edge contour point model with two-dimensional CAD model using model registration and error analysis to analysis the dimensional error.Finally, we use the Visual Studio 2010 to develop a vectorization software. The system contains a number of interactive features. The module of Bitmap Editor comprises the connection of broken edge, erase point, erase line, erase block, erase marquee, etc. The module of Vector editing comprises merge a straight line between two points, split the line segment, move the feature points, use three-point to the arc fitting.The practical application showed that the software can effectively convert an industrial CT image to a vector image, and can achieve an accuracy of 0.07 mm to the standard circle. It can meet the demands of practical application.