Constrained Sketch Generation Based On Image Contour Detection

Image based product modeling is a popular method for product development. Althouth research usually focuses on rebuilding3D model by multiple images, image contour tracking to get product sketch in different angle of view and assemble sketches into3D product model is also a valuable application. The famous3D product design software SolidWorks supplies image based sketch creation service. However it depends on user manual operation.Therefore there is no auto-extraction methods so far.This paper will present a new method which can auto-extract image contour and create constrained2D sketch automatically. The performance of model based CAD system lies on sketch image since sketch constraint determines the degree of quality in product design. The sketch this paper will generate is composed of lines, arcs, B-spline curves.This paper mainly focuses on the following research:Detect of image contour. It consists of gray processing, image binarization and binary image contour trace. Color images usually have a lot of useless information.We get binary image through OTSU method, and then trace the contour of the binary image.Geometry feature recognition of contour based on Hough Transform(HT) and randomized circle detect(RCD). The characteristic here means geometric characteristics such as line and circle. By recognition of straight line segments and arc segments on the contour data, make preparations before fitting work. Taking into account that the traditional method of feature points detecting based on curvature calculation is sensitive to noise points.By using Hough Transform to detect all possible point sets of line and Randomized Circle Detection algorithm to get all possible point set of circular arc, this paper successfully avoid the influence of noise points, and reach the ideal results.Constrained fitting of image contour.Although hough transform and randomized circle detection can also get line and circle equations. But its accuracy can not meets the needs of this paper. So I decide to use the least square method to fit lines and circles at last. This strategy successfully resolves those problems, and achieves a high fitting accuracy. As the B-spline curve have a lot of excellent properties, it is easy to implement the tangent continuous of its two end points.And at last, our algorithm can generate constranined contour sketch automatically.