Contour Matching Based On Hausdorff Distance

With the rapid development of scientific technique, shape matching become a very important technique in information processing field, especially in image processing field. It has been extensively applied in computer vision, resource analysis, medical image matching, weather report, traffic management and character recognition. When the objects are recognized by the machine, the contours of known image need to be matched with the unknown image partly or entirely in space. The process that finding the sub-present of the mode that is known based on the shape of the known mode(generally is the object that people fond of) is the shape matching.Firstly, the definition of shape matching is introduced. Then, this technique is applied to contour matching. The discussion and research is developed about this application, and two contour matching algorithms are proposed. The research work mainly contains:The first algorithm, Feature points are detected by using curvature. Feature points and point on both sides of them make up of the feature segment. Then feature segment is used to describe the contour segment by segment. Finally, the result of matching is gained by describing the segment contour by the structure segments and measuring the comparability of the segment contours taking the hausdorff distance as the measurement of the comparability. For geometrical characters of the contour itself have been taken good advantage,the algorithm ensures preferable matching precision,and accelerates computational speed.The second algorithm aims at problem of the inflexion and tangent point. Angle point, inflexion and tangent point of contour line are separately detected through study of curvature angle and curvature symbol. Treating the feature point as center and R as radius, feature segment is make from several feature points in its neighborhood. This algorithm is very precise at detecting the feature points. It has a strong ability of detection and orientation feature points.The matching rate of contour line for which inflexion and tangent points occur frequently can reach 98%.