A Contour Curve Matching Method

Contour curve matching is an important subject of computer pattern recognition, detection and query of remote sensing image. And it has been widely used in computer vision, resource analysis, remote sensing image detection, weather forecast, traffic management, character recognition and other fields. Shape is the intuitive response of target contour in the image, therefore the object contour curve is an important feature of object shapes. By matching the contour curves a match of the shape can be completed, then further the image matching can be done.Based on the study of object contour curve characteristics, the paper presented a characteristic, the ratio of arc length and chord length, finally contour curve matching was realized by the comparison of the characteristics. The main work carried out in the paper is as follows:1. Firstly described the relationship of shape matching and curve matching, then the related concepts, research background and research status of shape matching and curve matching were introduced.2. Studied the theory of shape matching problem. Mainly included the method of shape matching, the description methods of shape matching and the description of contour curve. Then the paper studied image preprocessing and contour curve extraction and processing.3. Key features and characteristics of the contour curve was studied, this paper took the curvature (constant displacement, rotation invariant, scaling and other related features) extreme points as feature points.The curvature feature points could be extended to the ratio of arc length and chord length, the size of the area, extreme angle and the centroid distance increment matrix etc. The features are similar and affine invariant, so you can use these features to characterize the information of contour curves.4. Presented a matching method of contour curve.The matching method is mainly focused on rough matching and accurate matching of curve, the first phase of the rough curve matching can remove the curve with a big difference,and the second stage the accurate matching can complete the overall curve detection. In these matches the design used the MSE of corresponding characteristic ratio to determine whether the curve is matched or not, finally through the comparison with these mean square error thresholds curve matching detection were completed.