Target recognition method based on elastic matching

Abstract

The invention discloses a target identification method based on elastic matching. The method comprises the following steps: S1, adopting the contour extraction algorithm to extract the contour features of a target; S2, calculating the curvature value of each point on the contour of the target; S3, achieving hierarchical description of the contour points according to the size of the curvature value; S4, adopting PCA-SC distance to achieve similarity measurement of the contour features of different targets in a segmented manner; S5, achieving reliability measurement of the contour features of the targets in a segmented manner; S6, normalizing the target similarity and reliability indexes and adopting the elastic matching distance for target post-identification. The target identification method can realize effective similarity measurement of shapes of the targets, has a larger distinction degree for target characteristic description, has dimension invariance, rotation invariance and traversing invariance, improves the accuracy and the robustness in identification and reduces the calculation complexity.

Description

technical field [0001] The invention relates to the technical field of target recognition and shape matching, in particular to a target recognition method based on elastic matching. Background technique [0002] How to make machine vision recognize and recognize objects automatically is a very challenging problem in the real world. The key to this research is to provide an efficient and accurate shape descriptor to measure the similarity of objects. Target recognition and shape matching are important research topics in machine vision, and have been widely used in engineering, such as target category recognition, wide baseline matching, specific target recognition, image and video retrieval, scene classification, robot navigation, data mining and In many fields such as texture recognition. [0003] According to the source of the feature, the shape description method is divided into two categories: the method based on the contour and the method based on the transformation dom...

AI Technical Summary

Problems solved by technology

Currently, shape descriptor similarity measures for contour methods mainly include: Shape Contexts, Inner-Distance ShapeContexts, Bayesian classifiers, skeleton graph methods, etc., but these shape descriptors only take into account the positional relationship of the target contour, and do not Taking into account the essential properties of the profile: curvature, relative length and concave-convexity, that is, the reliability index of the profile

Images