Based On The Shape Of The Object Shape Recognition And Matching Context

Generally speaking, People’s understanding of an image to a very great extent depends on the shape of the target in the image to distinguish, therefore the shape is the foundation of people’s vision system for analysis and object recognition. We can use the obvious shape features of the target to recognize and match. This process is generally divided into two steps:shape extraction and shape matching. This paper study the two aspects respectively, which combine a good outline integrated from wavelet multi-scale with the shape context descriptor, and describe and align the shape of the object as to analyze and improve the experimental results.For the target contour features, the traditional edge operator often do not have both the noise resistance and the accuracy. In this paper we use a edge detection approach which based on the wavelet modulus maxima integrated multi-scale, combining the advantage of different scales, which not only have good accuracy but also take into account the noise resistance much better. The basic thought about this is along gradient direction to detection the maximum value points of moulds under the restriction of certain threshold, combined with different scales of information to get the final edge image through the certain rules.The contours are often used to object matching, i.e., we must use a shape descriptor to express the shape and measure the similarity between the shapes according to certain rules. Serge Belongie et al present a novel approach to describe strongly the shape:the shape context. It has the good definition of matching cost, simple calculation, strong distinguish ability and insensitive to noise etc, which is one of the most successful shape descriptors in recent years. At its core, shape contexts can be understood as a point set matching technique. We approach this as a three-stage process:1. solve the correspondence problem between the two shapes,2. use the correspondences to estimate an aligning transform, and3. compute the distance between the two shapes as a sum of matching errors between corresponding points, together with a term measuring the magnitude of the aligning transformation.As a key advantage no special landmarks or key-points are necessary to solve for correspondences between points on the two shapes. Thus The shape context descriptor is tolerant to all common shape deformations. It has wide application in digital recognition, trademark retrieval, point set matching and image registration.In this paper, the edge detection of the wavelet multi-scale and shape context have been done some theoretical research and algorithm implementation, besides, the experimental results are also given. The good results show that this method is used to recognize the shape of the target is feasible. The experimental data shows that the shape context operator still exist some deficiencies in the matching process, for example, How to recognize the target according to the matching results etc. This paper gives some improvement methods from the matching rate, the matching cost, and the minimum bending energy three aspects etc., respectively, setting threshold to estimate the degree of matching and recognition results. The experimental results show it can reduce the iteration time and quickly get the optimal matching results.