Image Matching And Recognition Based On Hyper Graph

Image is a true reflection of objective things.It is usually an important carrier for people to obtain information about things.Research on images has attracted more and more attention.Image matching and recognition is an important branch of image research.It has many applications in the recognition of targets and the detection of motion trajectories of objects.Therefore,the research on matching and recognition of images has important value and significance.In the research of image matching and recognition,one of the effective ways is to extract features from the image,and then construct corresponding graph models according to the features,and then analyze the images.But most graph models are simple graphs based on unary or binary relationships.Simple graphs can only describe the relationship between two features.In order to make the connections between features more closely,some researchers have proposed hypergraphs instead of simple graphs to describe the relationship between features.Hypergraphs are extensions of simple graphs,hypergraphs can contain multiple vertices,so it contains more information than simple graphs.Using hypergraph instead of simple graph to describe the complex relationship between interested things can not only preserve the complex relationship between things,but also ensure the accuracy of describing the relationship between things to a certain extent.Therefore,this thesis studies the hypergraph,analyzes the properties of the hypergraph,and applies the hypergraph to the two directions of image matching and gait recognition.The following are the main contents and results of this thesis:(1)An image matching algorithm based on directed hypergraph is proposed.Firstly,the algorithm constructs 3 uniform hypergraph in the two feature points set to be matched,calculates the weight of the triples contained in each hyperedge,and then uses these weights to construct the weighted adjacency tensor.Finally,image matching is realized by convex-concave relaxation algorithm.The simulation and real experiment results show that the algorithm can improve the accuracy of matching,and has a good matching effect for complex image transformation.(2)An touch gait recognition algorithm based on hypergraph is proposed.Firstly,a complete cycle of dynamic gait data is processed,and a complete plantar pressure image is obtained by weighted averaging.Secondly,the processed plantar pressure image is divided into regions and the pressure maximum point and the pressure center point of each region are used as feature points of the plantar pressure image,these feature po:ints are used to construct the hypergraph and calculate the tensor.High-order singular value decomposition of tensors(HOSVD)extracts high-order spectral features as touch gait features.In order to further improve the gait recognition rate,the Fourier descriptor is used to describe the contour of the plantar.Finally,the support vector machine(SVM)is used for classification and recognition.Experiments show that this method has a higher recognition rate than other methods for both static and dynamic gait data.