Recognition And Matching Of Cuboid Shape Object In The Image Data

In recent years,with the rapid development of the network and increasing of the size of the image database,image retrieval has begun to enter people's life.The security problem of luggage and other objects in important occasions such as airport,train station,subway and others,requires us to use matching technology of object to retrieve massive data of monitoring system.The needs of Image retrieval of people's daily life,for example,the object retrieval what cannot be described by the use of text,can be achieved by the use of matching technology of object based on image.In real life,from the construction,transportation,home to many fields of daily necessities,food,entertainment activities,there are a large number of cuboid shaped object.This makes the study of object recognition of cuboid shape object be particularly important.But at present the application of object recognition is limited to 2D object recognition,such as word recognition,license number recognition,face recognition and fingerprint recognition.For there are many new problems need to be studied.Therefore,this paper focuses on the research of recognition and matching techniques of cuboid objects.The image of same object with different perspective in shape and pattern will have a larger deformation.The use of existing technology is difficult to solve the matching problem of image from different angles of view.Model based method,appearance based methods and methods based on local feature matching which study on the method of 3D object recognition are difficult to solve the situation of the larger changes of objects view.The recognition of 3D object can be used to identify depth image took by a specific camera,which is difficult for the existing two-dimensional image matching.In order to solve the matching problem under the change of object perspective,this paper presents a cuboid shape object matching model,the model of comprehensive utilization of the cuboid object pattern feature and shape feature,pattern for cuboid object pattern matching,shape features to adapt to the change of angle shape feature.In the aspect of pattern matching feature,use segmentation matching algorithm.Firstly,label out the frame of cuboid object and split the object into three surface,then use the pattern matching ASIFT algorithm,and give full play to the advantages of the ASIFT matching algorithm in 2D image matching.In the aspect of shape feature matching,using reverse perspective method for frame size ratio with cuboid object,restore the true size ratio of cuboid objects,which is not affected by the changes of view.This paper design a matching model by using two types of information,adjusting the weights of two kinds of information according to the actual characteristics of the image,the effective adaptation of pattern features and characteristics of different situations of framework to further improve the accuracy of the algorithm.This paper reference to standard subset of SUN image database,consider four influence factors,the definition,angle,distance and pattern complexity,to construct the cuboid object standard data set.And we use this data set for comparing the matching accuracy of the proposed algorithms and evaluate matching accuracy the proposed algorithms under different influence factors of the proposed algorithms.Experiments results show that the matching accuracy of matching algorithm of this paper is higher than the other two kinds of matching algorithm in the various factors.In the end,the experiments in the image data with the comprehensive effect verify the effectiveness of the proposed algorithm by this paper.The problem of recognition and matching of cuboid objects proposed in this paper is significant.And the matching algorithms that we proposed can be applied to detection of cuboid objects,picture retrieval,and the other fields.