Design And Implementation Of Specific Target Recognition System And Its Database

With the gradual increase of digital image information,the specific target recognition system has been playing an irreplaceable role in every field.Confronting a great deal of image information,it has been one of the most significant assignment to identify the specific target in the image and classify them.The thesis aims at the current problems in target recognition system,and constructs a complete set of target recognition algorithm and the matched image database,complete to identify a variety of specific target and store them.In the aspect of target recognition algorithm,this thesis proposes a scheme that combines Canny edge detection algorithm and Haar-like classification recognition algorithm,and improved Canny edge detection algorithm and SIFT matching algorithm.Firstly,the scheme use AdaBoost classifier which based on Haar-like matching algorithm to recognize the image,so as to obtain the sub-images including specific object,then extract the edge information of the sub-images using the improved Canny algorithm,and the improved SIFT algorithm is used to complete the similarity matching,thus identifying specific targets in the image.Aiming at the phenomenon of poor filtering effect of Canny algorithm,this thesis selects the bilateral filter to replace the original Gauss filter on the basis of testing a variety of filtering,effectively decrease the intensity of peripheral noise,and improve the precision of target recognition.Meantime,considering that SIFT algorithm cannot find enough feature point in edge image,linear-feature extraction algorithm based on the Hough Transformation is used to extract the feature lines from edge images and increase the feature points,so as to improve the matching effect.In the management of database,base class table is established depending on the types of specific target.For each different sample,a Haar-like features database,an edge features database and a image database is established.This thesis accomplishes target recognition system and database system.The target recognition system is able to identify the input image with the accuracy higher than 90%,and is also able to process large image efficiently.The image which has been recognized is stored by database system depending on the type of the target in the image,the training sample and the feature of the specific target are also stored in the database.