Research On Object Recognition Based On Local Neural Response

With the development of pattern recognition and computer vision, object recognition has been widely used in military, aerospace, navigation, astronomy and intelligent video surveillance, and it has always been concerned by domestic and foreign researchers. Identifying performance goals depends on the choice of similarity measure, deep learning and manifold learning theory is used to extract target feature representation, while the effectively template selection method can improve the accuracy of image feature matching. The main contents are as follows:Firstly, the feature extraction and feature extraction method based on the selected template, describes the purpose and conditions of each feature extraction, and discrimination on the basis of a brief description of the good performance of several common feature extraction algorithm. Template-based feature extraction method in the target identification process for processing and feature extraction of target information to have a better image representation, so that the target identification can get higher accuracy.Secondly, local neural response characteristics extraction algorithm uses depth study and manifold learning algorithms to extract target features to make feature information indicates more representative. Scale and complexity of the algorithm is rotation invariant, and obtained by the partial coding salient features of partial images, maximize joint operation remains extracting features of translation invariance in target identification process can improve classification accuracy, irregularly distributed processing Data and efficient handling of high-dimensional data.Then, based on local neural response presents a good feature selection algorithm represents a template capabilities. The algorithm uses the tag information to obtain a smaller number of training samples and has a strong ability to distinguish the template, the local neural response more suitable image recognition. By carrying out experiments on the standard image database, the result shows the algorithm in the case of a small amount of template image recognition accuracy can still improve local nerve reactions.Finally, the GUI interface using MATLAB for simulation of human face recognition system. The result shows that, based on local nerve reflect templates selected algorithm used for local deformation and complex background of robust, can achieve better target recognition.