| With the rapid social, politics and economic development, a growing number of social events and crimes take place in the world, and more and more attention are paid on video surveillance. Therefore, the study of object classification is of great practical significance to the video surveillance.This article studies similarity network fusion(SNF) algorithm, and based on previous studies, we propose a fast similarity network fusion algorithm that could be used in the online video object classification. By constructing sample-similarity network for each data type and then efficiently fusing these networks into one single similarity matrix, SNF can efficiently predict labels for new samples based on the constructed network. The main problem of online object classification using SNF is its complexity. We propose fast SNF(FSNF) in this work consists of two main steps: dividing the matrix into two parts and replacing the main part of testing matrix using the same part of training matrix and this replacement can reduce most of the computation load. Experiments show that compared with SNF, the proposed FSNF can gain 13-16 times speed increasing with only 0.5%-0.6% accuracy losing.This article also design an online object classification system based on the FSNF algorithm, and complete software system programming on MATLAB platform. Add the feedback and retrain mechanism into this system to strengthen the classifier under different surveillance scenarios to achieve better classification performance. Experiments are conducted to verify that the effectiveness of the system and FSNF algorithm, and verify that the feedback mechanism significantly improves the classification accuracy.Based on the similarity network fusion algorithm, this article propose a new fast similarity network fusion, a design and implementation of the object classification system based on FSNF. The work has certain practical significance to apply the algorithm to object classification in video surveillance. |
PDF Full Text Request