Research On Image Recognition And Application Based On High Dimensional Data Reduction Algorithm

In the era of big data,dimensionality reduction of high-dimensional data is an effective form of data processing,which can transform data of large scale and high complexity into low-dimensional data that is easy to analyze,calculate and store,and effectively improve the efficiency of data processing for users.Based on the field of face recognition,this paper discusses the background,significance and principle of dimensionality reduction of high-dimensional data,and discusses the application form of dimensionality reduction algorithm based on empirical research,so as to provide some reference for the optimization and application of dimensionality reduction algorithm in face recognition system.Based on the introduction of the concepts,technologies and methods of face recognition and dimensionality reduction of high-dimensional data,this paper proposes a spacing discriminant projection algorithm for the deficiency of global and local considerations in the dimensionality reduction of boundary Fisher analysis and maximum spacing criterion,and proposes the NMF+SDA algorithm for the problem of lax constraints in the dimensionality reduction of NMF non-negative matrix factorization.The effectiveness and practicability of the improved algorithm are verified by face recognition experiments.The findings:Jaffe and CK+datasets were used to conduct a single variable face recognition experiment on principal component analysis,boundary Fisher analysis,maximum spacing criterion and spacing discrimination projection.The experimental results show that the principal component analysis has the worst face recognition effect and the spacing discrimination projection recognition rate is the highest no matter how the number of training samples changes.YALE data set or ORL data set is used to carry out face recognition single variable recognition experiments on principal component analysis(PCA),singular value decomposition(SVD),NMF non-negative matrix decomposition(NMF)and NMF+SDA.The experimental results show that the face recognition rate from high to low is NMF+SDA,NMF non-negative matrix decomposition(NMF),principal component analysis(PCA)and singular value decomposition(SVD).