| The emergence of big data,cloud computing and other technologies has pressed the shortcut key for the innovation and development of information technology in the digital era.The way of data and information processing is also constantly updated.In the field of machine learning,the development pace of classification algorithm is also accelerating.The improved k-means hybrid clustering algorithm and a semi supervised learning classification algorithm based on the minimum and maximum neighborhood order composition are proposed.Hyperspectral images of cucumber leaves are collected by Hyperspectral spectrometer and classified after image spectral data processing.The performance of different algorithms is comprehensively evaluated by comparing and analyzing the relevant data extracted from the experiments of different types of classification algorithms.The specific research work is as follows:1.The average spectral data of ROI were extracted from the hyperspectral image of cucumber leaves,and the multi-element scattering correction(MSC)and standardization processing(SNV)were used for preprocessing to eliminate the scattering effect and reduce the irrelevant spectral information.Principal component analysis(PCA)was used for dimension reduction to extract the principal components with more information expression and eliminate the redundancy of data.2.An improved unsupervised classification method is proposed,which is based on the improved hybrid clustering algorithm of artificial bee colony and K-means,and combines the global artificial bee colony algorithm with K-means + + algorithm,so that the algorithm has the characteristics of global optimization,optimization of the initial clustering center and fast convergence.Experiments are conducted on iris,wine,glass and balance sacle datasets in UCI database.The results show that the proposed hybrid clustering algorithm has good stability and the clustering effect is improved.3.A graph based semi supervised learning classification algorithm is proposed,which combines KMM with bb-llgc,that is kmm-bb-llgc algorithm.It takes into account the symmetry of edges and the connectivity of the whole graph,and simplifies the objective function on the graph,so that it is not affected by parameters.Experiments are carried out on UCI database,and compared with knn-llgc,knn-bb-llgc and kmm-llgc.The experimental results show that the proposed method has higher classification accuracy and can achieve efficient and accurate classification of samples.4.The improved unsupervised learning,semi supervised learning and traditional supervised learning three different types of classification methods in cucumber leaf data experiment,comprehensive comparative analysis,to explore the applicability of different classification algorithms in hyperspectral image data and verify the practicability of the improved algorithm... |
