Font Size: a A A

Research On Data Fusion Method Based On Ensemble Learning

Posted on:2018-08-18Degree:MasterType:Thesis
Country:ChinaCandidate:Y W XuFull Text:PDF
GTID:2438330602961082Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Data fusion has solved some information problems difficult to deal with,and has improved the accuracy of the information system.However,Data fusion has not yet formed a basic theoretical framework and an effective and generalized model.The research on specific algorithms is still in the primary stage.As an important research direction in the field of data fusion,image fusion has wide application prospect in terms of the military and the civil use.In this paper,the image fusion was used as the research object,and the data fusion model and algorithms were studied from the levels of the data fusion and decision-making fusion,combined with the knowledge of the Kalman algorithm and ensemble learning.The paper presented a kind of data fusion model and algorithm which based on adaptive Kalman.Based on the algorithm which using the Kalman algorithm to complete image denoising,this algorithm used the sensor confidence calculation method to distribute the weight of different sensors,and adjusted the measurement error covariance matrix of the sensors to improve the accuracy of the data fusion.The paper also presented a kind of fusion model and algorithm which based on the ensemble learning.Firstly,the evaluation of base-classifier algorithm was used to eliminate the unqualified classifiers,and then the classifiers were sorted by the accuracy and difference.The classifier with the highest precision and the biggest difference was obtained.Then,the FSE fusion algorithm(Fast Strong-classifiers Ensemble)was used to break the sample distribution and re-samplinged to make the classifiers pay more attention to the samples difficult to learn,and thus determined the weight of each classifier.For the proposed models and algorithms,the experiments were done to verify them,and the results showed them effective and feasible.Finally,based on the proposed models and algorithms,the overall framework of data fusion based on ensemble learning was designed,and the data fusion system with image fusion was realized.The system enabled the integration of the data fusion and the decision-making fusion,and provided fusion images which having higher quality for military or industrial areas,and improved the image recognition in the accuracy and generalization at the same time.
Keywords/Search Tags:Data fusion, image fusion, data fusion model, Kalman algorithm, ensemble learning
PDF Full Text Request
Related items