Research On Image Object Recognition Based On Ensemble Extreme Learning Machine

With the progress of the information age,image recognition technology plays an important role in the development of many fields.The rise of neural network provides a new way for the research of image recognition technology.The learning process of extreme learning machine has the advantages of short learning time and good generalization performance.The extreme learning machine used in the field of image recognition can greatly improve the accuracy of image recognition.On the basis of investigation and analysis of the principle of extreme learning machine,this paper focus on exploring the practical performance of extreme learning machine by optimizing the extreme learning machine and utilizing the image recognition experiments.This paper studies the basic theory of traditional image recognition system and the traditional neural network.Aiming at the defects of traditional extreme learning machine algorithm,we introduce the differential evolution algorithm into the learning process of extreme learning machine.The input weight vector of extreme learning machine and the hidden layer node offset value are optimized.Besides,the performance of the integrated extreme learning machine is tested by the experiment of the corn seed variety recognition and handwritten number recognition.In maize seed identification experiment,aiming at the differences between the varieties of maize,through extracting maize seed's geometry,texture,color feature parameters and through learning characteristic data of different maize based on integrated extreme learning machine,the purpose of identification of maize varieties and classification accuracy judgment is achieved.This paper also studies the influence of the selection of the number of hidden layer units and the characteristic value of the corn seed on the performance of the extreme learning machine.The experimental results show that the integrated extreme learning machine applied to the identification of maize seed varieties can achieve95.77% classification accuracy.In the handwritten number recognition experiment,by extracting the characteristic value of the 35 dimension of the handwritten number,the practicality of the extreme learning machine is further verified.The experimental results show that the integrated extreme learning machine can be effectively applied to thedigital recognition of handwritten digits.