Multi Model Fusion of Binary Information Mining on Impact Velocity Prediction of Abnormal Projectile

Abstract

Aiming at the problem of predicting the impact velocity of abnormal projectile,RGB image information was extracted from the radar waterfall before and after the projectile impacting,and digital information of key points was extracted from the radial velocity before and after the projectile impacting. Firstly,binary information such as image information and digital information were used as feature vectors, corresponding to the measured impact velocity as the target vector. Based on training data,support vector regression machine models were established to mine nonlinear features of the impact velocity. By introducing test data into the established model,the corresponding impact velocity can be predicted. At the same time,a GM(1,1)grey model was established using training data to mine linear features of the impact velocity,and then the corresponding target velocity of the test data can be predicted. Secondly,the fitting values of the three models to the training data were constructed as feature vectors,corresponding to the measured impact velocity as the target vector,and a genetic algorithm was established to optimize the LSSVM model. Finally,the predicted values of the three models on the test data were substituted into the established genetic algorithm optimized LSSVM model,and the predicted target velocity of the model was obtained. The results show that compared with support vector regression machine,multiple linear regression and random forest,genetic algorithm optimized LSSVM has higher prediction accuracy,and the error is far less than 1%｡,which can be used as the prediction model of impact velocity of abnormal projectile.