Research On Infrared And Millimeter Wave Compound Detection Technology For Aerial Small Targets

With the rapid development of anti-armor guided weapons,armoured vehicles are facing more and more serious air threats.The vehicle borne aerial small target early warning system is of great significance to improve the survival ability of armoured vehicles.Most of these early warning systems use single band detection technology.However,infrared and millimeter wave compound detection technology are based on heterogeneous sensor information.It can effectively improve the target feature dimension.Compared with single band detection technology,it has lower false alarm probability and stronger adaptability.In order to improve the detection ability of vehicle borne early warning system for aerial small targets,the thesis studied the infrared information processing,millimeter wave information processing,infrared and millimeter wave compound detection technology.The main research contents are summarized as follows:(1)A design idea of infrared and millimeter wave compound detection system for aerial small targets is proposed.According to the electromagnetic environment of the battlefield and the operational requirements,two kinds of composite detection methods are designed:guiding and parallel.(2)The infrared characteristics of target and background clutter are analyzed.Aiming at the problems of high clutter intensity and hard-to-remove false alarm in the complex infrared image,a small infrared target detection method based on local gradient characteristics is proposed.The Local Gradient Second-order origin Moments(LGSM)is proposed to distinguish target from high-intensity clutter.Two thresholds are set on LGSM value and Laplace-Gaussian filtered value respectively for hysteresis threshold segmentation,which can effectively reduce false alarm probability while ensuring detection probability and algorithm efficiency.(3)The thesis analyzes the millimeter wave characteristics of typical aerial targets and the shortcomings of millimeter wave phased array radar in estimating target velocity characteristic parameters,proposes a velocity characteristic parameter estimation method based on u-domain modified Newton method.The grid search in Fractional Fourier transform(FRFT)domain is used as rough estimation,and the peak parameters u₀ and ?₀ in FRFT domain are accurate estimated by u-domain interpolation and ?-domain Newton method,and the velocity and acceleration of the target are derived.The simulation results show that the proposed estimation can reduce the Root Mean Squared Error of Linear Frequency Modulation signal.(4)In view of the situation that the millimeter wave radarcannot search the airspace due to the battlefield environment and operational requirements,a guided mode compound detection method based on visual attention mechanism is proposed.In this thesis,we use naive Bayes classification to get the focus area,and construct the global saliency map of the infrared image.According to the mastery of target feature information,two decision methods,fuzzy pattern recognition and fuzzy binary classification decision tree,are proposed to classify the target by using the millimeter wave echo feature of the target.The simulation results show that the guided compound detection can fuse multiple infrared characteristic saliency maps,and provide guidance information for millimeter wave radar,and remove false alarm through the verification of radar.(5)When the infrared search and tracking system and millimeter wave radar work in parallel,aiming at the heterogeneity of infrared image and millimeter wave echo information,the thesis proposes a parallel mode compound detection method based on heterogeneous data processing,which transforms the infrared images into feature vectors,integrates the infrared images and millimeter wave feature vectors into joint feature vectors,and carries out classification based on support vector machine.Heterogeneous data processing is optimized in the three stages of infrared feature extraction,data association and feature selection.The infrared feature extraction based on human vision system is proposed to remove redundant data in infrared data set.The grid global nearest neighbor data association is proposed to simplify the construction of correlation matrix.The mutual information feature selection method based on prior information is proposed to improve the detection probability when the difference between training set and test set is large.Simulation results show that heterogeneous data processing can improve the classification performance.