The Optimized Statistical Modeling For BRDF And Its Application

Bidirectional reflectance distribution function (BRDF) can be used to describe thescattering and radiation characteristics of rough surfaces and it has importantapplication value in the field of target detection, tracking, recognition, feature extraction,stealth technology and so on. In this paper, the measured data of BRDF is optimizedthrough different algorithms and the detailed work is as follows.Firstly, some BRDF statistical and empirical models for the surfaces of severaltypical targets are analyzed, among which the five-parameter model is discussed indetail.Then the experimental data of different samples are optimized with thefive-parameter BRDF statistical model on the basis of the introduction of the geneticalgorithm (GA), the simulation annealing algorithm (SA) and the particle swarmoptimization algorithm (PSO) and the concrete statistical models are created. Themodels can agree to the experimental data well.Meanwhile, given that BRDF is affected by many factors and its complexnon-linear characteristic, the fitting and forecasting results obtained by traditionalmethods often cannot meet the requirements. This paper proposes an approach toestablish the BRDF model by least squares support vector machine (LS-SVM). Thebasic principle of LS-SVM is introduced and then a GA is applied to optimize the modelparameters to raise the fitting accuracy and enhance its generalization ability. At thesame time, the BRDF models based on BP neural network are built and a comparison ofthe data calculated by the two kinds of models is carried out．The result shows that theLS-SVM model has good integrative performance, strong generalization ability andhigh precision and is of great value in engineering applications.In addition, the near field radar cross section (RCS) of two complex targets iscalculated based on the five-parameter BRDF model and the LS-SVM modelrespectively. The calculation results indicate that it has the advantage of intuition andhigh computation efficiency to study the scattering characteristics of complex targetswith BRDF statistical models.