Research On Fast Target Recognition Algorithm In Optoelectronic Reconnaissance

Using Bayesian Networks and Dempster-Shafer theory as a theoretical tool, this paper focus on the fast target recognition algorithm, the work was done as follows:1. A Bayesian Networks interval estimation algorithm (BNIE) was proposed to solve parameter learning problem, it used the form of interval parameter estimation instead of traditional point ways, generalized the point value to the form of interval estimation, decreased the over-fitting data phenomenon in certain extent as the prior knowledge was rare.2. An Interval Reasoning Algorithm (IRA) was put forward to solve estimation reasoning problem, it used observed variables as an input, the inference procedure among Bayesian Networks nodes was in the form of interval, and the final output was interval too, this method extended the probability’s propagation and reasoning way which always been done in the point value form.3. An Ensemble Reasoning Algorithm (ERA) was presented to solve inference problem, it was inspired by many experts working together to make a decision. The algorithm used multi section mean method to discretize the interval value firstly, and then assigned these values to the homogeneous Bayesian inference network, integrated the final inference results in some way to obtain the type of target recognition in the end.4. A Weighted Fuzzy Logic Dempster Algorithm (WFL) was advanced to solve IFF problem, it used weighted fuzzy logic as the proposition antecedent, utilized an improved weighted Dempster combination rule to achieve evidence fusion, reached“whether recognition”and“how to identify”problems, this method has a certain significance for the design of target recognition system.The simulation results show that the algorithms proposed in this paper were fast and stable, reached the expectation index, and have practical value.