Research On Face Recognition Based On Local Visual Model

Vehicle Routing Problem is proved to be a NP-Hard problem. As it is applied widely and solved hard, many domestic and foreign scholars always try hard to explore and study the method of the problem. As a new algorithm which is generated and developed in recent years,NSGAⅡis applied widely and studied because of its efficient in solving multi-objective optimization. But there is little literature about solving VRP through NSGAⅡ.So this paper solves the VRP based NSGAⅡafter studying the current research situation. And it achieves applying the NSGAⅡto the VRP and emulating through software.Then it advances the improvement against the result of emulating.And this paper proves the feasibility of the improvement through comparison.The achievements of this paper are as follows:Ⅰ.This paper achieves to apply the NSGA II for VRP. Analysis and generalizes three object functions:the shortest of the totle driving path, the most punctual of arrival, the least number of the vehicles.And also three constraints:each vehicle can't exceed the load; the number of customers in each path can't exceed the totle number; there is only one vehicle for each customer. Then builds the math model for VRP, and designs every link according the feature of NSGAⅡ:generating coding (using natural number coding); determining initial population (judging if it is overload);evaluating the fitness functions (considering the object functions);selection (tournament rule); crossover (generating the genes for crossover randomly);mutation (generating the point for mutation).Ⅱ.The paper achieves to emulate with software.It codes in Matlab, chooses C101 in Benchmark Problems as the data for experiment. Then it gets the distribution feature of costomers'demond amount and demond timing after the experiment. Further the feature of NSGAⅡwhich is strong solving ability, fine convergence situation and fast convergence rate is got by analysising the convergence genetation and convergence situation of object funtions.Ⅲ. The paper advances the improvement for NSGAⅡ.It indicates the shortcomings of designs of NSGAⅡafter studying the emulation result and the feature of NSGAⅡ.Then it advances the improvement for two links of NSGAⅡ's design. For improving the determining of initail population, it recommends the strategy of Greedy Algorithm.And for improving the crossover operator, it references the idea of Or-opt, changes the original algorithm.By comparing the two algorithms, it verifies the improved algorithm is effective in avoiding premature convergence, raising efficiency and stability.