The Application Of Fruit Fly Optimization Algorithm In Face Recognition Technology

When using Face recognition,we need appropriate accuracy rate,precision ratio and recall rate in different cases and consider the influence of light,posture and the degree of exposure.Therefore,it is important for us to determine appropriate parameter in different cases.In this paper,we put forward to a kind of method which combines the group intelligent algorithm and deep neural network and can adjust appropriate parameter in different cases automatically.Green hand can master our method easily so that we can reduce maintenance costs significantly.Our main work is as follows:1.We analysed and compared MTCNN(Multi-task Cascaded Convolutional Networks)and FaceNet carefully.2.We proposed a double population fruit fly optimization algorithm with logistic transform(DFOAL)which is based on the Fruit Fly Optimizational Algorithm(FOA).We introduced multi-group cooperative evolution strategy and the dynamic step size design based on the Logistic function into DFOAL and make up for the disadvantage that the traditional Drosophila algorithm can only gain the local optimal solution in the optimization calculation.The classical test function data shows that our method are significantly effective.3.We apply DFOAL to FaceNet face recognition system,analyzed the optimized parameters optimization methods of Drosophila optimization algorithms and apply DFOAL to Tensorflow deep learning platform with Python language.Simulation data shows that DFOAL can fast adapt to new project without relevant experience.