Research On The Improved Model Of Artificial Bee Colony Algorithm Based On GPU And Its Application

With the rapid development of large-scale parallel computing technology and the continuous improvement of the level of GPU hardware,more and more algorithms with parallel characteristics have been applied to the GPU parallel computing technology in the parallel phase.The artificial bee colony algorithm has the characteristics of partial parallelization,in the process of finding the optimal solution,the particle is relatively independent.On the basis of summarizing the traditional artificial bee colony algorithm on CPU,an artificial bee colony algorithm model based on GPU(PABC)is proposed.PABC algorithm is used to run in the GPU kernel,which is based on the search phase of the bees and the neighborhood search phase and the fitness calculation phase,and the independence of each bee in the artificial bee colony algorithm is used to achieve the purpose of parallel computing.PABC model data initialization,random number generation and other data organization phase running in CPU,and increased the data communication between GPU and CPU.Due to the use of CPU-GPU collaborative work mode,PABC has faster processing speed than the traditional ABC model,and can show excellent performance in the context of large populations.The demonstration to the PABC model in the experiment of 5 test function are also carried on.In addition,the PABC model is applied to the Canny operator image edge detection algorithm and particle filter video tracking technology.By the combination of the traditional Canny operator image edge detection algorithm,the experiment results show that the algorithm can improve the speed of the edge detection algorithm in different population size and iteration number.When the PABC is applied to the particle filter,the improved algorithm is proved to have better performance in the stability,accuracy and efficiency of the improved algorithm based on the 6 standard image sequences.