Research On Artificial Bee Colony Algorithm Theories And Its Application In Information Processing

The Artificial Bee Colony (ABC) algorithm is a new swarm intelligence algorithm which has advantages like less control parameters, fast convergence, high searching precision and strong robustness. It has been widely used in scientific research and industrial production field, and has achieved good optimization results. However the classical ABC algorithm also has disadvantages such as the lack of knowledge of global information, weak local search capability, easy to fall into local extreme and so on, these disadvantages lead to insufficiency in both searching precision and efficiency.This thesis focus on different improvement strategies of ABC algorithm to improve its performance and use ABC algorithm in image processing and voice recognition to expand its application field. The main content of this thesis are summarized as follows:Firstly, for the problem of the neighborhood search formula in classical ABC algorithm which cannot effectively use the population information, this thesis present a Global Artificial Bee Colony algorithm based on Crossover operator (CGABC). CGABC algorithm improve the convergence speed through crossover the solutions after neighborhood search and the current global optimal solution, balance the capability of global optimization and local search through select a reasonable crossover operator, meanwhile set up random disturbance to increase diversity of the swarm. The crossover operation parameter is chosen through experiment and analysis. Optimization results of the benchmark functions show the effectiveness of the CGABC algorithm.Secondly, in order to solve the problem in classical ABC algorithm that the solutions generated by scouts are random and lack of knowledge of the global information, this thesis proposes an improved ABC Algorithm based on automatic Chaotic Tube search (CTABC) combined with the Tube Search algorithm (TS) and Tent chaotic mapping. CTABC algorithm add the local extremes into tube table which can avoid repeated search around local extremes, improve the efficiency of the algorithm and enhance the ability of global optimization. Scout selects new solution from candidate solutions generated by local extreme using Tent chaotic search to make the stagnant solution jump out of local extreme and continues to evolve, which can improve the convergence speed and the local search capability of the algorithm. Optimization results of the benchmark functions indicated that convergence speed, precision and robustness of CTABC algorithm are obviously better than that of basic ABC algorithm and TS algorithm.Thirdly, in order to extract the exact edge of image, this thesis transform edge detection problem to an optimization problem with a large number of local extreme values and using CTABC algorithm to solve it. The new solution which created by the scout using chaos tube search is correlated with the original local extreme, which makes the edge points more concentrated and avoids the repeated search for local extreme that can find more edge points in limited cycles. Experiments showed that the CTABC algorithm can extracts edge with better continuity, more accurate and less noise compared with ABC algorithm and classic image edge detection method.Finally, in order to solve the problem in LBG algorithm that it depends on the initial codebook, we propose a new codebook generating method which is alternate using ABC algorithm and LBG algorithm. It improves the quality of codebook and reduces the distortion of optimal codebook. Experiments showed that ABC-LBG algorithm increase recognition ratio of isolated word recognition system.