IBAS-LMBP神经网络算法在图像压缩中的应用  

Research on Application of IBAS-LMBP Neural Network Algorithm in Image Compression

作　　者：王海军 金涛 门克内木乐 WANG Hai-jun;JIN Tao;MENKE Nei Mu-le(Department of Mathematics and Computer Engineering,Ordos Institute of Technology,Ordos 017000,China;Department of Information Engineering,Ordos Institute of Technology,Ordos 017000,China)

机构地区：[1]鄂尔多斯应用技术学院数学与计算机工程系,内蒙古鄂尔多斯017000 [2]鄂尔多斯应用技术学院信息工程系,内蒙古鄂尔多斯017000

出　　处：《火力与指挥控制》2021年第5期12-17,共6页Fire Control & Command Control

基　　金：国家自然科学基金(61741509);内蒙古自治区高等学校科学研究基金资助项目(NJZY19260)。

摘　　要：鉴于天牛须搜索算法在对多维非线性优化问题求解时容易出现局部收敛现象,从而导致无法求出全局最优解,设计了改进天牛须算法(Improved Beetle Antennae Search,IBAS)。该算法中在天牛位置计算时引入惯性权值,设计了基于自适应步长衰减系数,增加天牛在位置计算时的自适应能力,为了控制算法优化过程解的取值范围,参照粒子群算法对单点位移进行限制。实验结果表明,在相同误差精度、相同迭代次数情况下,将改进天牛须算法与LMBP算法相结合建立IBAS-LMBP算法模型并应用于图像压缩,模型的运行效率明显高于GA-LMBP、PSO-LMBP及BAS-LMBP算法模型。In view of the local convergence of the beetle antennae search algorithm in solving multi-dimensional nonlinear optimization problems,which leads to the inability to find the global optimal solution,an beetle antennae search improved algorithm is designed.In this algorithm,the inertial weight term is introduced in the calculation of the position of the beetle,and the adaptive step attenuation coefficient is designed to increase the adaptive ability of the beetle in the calculation of the position.In order to control the value range of the optimization process solution of the algorithm,the single point displacement is limited by reference to the particle swarm optimization algorithm.The experimental results show that under the same error accuracy and the same number of iterations,the improved IBAS algorithm and the LMBP algorithm are combined to establish the IBAS-LMBP algorithm model and applied to image compression.The operating efficiency of the model is obviously higher than that of GA-LMBP,PSO-LMBP and BAS-LMBP algorithm models.

关 键 词：图像压缩 天牛须算法 莱文伯格·马夸德算法 BP神经网络 仿生算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]