基于密度自适应粒子群优化的图像可用度分类  

Research on Availability of Satellite Images Based on Density Adaptive Particle Swarm Optimization

作　　者：吴兴华[1] 郑红[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《计算机仿真》2010年第9期274-277,285,共5页Computer Simulation

基　　金：国家自然基金(60543006)

摘　　要：研究图像优化评估方法,针对卫星遥感图像可用度模糊神经网络分类器学习过程中存在的学习速度慢且容易陷入局部极小值的问题,为提高评估的准确率,提出了一种密度自适应粒子群优化(DensityAdaptivePSO,DAPSO)算法。通过信息熵建立粒子群密度评估测度并更新粒子群的粒子数量,以保持种群的多样性,有利于提高算法全局寻优能力。利用120幅中巴资源02B卫星(CBERS-02B)高分辨率相机(HR)采集的遥感图像,比较BP算法,标准粒子群优化算法以及密度自适应粒子群优化算法的模糊神经网络可用度分类器识别结果,表明新的算法收敛速度更快,有更强的全局寻优能力,并为应用提供依据。Due to the shortage of learning algorithms available in Adaptive Neuro-Fuzzy Inference System （ANFIS）,a density adaptive particle swarm optimization（DAPSO）algorithm was proposed here. In order to improve the ability of global optimization,we used information entropy to calculate the density of particle swarm,and updated it to maintain the population diversity. We also extended particle swarm to improve the ability of local optimization. Using 120 images obtainedd by high-resolution camera in CBERS-02B satellite,we compared BP algorithm,PSO algorithm and DAPSO algorithm,which were used to train the parameters in ANFIS. The results show that DAPSO converges faster and has a stronger ability of global optimization.

关 键 词：自适应神经模糊推理系统 信息熵 粒子群优化 图像可用度评估 遥感图像 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]