摘要
针对轮毂识别系统前期图像特征提取误差较大时分类准确性降低的问题,提出了基于改进粒子群算法优化BP神经网络的轮毂识别模型。在标准粒子群中引入遗传算法的变异因子、惯性权重、时间因子、速度边界限制和反弹策略,以改进粒子群算法,从而提高寻找最优阈值与权值的性能。经过与不同算法的对比数据看出,采用改进粒子群优化BP神经网络算法的分类识别率比其他算法提高了9%左右,且收敛速度、收敛精度均有提高,证明了所提IPSO(improved particle swarm optimization)算法的有效性。
For the problem that the low classification accuracy caused by image feature extraction with large error in the hub recognition system,a hub recognition model based on the BP neural network optimized by modified PSO is proposed.Variation factor related to the genetic algorithm,inertia weight,velocity boundary limit,bounce strategy and time factor are introduced into the standard particle swarm to improve the particle swarm optimization algorithm,further improving the efficiency to find the optimal threshold and weight.The data of experiments and comparisons show that the classification accuracy of BP neural network optimized by modified PSO is about 9%higher than other algorithms.The convergence speed and precision are improved,which proves that the improved particle swarm optimization(IPSO)algorithm is effective.
作者
葛艳
刘杏杏
谢俊标
GE Yan;LIU Xingxing;XIE Junbiao(Institute of Information Science and Technology,Qingdao University of Science and Technology,Qingdao,Shandong 266061,China)
出处
《中国科技论文》
CAS
北大核心
2019年第7期773-777,788,共6页
China Sciencepaper
基金
国家自然科学基金资助项目(61273180)
山东省高等学校科技计划项目(J14LN74)
关键词
粒子群改进算法
BP神经网络
轮毂识别分类
特征提取
modified PSO
BP neural network
hub recognition and classification
feature extraction
