摘要
针对BP网络收敛速度慢、计算量大等缺点,采用学习样本产生的总误差调整权值,以提高BP算法的速度。文中通过仿真说明其算法的有效性,对改进BP算法用于纸病检测有一定的参考价值。实践表明,改进的BP网络能够满足在线纸病检测要求,平均识别率达94%以上。
According to the BP network was slow convergence and large amounts of calculation,this paper proposed the line detection method of eliminate the influence of sample order and adaptive correction weights of BP network method. To improve speed and efficiency of the training at the same time,the recognition process was more stable and reliable,more accurate identification results. The experiments show that improved BP network ean meet the re- quirements of online testing ,and recognition rate can reach more than 94%.
出处
《自动化与仪表》
北大核心
2013年第1期6-9,共4页
Automation & Instrumentation
基金
国家自然科学基金项目(30972322)
关键词
纸病
BP网络
模式识别
在线检测系统
paper defect
BP Network
pattern recognition
on-line detection system
