摘要
简化和改进了脉冲耦合神经网络(PCNN),建立了基于时间索引图的脉冲耦合神经网络海冰 SAR 图像分类器,用于海冰 SAR 图像的分割和海冰分类。在此基础上建立了基于人工解译的半自动海冰分类判读系统。将发展的分类器用于辽东海湾冰探测,结果表明这个分类器能够区分海冰和海水,识别不同海冰类型,且具有高效率。为了选择适合辽东湾海冰分类的 PCNN 参数,分析了链接半径、链接强度和索引图等级等参数,给出了各参数合适的取值范围及调节原则。
The pulse-coupled neural network (pCNN)was simplified and improved,and a PCNN SAR sea ice image classifier based on time-index images was proposed.Then a semi-automatic sea ice classification system with a man-machine interface was built based on the artificial interpretation.The classifer proposed was used in the detection and classification of sea ice in Liaodong bay.The results show that the classifer can distinguish sea ice from open water and recognize the different ice types with high efficiency.In order to select PCNN parameters suitable for classification of sea ice in Liaodong bay,the linking radius,the linking strength and the time-index rank were analyzed,and their value ranges and regulatory principles were given.
出处
《高技术通讯》
CAS
CSCD
北大核心
2008年第2期190-195,共6页
Chinese High Technology Letters
基金
国家863计划(2001AA633080)资助项目