基于模糊度与简化PCNN的遥感图像变化检测

Change detection method based on simplified PCNN with minimu fuzziness

遥感图像变化检测必须充分考虑不同时间、不同环境背景等具体情况对于图像的影响,尽可能消除这些"伪变化"影响,以获得比较客观的感兴趣区域变化检测结果。针对遥感图像这一特点提出了一种基于脉冲耦合神经网络的变化检测法。该方法利用脉冲耦合神经网络实现多时相遥感图像的多层次分类,将分类结果进行差值比对,即可检测出感兴趣的变化区域。并提出根据最小模糊度准则,自动确定PCNN循环迭代次数和最佳阈值。实测数据的实验结果表明该方法的变化检测效果优于基于最大熵变化检测算法。

Remote sensing image change detection must fully consider the effect of different time and different environmental background,as possible as to eliminate these ＂false change＂ effect,in order to obtain more objective results of interest region.A novel change detection algorithm based on pulse coupled neural networks（PCNN）is proposed.Different regions in multi-temporal remote sensing images are hierarchically classified by PCNN.Comparing the classification result,the interested change can be extracted.Least fuzziness is used to determine the cyclic iterative times and the best segmentation threshold.The performance on real data shows that our algorithm is superior to the PCNN based on the criterion of Shannon entropy.