高空间分辨率遥感影像中地物目标内部光谱信息复杂性的增强,使得传统基于光谱特征值的数据处理方法效果不再显著,影像分割为解决这一问题提供了一种思路,成为当前高空间分辨率遥感影像处理的研究焦点.时刻独立脉冲耦合神经网络具有状态相近、空间相邻神经元相互耦合同步脉冲激发和区域之间神经元脉冲激发时刻独立两大特点,已被应用于非遥感影像分割中,并取得较好效果.本文结合高空间分辨率遥感影像特点,通过对网络参数进行实验和分析,提出一个基于时刻独立脉冲耦合神经网络的高空间分辨率遥感影像分割方法,并利用空间分辨率0.3m的航空影像进行了数据试验,将分割结果进行讨论并与现有时刻独立脉冲耦合神经网络方法和ISODATA方法分割结果进行对比分析.结果表明:时刻独立脉冲耦合神经网络在高空间分辨率遥感影像分割处理中具有很好的应用前景.

High spatial resolution remote sensing images represent the surface of the earth in detail.As spatial resolution increases,spectral variability within the land cover units becomes complex in high spatial resolution remote sensing images,which makes traditional remote sensing image-processing methods on pixel basis such as ISODATA not suitable.Image segmentation that takes spatial information of image into account provides an alternative solution to this problem,and becomes a hot spot in the processing of high spatial resolution remote sensing image nowadays.Temporal Independent Pulse-Coupled Neural Network(TI-PCNN for short) is an improved PCNN,which is a useful biologically inspired image-processing algorithm. It has two properties including a neuron which has the ability to capture neighboring neurons in similar states and regions of neurons which are not connecting with each other,no matter in which states they are,have different pulsing time.These properties of the TI-PCNN ease difficulties of optimal parameters selection process commonly encountered in the usage of traditional PCNN,and make it a useful new tool in non-remote sensing image segmentation.However,due to its heavy computational cost and over-segmentation of objects within the range of low intensity,the original TI-PCNN method is ineffective at segmenting high spatial resolution remote sensing image.By taking account of spatial and spectral characteristics of high spatial resolution remote sensing image,this paper studies the function of parameters in the TI-PCNN and proposes a segmentation method based on the TI-PCNN.A subset of aerial images with spatial resolution of 0.3m is used for experiment and analysis.Segmented result is compared with that of current TI-PCNN method and ISODATA.Result shows that our method can reduce variability within the land cover units to a large extent while maintaining geometric structure in the image.It provides a great potential in high spatial resolution remote sensing image segmentation.