基于分形特征的高标准农田遥感分类方法研究
Remote Sensing Classification for High Standard Farmland Based on Fractal Characteristics

摘要： 目前全国高标准农田面积数量已具一定规模，由于人工解译的工作效率较低，如何实现对全国大面积的高标准农田建后利用情况进行实时、精准遥感监测成为亟待解决的问题。由于监测面积大，精度要求高，迫切需要研究一套遥感自动监测方法在全国推广。以广东省东莞地区作为研究区，选择2017年2月15日的高分二号遥感影像，基于分形图像分割并结合BP神经网络对区域高标准农田进行分类，并加以人工解译和实地验证。 结果显示，该分类方法总体精度为 80112 2%，Kappa 系数为0761 1。表明分形图像分割结合BP神经网络的遥感分类方法总体精度较高，能较好地满足高标准农田建后利用情况遥感监测的需求。此方法可以在全国范围推广应用，为高标准农田建成后的实时监管提供技术支撑。
Abstract: The highstandard farmland area in China has currently reached a considerable scale. The low efficiency of manual interpretation has called for more precise and realtime remote sensing monitoring for the construction of large scale farmland in the country. Because of the large monitoring area and high precision demand, a set of automatic remote sensing monitoring classification method is needed to develop. The study area is located in Dongguan area of Guangdong Province, South China. Using GF2 remote sensing images (15th February, 2017), the fractalbased image segmentation combined with the BP neural network remote sensing classification method is studied, which is supported by the artificial interpretation and field verification. The experimental results show that the overall precision of the classification method is 80112,2% and the Kappa coefficient is 0761,1，which indicates that the overall precision of fractal image segmentation and BP neural network remote sensing classification method is higher. This can better meet the needs of remote sensing monitoring after the construction of the high standard farmland.A nationwide adoption of this method can provide technical support for the realtime monitoring of the high standard farmland