基于模糊判别成分分析法的高光谱作物信息提取与分类

东北地区是中国主要的玉米种植区，同时也是中国易发生干旱的地区，干旱常态化严重制约着该地区玉米生产的稳定发展。以辽宁省春玉米为研究对象，在明确春玉米不同发育期干旱变化特征的基础上，基于FY-3A、MODIS、春玉米发育期和土壤相对湿度观测等数据，建立春玉米干旱遥感监测指标集，构建各发育期不同土层深度的土壤相对湿度遥感监测模型，并以2000年为例开展了辽宁省春玉米干旱监测的应用研究，结果表明：1993-2012年辽宁省春玉米在各个发育期均有干旱发生，其中1999-2002年为干旱高发期，乳熟期干旱最为严重；多指数协同配合能提高遥感手段对土壤相对湿度的监测能力，其中陆表水分指数对土壤相对湿度监测能力较强，其次是水分指数；利用构建的春玉米各发育期土壤相对湿度遥感监测模型，监测2001-2004年部分发育期和土层深度的干旱状况，总体监测准确率为73.32%；实现了2000年辽宁省春玉米发育期干旱等级动态监测，所得监测结果与当年农业气象观测记录在发育阶段和空间上都有很好的一致性，遥感监测结果正确。因此，此项研究对于大范围准确跟踪监测春玉米干旱，以及提高春玉米生产的防灾减灾能力具有重要意义。

Hyperspectral feature extraction using fuzzy-statistics-based discriminative component analysis method for crop classification

Drought has become a problem that is universally faced by global terrestrial ecosystems. Northeast China is dominated by a temperate monsoon climate and located in an area sensitive to global climate changes, and one of the main impacts of climate changes in Northeast China is manifested as drought in growing seasons. The drought area has a gradual increase trend, and drought has become the main agro-meteorological disaster in this region which is also the main maize planting area in China. Drought normalization seriously restricts the stable development of maize production in Northeast China. So in this paper, we took spring maize in Liaoning Province as an example and made clear the drought variation characteristics in different growth stages by using the data of MODIS, FY-3A, relative soil moisture and growth stages and the methods of RS, GIS and statistical analysis. Then we analyzed the correlation between multi-time scale remote sensing indexes and relative soil moisture in different soil depth, and established the remote sensing monitoring index set and models of relative soil moisture for each growth stage and soil depth. The accuracy of the monitoring models was verified and the application research of monitoring the drought of spring maize in Liaoning Province was studied. The results showed that: From 1993 to 2012, drought occurred in each growth stage of spring maize in Liaoning Province and the highly frequent drought occurred in the period from 1999 to 2002. The drought grade in each year was mainly light drought, and the drought was the most serious at milk stage, followed by heading stage. There was no significant correlation between the monitoring indexes and relative soil moisture at sowing and emergence stage, but there was a significant or extremely significant correlation between them at three-leaf and maturity stage. The monitoring index LSWI (land surface water index) had a strong ability to monitor relative soil moisture, followed by WI (water index). Multi-index coordination could improve the monitoring ability of relative soil moisture by remote sensing means, and the monitoring ability at 10-day scale was generally higher than that at 5-day scale and 20-day scale. Based on the monitoring models of soil relative humidity, the drought condition of partial growth stages and soil depth from 2001 to 2004 was monitored. The overall monitoring accuracy was 73.32%, in which the monitoring accuracy of jointing and three-leaf stage was more than 85%. The dynamic monitoring of drought grades in different growth stages in 2000 was realized. The monitoring results were consistent with the agro-meteorological observation records in terms of the growth stages and space. This study would be very significant to improve the capacity of disaster prevention and mitigation by monitoring spring maize accurately and synchronously on a large range.