摘要
风灾作为水稻气象灾害之一,严重影响着水稻的生长、产量以及品质。遥感技术作为倒伏作物监测的新途径,给作物受灾面积统计的实时、动态、宏观监测带来新思路,一定程度上弥补了传统方法的不足。以黑龙江省五常市典型水稻倒伏地块获取小型无人机多光谱数据,并基于此数据选取光谱特征、纹理特征、植被指数三个指标对比水稻倒伏前后的差异,接着以水稻倒伏前后的光谱特征、纹理特征结合最大似然法分类,植被指数差异选择最大阈值进行分类,最后对倒伏水稻的提取精度进行评价与分析。研究发现:①植被指数特征相对差异值最小(7%),不适用于准确区分正常和倒伏水稻;光谱特征的相对差异值居中(18%);纹理特征相对差异值最大(28%),最适于区分正常和倒伏水稻。②从提取的水稻面积结果显示可知,基于均值纹理的分类方法的精度最高(94%);基于光谱特征分类方法的精度居中(72%),介于两者(植被指数和均值纹理特征)之间;基于植被指数最优阈值分类方法的精度最低(66%)。该结果可为制定灾后生产管理、防控措施、评估产量损失具有重要参考依据;为动态监测水稻倒伏面积具有重要的借鉴意义。
As one of the meteorological disasters of rice,wind disaster seriously affected the growth,yield and quality of rice.As a new way of monitoring lodging crops,remote sensing technology brought new ideas to the real-time,dynamic and macro monitoring of crop affected area statistics,and made up for the shortcomings of traditional methods to a certain extent.The multispectral data of small unmanned aerial vehicle was obtained from the typical rice lodging plot in Wuchang City,Heilongjiang Province.Based on this data,three indexes of spectral characteristics,texture characteristics and vegetation index were selected to compare the differences before and after rice lodging.Then,the spectral characteristics and texture characteristics before and after rice lodging were combined with the maximum likelihood method for classification,and the maximum threshold was selected for the difference of vegetation index for classification.Finally,the extraction accuracy of lodging rice was evaluated and analyzed.The results show these as follows.The relative difference of vegetation index characteristics is the smallest(7%),which is not suitable for accurately distinguishing normal and lodging rice.The relative difference of spectral characteristics was in the middle(18%).The relative difference of texture features was the largest(28%),which was most suitable to distinguish normal and lodging rice.The results of the extracted rice area show that the classification method based on mean texture has the highest accuracy(94%).The accuracy of the classification method based on spectral features is in the middle(72%),which is between the two(vegetation index and mean texture features).The accuracy of the optimal threshold classification method based on vegetation index is the lowest(66%).The results can provide an important reference for formulating post disaster production management,prevention and control measures and evaluating output loss.It has important reference significance for dynamic monitoring of rice lodging area.
作者
宁静
周芳琪
周杰
NING Jing;ZHOU Fang-qi;ZHOU Jie(College of Public Management and Law,Northeast Agricultural University,Harbin 150030,China)
出处
《科学技术与工程》
北大核心
2022年第31期13723-13729,共7页
Science Technology and Engineering
基金
国家自然科学基金(41971217)。
关键词
无人机
水稻
倒伏
多光谱
面积估算
unmanned aerial vehicle(UAV)
rice
lodging
multispectral
area estimation
