Optimum spectral and geometric parameters for early detection of laurel wilt disease in avocado
Optimum spectral and geometric parameters for early detection of laurel wilt disease in avocado
复制标题
DOI：
10.1016/j.rse.2015.09.011
复制
发表时间：
2015-12-15
影响因子：
13.5
通讯作者：
Abdulridha, J.
中科院分区：
文献类型：
--
作者：
De Castro, A. I.;Ehsani, R.;Abdulridha, J.
鳄梨（Persea americana）是佛罗里达州仅次于柑橘的第二大重要作物，批发价值3500万美元，约占热带水果作物面积的60%。月桂枯萎病（LW）是一种致命疾病，沿美国东南沿海迅速蔓延，影响商业鳄梨生产。本文评估了快速准确地检测LW的空间和光谱要求。采用方差分析和多层感知器（MLP）和径向基函数（RBF）两种神经网络对健康(H)、疫霉根腐病（PPR）和LW叶片的光谱数据进行分析。确定了最有效的波长，并将滤光片更新为MCA-6 Tetracam相机（580-10 nm, 650-10 nm, 740-10 nm, 750-10 nm， 760-10 nm和850-40 nm）。然后，使用MCA相机从直升机上拍摄多光谱航空图像，在三个高度（180,250和300 m）的牛油果田里，树木处于LW发育的不同阶段，早期，中期和晚期。分析分别基于2级和4级系统进行。2级系统用于区分H树和LW树，足以识别需要移除的树木；4级系统用于区分H树和LW发育的三个阶段。航空图像分析证明了所选过滤器在成功识别LW方面的实用性，即使对于处于疾病发展早期且症状最小的树木也是如此。根据m值和林冠大小、果园大小等生物学参数，选择理想飞行高度为250 m （153 cm像素大小）。较高m值确定的最佳VIs为TCARI（760 ~ 650）、GNDVI、NIR/G、Redge/G和viggreen，使用与Redge相关的任何波段（740和750 nm）或NIR区域（760和850 nm）。2级和4级系统使用上述Vls区分LW的效用的结果报告；而基于H、早、中、晚4类系统的算法开发更为方便。通过本研究提出的方法早期发现LW可以使农民通过适当的管理策略控制这种疾病的传播。(C) 2015爱思唯尔公司版权所有。
Avocado (Persea americana) is a crop that is second in importance in Florida behind citrus with a wholesale value of $35 million and represents approximately 60% of the tropical fruit crop acreage. Laurel wilt (LW) is a lethal disease that has spread rapidly along the southeastern seaboard of the United States affecting commercial avocado production. This article evaluates the spatial and spectral requirements for quick and accurate detection of LW. Spectral data from healthy (H), Phytophthora root rot (PPR) and LW leaves were analyzed using ANOVA and two neural networks, multilayer perceptron (MLP) and radial basis function (RBF). The most effective wavelengths were identified and the filters were updated to a MCA-6 Tetracam camera (580-10 nm, 650-10 nm, 740-10 nm, 750-10 nm, 760-10 nm and 850-40 nm). Then, the MCA camera was used to take multispectral aerial images from a helicopter at three altitudes (180,250 and 300 m) in an avocado field with trees at different stages of LW development, early, intermediate and late. The analyses were conducted based upon 2-class and 4-class systems. The 2-class system was designed to differentiate H and LW trees sufficient to identify trees for removal and the 4-class system was used to differentiate H plants and the three stages of LW development. Aerial image analysis proved the utility of the selected filters for successful identification of LW, even for trees in early stage of disease development with minimal symptoms. The ideal flight altitude of 250 m (153 cm pixel size) was selected according to the M-values and biological parameters such as canopy size and orchard size. The optimum VIs determined by higher M-values were TCARI(760-650) as well as GNDVI, NIR/G, Redge/G and VIGreen using any of the bands related to Redge (740 and 750 nm) or NIR regions (760 and 850 nm). Results reported on the utility of the 2-class and 4-class systems using the above Vls to discriminate LW; however it would be more convenient to develop the algorithm based on the 4-class system (H, early, intermediate and late). The early detection of LW through the methodology proposed in this research could allow farmers to control the movement of this disease through proper management strategies. (C) 2015 Elsevier Inc. All rights reserved.