云南切梢小蠹危害云南松监测模型与判定规则

Research of Damage Monitoring Models and Judgment Rules of Pinus yunnanensis with Tomicus yunnanensis

目的]通过分析云南切梢小蠹危害下的云南松枝梢高光谱特征,建立其危害程度监测模型和判定规则。[方法]使用便携式地面成像光谱仪采集云南切梢小蠹蛀梢期的云南松中、幼龄林枝梢光谱反射率数据,分析光谱特征并提取特征参数,以此构建云南松受云南切梢小蠹危害程度的监测模型和判定规则。[结果]随危害程度加重,在绿波段(510560 nm)和近红外波段(7201 036 nm),光谱反射率逐渐降低;一阶微分曲线在红边(680760nm)的峰值向短波方向移动;云南切梢小蠹危害程度与光谱反射率及其一阶微分在509539、549564、595677、687692、702807、838875、8911 031 nm显著相关;以高光谱特征参数S_(Dr)、D_y、(D-H)/(D+H)、S_(Dnir)、(S_(Dnir)-S_(Dr))/(S_(Dnir)+S_(Dr))构建的4个监测模型的实测值与预测值的线性拟合关系较好(R^2>0.9),可准确估算云南切梢小蠹危害程度;根据4个监测模型建立的判定规则准确率高(≥80%),其中,多元线性回归模型y=-7.720x_1+1.275x_2+1.251x_3-4.835x_4+1.135x_5+6.632的判定规则准确率最高(健康(<1.589)、轻度受害[1.589,2.465)、中度受害[2.465,3.381)、重度受害(≥3.381)),达93.333%。[结论]根据云南松高光谱特征参数,建立的监测模型和判定规则可有效监测云南切梢小蠹危害程度,研究结果可用于云南切梢小蠹危害发生发展的监测。

[ Objective ] By analyzing the hyperspectral features of Piuus yunnanensis in different damage levels and building monitoring models, to establish the damage monitoring model and judgment rules for integrated control of Tomicus yuuuanensis. [ Method ] In shoot damage period of T. yunnauensis, the imaging hyperspectral data of young and middle-aged P. yuuuaueusis in study area were obtained by SOC710VP, and hyperspectral features were analyzed to extract hyperspectral features parameters to build damage levels detection models and judgment rules. [ Result ] With the aggravation of damage level of T. yunnanensis, the reflectance spectral curves of P. yunnauensis gradually declined at green bands （510 -560 nm） and near-infraled bands （720 -1 036 nm） , and the peak values of spectral first derivative curves gradually decreased at red edge （680 -760 nm）. In 509 -539,549 -564,595 - 677, 687 -692, 702 -807, 838 -875 nm, and 891 -1 031 nm, the damage levels and leflectance and first deriv- ative of P. yuuuaueusis needle were significantly colxelated. Hyperspectral parameters Svr, Dy, （ D -H）/（D ＋ H）, SDnir, and （SDnir- SDr）/（SDnir ＋ SDr） were used to establish the monitoring models, the R2 of measured value and pereicted value all reached 0.9. The accuracy of quantitative judgment rules based on 4 monitoring models were higher than 80%, the rules （Healthy （ 〈 1. 589）, Slight damage [ 1. 589, 2. 465）, Moderate damage [2. 465, 3. 381 ）, Severe damage （ i〉3. 381 ） ） based on nmhivariable linear regression model y = -7. 720x1 ＋ 1. 275x2 ＋ 1. 251x3 -4. 835x4 ＋ 1. 135x5 ＋6. 632, reached the highest precision （93.33%）. [ Conclusion] The monitoring models and judgment rules based on hyperspectral feature parameters can monitor the damage level of T. yunnanen- sis effectively.