Abstract: Mastering the spectral differences and changes in the identification effect of poisonous weeds after chemical control is the basis for realizing the control effect of large-scale remote sensing monitoring. This study took Aconitum leucostomum as the identification object,and the experimental was performed at late May (before chemical control),late June (15 days after chemical control),and mid-July (30 days after chemical control). We analyzed spectral changes,filtered characteristic wavelengths,and established the Fisher discriminant model to identify it. The results showed that:after chemical containment,the reflectivity of A.leucostomum increased in the 600~710 nm band,and decreased in the 770~1 075 nm band;the reflectivity of other forages in the 720~1 075 nm band increased in June,and there was no significant difference in July;The sensitive identification bands varied from month to month,mainly in the 399~404 nm band,512~578 nm band,684~718 nm band,775~778 nm band and 979~1 075 nm band;The recognition accuracy of A.leucostomum was the best in May without prevention and control,and the recognition accuracy was 96.00%;After prevention and control treatment,the identification accuracy of A.leucostomum affected by the herbicide in June and July was 73.33% and 78.00%,respectively. In general,chemical prevention and control reduced the recognition accuracy of A.leucostomum,but after 30 days of prevention and control,the spectral characteristics of A.leucostomum affected by herbicide would change significantly,and the recognition accuracy would increase.

Key words: Aconitum leucostomum, Chemical prevention and control, Spectral analysis, Stepwise discrimination, Fisher discriminant

摘要： 掌握毒害草在化学防控后的光谱差异及识别效果变化是实现大面积遥感监测防控效果的基础。本研究以白喉乌头(Aconitum leucostomum)为识别对象，试验在5月下旬(化学防控前)、6月下旬(防控后15 d)以及7月中旬(防控后30 d)进行，分析其光谱变化，筛选特征波长并建立Fisher判别模型进行识别。结果表明：防控处理后，白喉乌头在600~710 nm波段内反射率升高，在770~1 075 nm波段内反射率降低；其他牧草在720~1 075 nm波段内的反射率表现为6月升高，7月无明显差异；各月份的敏感识别波段不相同，主要分布在399~404 nm波段、512~578 nm波段、684~718 nm波段、775~778 nm波段以及979~1 075 nm波段内；5月未防控乌头识别效果最好，识别精度为96.00%，防控处理后6月、7月的药害乌头识别精度分别为73.33%，78.00%。总体而言，化学防控降低了白喉乌头的识别精度，但防控30 d后药害乌头的光谱会产生明显变化，增加其识别精度。

关键词: 白喉乌头, 化学防控, 光谱分析, 逐步判别, Fisher判别