Abstract: Taking the citrus （Citrus reticulata Blanco） in Jiujian Village, Pingshan Town, Luzhai County, Liuzhou, Guangxi as the research object, the citrus huanglongbing （HLB） plants were identified through the ground manual measurement, and the hyperspectral imaging of the calibrated citrus planting plot was obtained by cooperating with UAV low altitude remote sensing; the average spectra of the regions of interest （ROI） in the canopy of healthy citrus plants and HLB plants were calculated, and outlier removal, smooth denoising, and spectral transformation on the initial spectra were performed to obtain the original spectra, first-order differential spectra （FDR）, and second-order differential spectra （SDR）;after dimensionality reduction using principal component analysis, a support vector machine （SVM） classification model was constructed. The results showed that by selecting feature bands ranging from 400 to 1 000 nm and using ArcGIS software to extract sample average spectra, the classification accuracy of the training and testing sets of the full band first order differential spectra reached 87.41% and 84.67%, respectively. The SVM classification model parameters were C=35.39 γ= 0.01;using ENVI software to extract the average spectrum of samples, the classification accuracy of the training and testing sets for the full band first-order differential spectrum reached 92.39% and 96.43%, respectively. The SVM classification model parameters were C=5.06 γ=1.02. UAV low altitude remote sensing and hyperspectral monitoring of citrus HLB was feasible, which could quickly identify HLB plants in citrus plantations.

Key words: huanglongbing （HLB）, UAV, hyperspectral, support vector machine, low altitude remote sensing, citrus（Citrus reticulata Blanco）