Citrus ''Unshiu'' is the most popular citrus variety of Jeju Island in Korea. To meet the increased demand of citrus consumption, the citrus yield should be managed properly and systematically. However, current cultivation technique used in citrus farms is still dependent on manual works, which cannot provide on-time prescription to citrus trees exposed to abiotic stresses. Remote sensing techniques using drones can be an effective alternative for this problem.
The ''Drone'' is an unmanned aerial vehicle(UAV) which can be controlled by electromagnetic signal, and the fields of using UAV-platformed remote sensing technology is being expanded rapidly. UAV remote sensing at low altitude is capable of detailed observations with high resolution, and it can provide valuable information to field of precision agriculture. Hyperspectral/Multispectral imaging techniques have shown reliable performance to measure physico-chemical characteristics of biological materials. In particular, multispectral imaging using only a small number of specified wavelengths has the advantages of much lower cost and faster measurement speed compared to hyperspectral imaging.
In this study, a remote sensing system using an UAV mounted with multispectral imager was constructed to monitor the abiotic stress of citrus Unshiu. After optimizing the UAV-multispectral remote sensing system through several tuning operation, the optimal auto-flight path considering the sensor property was set. Using automatic flight function, multispectral Vis/NIR images were acquired for citrus trees exposed to water and nitrogen stress. To minimize errors caused by changing light conditions and atmospheric effect, radiometric calibration and image preprocessing techniques were applied. And a 3D multispectral image cube of citrus farm was created through the image stitching process using SIFT algorithm. Several vegetation index analysis, principal component analysis (PCA), and partial least squares discriminant analysis (PLS-DA) were used to discriminate the citrus trees exposed to abiotic stresses.
It was possible to discriminate the water stress by using the wavelengths of red and red-edge range such as 670 nm, 717~731 nm and 770~780 nm. In case of nitrogen stress, it was possible to discriminate by using the wavelengths associated with red, red-edge and near-infrared range such as 679~693 nm, 735~770 nm, and 850 nm. In common, these wavelengths are related to chlorophyll changes in leaves of vegetation, and it is considered that the chemical changes related to chlorophyll appeared due to the water and nitrogen stresses. Also, these results show consistent tendency with other studies in related fields.
Therefore, it is confirmed that the water and nitrogen stresses of citrus Unshiu can be discriminated by analyzing the spectral data of corresponding wavelength gathered by the multispectral remote sensing system. The results demonstrated that the UAV system combined with the multispectral imager has good potential for stress monitoring of citrus Unshiu in a rapid and nondestructive manner.