Rapid detection of chlorophyll content and distribution in citrus orchards based on low-altitude remote sensing and bio-sensors

• Published in: International journal of agricultural and biological engineering 2018-03, Vol.11 (2), p.164-169
• Main Authors: Wang, Kejian, Li, Wentao, Deng, Lie, Lyu, Qiang, Zheng, Yongqiang, Yi, Shilai, Xie, Rangjin, Ma, Yanyan, He, Shaolan
• Format: Article
• Language: English
• Subjects: Altitude; Analytical chemistry; Biosensors; Canopies; Chlorophyll; Detection; Fertilization; Fluorescence; Fruits; Goodness of fit; Least squares method; Leaves; Low altitude; Mathematical models; Multiplexing; Nitrogen; Nutrition; Orchards; Prediction models; Production capacity; Reflectance; Regression analysis; Remote sensing; Remote sensors; Sensors; Soil analysis; Spatial distribution; Spectral reflectance; Vegetation
• ISSN: 1934-6344; 1934-6352
• DOI: 10.25165/j.ijabe.20181102.3189

The accuracy of detecting the chlorophyll content in the canopy and leaves of citrus plants based on sensors with different scales and prediction models was investigated for the establishment of an easy and highly-efficient real-time nutrition diagnosis technology in citrus orchards. The fluorescent values of leaves and canopy based on the Multiplex 3.6 sensor, canopy hyperspectral reflectance data based on the FieldSpec4 radiometer and spectral reflectance based on low-altitude multispectral remote sensing were collected from leaves of Shatang mandarin and then analyzed. Additionally, the associations of the leaf SPAD (soil and plant analyzer development) value with the ratio vegetation index (RVI) and normalized differential vegetation index (NDVI) were analyzed. The leaf SPAD value predictive model was established by means of univariate and multiple linear regressions and the partial least squares method. Variable distribution maps of the relative canopy chlorophyll content based on spectral reflectance in the orchard were automatically created. The results showed that the correlations of the SPAD values obtained from the Multiplex 3.6 sensor, FieldSpec4 radiometer and low-altitude multispectral remote sensing were highly significant. The measures of goodness of fit of the predictive models were R2=0.7063, RMSECV=3.7892, RE=5.96%, and RMSEP=3.7760 based on RVI(570/800) and R2=0.7343, RMSECV=3.6535, RE=5.49%, and RMSEP=3.3578 based on NDVI[(570,800)(570,950)(700,840)]. The technique to create spatial distribution maps of the relative canopy chlorophyll content in the orchard was established based on sensor information that directly reflected the chlorophyll content of the plants in different parts of the orchard, which in turn provides evidence for implementation of orchard productivity evaluation and precision in fertilization management.