Vegetation detection based on spectral information and development of a low-cost vegetation sensor for selective...

Real-time site-specific selective spraying for crop protection against weeds, pests and diseases requires fast and precise detection of targets. This study investigated several aspects of vegetation target detection based on spectral information and developed a low-cost vegetation sensor. Specific objectives included: (1) compare vegetation-detection performance using reflected and fluorescence data; (2) evaluate the effect of light source and bandwidth, and select the optimal light source and waveband; and (3) develop a low-cost vegetation sensor and test its performance under different light conditions.

The main outcomes of this study are as follows. (1) The fluorescence excited by blue and red light-emitting diodes (LEDs) was an effective attribute with which to differentiate vegetation from non-vegetation and provided results comparative with reflected spectroscopy. (2) A blue LED could excite strong fluorescence and was recommended as the light source. (3) Under illumination by blue and red LEDs, bandwidth did not have any obvious effect on classification accuracy within the studied bandwidths from 0 nm (single wavelength) to 120 nm. (4) A low-cost vegetation sensor was developed and tested, with 100% detection accuracy in both dark and outdoor environments.

This study suggests that single-waveband fluorescence spectroscopy is an effective approach to detect vegetation targets and low-cost LEDs can be used for illumination. Light source modulation with a sinusoidal signal is an effective way to resist the influence of environmental light. Overall, the findings of this study contribute to an improved understanding of developing low-cost and effective vegetation sensors for selective spraying from theory to application. © 2022 Society of Chemical Industry.