论文部分内容阅读
Hyperspectral imaging system is increasingly being used for plant phenotyping, which provides comprehensive assessment of plant traits such as growth, tolerance, resistance, physiology,and ecology.To improve the performance of the stationary hyperspectral imaging system, it is imperative that the system is spatially and spectrally calibrated for the image acquisition.First, the smile and keystone effect from the system was removed by using a modified geometric control point correction.Then, the geometric calibration was performed with a white paper printed with thin parallel lines 2 mm apart, and the spectral calibration was implemented using spectral calibration lamps (xenon lamp and mercury-argon lamp).The pixel-by-pixel flat-field correction was performed at all the wavelengths by subtraction of the dark current and then division of the difference between the spectralon reference and the dark current.The light source and source-detector distance for the stationary image acquisition of plant materials were also optimized.Both tungsten halogen lights and LED lights including two cool white LED lights with the spectral region of 400-700 nm and four 830-nm infrared LEDs were investigated related to the incident angle, intensity, and exposure time during the image acquisition.System response, repeatability,and stability were evaluated by the standard Teflon samples as well as leaves of five different plants.This research provided a systematic guide for calibrating and evaluating hyperspectral imaging systems, and offered solutions to improve accuracy in measuring plant phenotypes.In addition, the proposed method could be easily adapted for developing and designing a general-used hyperspectral imaging platform applied for different plant materials.