Ngoun

Ngoun Sreymey

University of Hohenheim, Institute of Agricultural Engineering.

Department of Agricultural Engineering in the Tropics and Subtropics

 

Challenges

The basic advantage of crop imaging is the gapless and detailed detection of the desired plant characteristics even from large areas. Thus, it is fundamentally superior to point measurements, especially under heterogeneous conditions. In multi- or hyperspectral imaging, several images of the light reflexion of different wavelengths, or wavelength bands, are captured from a surface simultaneously. Based on spectral images, relevant wavelengths are identified, and indices calculated. Additional indices are used to detect plant stress. For the detection of wavelengths >1000 nm, mainly InGaAs (indium gallium arsenide) sensors are used due to the lower available light energy. In thermography, the surface temperature of leaves is usually recorded in the spectral range between 5μm and 15μm with infrared cameras and displayed, for example, via false-colour images. RGB colour images reflect visible light information between 400nm and 750nm and its color space is done with CMOS sensors. Since the light-sensitive sensors cannot distinguish between colours, a filter matrix of red, green, and blue filters is placed on the sensors. Currently, the main focus of RGB imaging in agriculture is the determination of biomass.

 

Objectives

The objective of this project is to investigate low-cost imaging methods with RGB cameras that are available worldwide at high resolutions: i) Are RGB images in principle suitable to indicate irrigation need and lack of fertilizer? ii) How is the information quality of RGB images compared to the use of more detailed and wider reflection spectra, what is the relevant image information of RGB images, and can the information content be improved by using spectral filters? iii) Do RGB images from the field show spatial patterns with respect to possible significant image information, which make automatic pattern recognition possible?

 

Aims

  •   Optimize water and nutrient management using optical information.
  •  Use multi- or hyperspectral imaging with discrete wavelengths, narrow wavebands or wider wave ranges up to the near infrared (NIR).
  •  Low-cost but high-resolution RGB imaging.

 

Methods

RGB color images reflect visible light information (400 - 750nm) and image acquisition is generally performed with CMOS sensors. During the study Images will be acquired on different wavelengths with IR-cameras, hyperspectral imaging and high resolution RGB imaging. This will be done at different levels of plant water and nutrient status under controlled conditions in order to build up a data base for remote sensing data specially emphasizing on cassava plants. The bands will be compared in order to develop an algorithm for analysing cassava plant status base on RGB colours only.

In field trials the developed method will be tested for its practicability under realistic conditions of cassava production in Cambodia. Related ground-based data, such as soil moisture, stomatal conductance of the plants and weather data will be monitored in order to assess the method. Feedback from farmers shall be used to improve the technique.

Expected results

Specifically the expected results are:

  •  A database for improving the remote assessment for plant water and nutritional status in cassava is created
  • A method was develop to enable the remote assessment of plant and nutrient status in cassava based on the UAV based acquisition of RGB color images
  • The method to assess water and nutrient status in cassava has been field tested in Kratie, Cambodia with local smallholder farmers