Conventional RGB cameras capture images that are similar to what a human could see. Therefore, they are very usable, because as humans we are easily able to interpret the images and come up with pipelines for analysis. Nevertheless, plants emit and absorb more wavelengths than the ones which are collected in RGB; plants absorb solar radiation to execute their photosynthesis, which is subsequently partly re-emitted in the near-infrared (NIR) spectral region to prevent overheating. The NIR spectral region is not visible by humans, but can be captured by spectral cameras and subsequently used to estimate plant stress (caused by for example water shortage, pests, diseases, nutrient shortage, or heat stress). An interesting current trend is to investigate the potential of the NIR spectral region for automatically identifying plant stress and health.
The task for this project is to investigate which potential NIR spectral regions may offer, and what the limitations and advantages of these offers are. Subsequently, data analysis can be executed on data that is already collected in a greenhouse to investigate the outcomes of different analyses using the NIR spectral region. Also, an AI algorithm could be trained to automatically detect plant stress in an early stage. Subsequently, this AI algorithm can be used to inform the farmer about possible shortcomings.
20% Theory, 20% research potential NIR based on already collected data, 40% implementation, 20% writing report
Le Viet Duc, firstname.lastname@example.org