AGRiculture with Intelligent POLarimetric and HYperspectral Sensing (AGRIPOLHYS) – AGRIPOLHYS

Hyperspectral Imaging (HSI), which allows contact-less, non-destructive and fast optical sensing, is a good candidate for massive plant health monitoring in smart agriculture. To allow a dramatic reduction of both plant losses and chemical treatments, it is essential to provide the farmers with tools capable of detecting and identifying plant diseases, at very early stages, when the symptoms are not detectable by the human eye. Constructing such tools constitutes, today, an unaddressed challenge. In AGRIPOLHYS project we propose to perform early detection and identification of four tomato plant diseases (powdery mildew, late blight, gray mold and leaf mold), firstly, considering each disease separately, secondly, in a consortium. We propose to follow and combine three distinct paths: to use HSI system with an extended spectral range covering both VNIR (Visible Near Infrared) and SWIR (Short Wave Infrared); to add polarization sensing to the VNIR and SWIR HSI camera; to use innovative artificial intelligence algorithms.

Firstly, AGRIPOLHYS project includes research efforts in new hardware HSI camera conception, for constructing a high-speed system capable of simultaneous polarization sensing and hyperspectral sensing. Secondly, AGRIPOLHYS project targets the improvement of the performances of the Artificial Intelligence algorithms realizing plant disease detection and identification. On one hand, we research new routes in Artificial Intelligence for the analysis of the colossal amount of data produced by the new high-speed spectro-polarimetric camera. On the other hand, we create a high-quality database containing ground truth by both imaging plant leaves and detecting plant diseases at early stages with new molecular and biochemical tools (as for example Volatile Organic Compounds). This database serves for algorithms calibration and testing.