Institut für Rebenzüchtung
The interdisciplinary MuSik4Vine project combines innovative processes and expertise from breeding research, micro-optics and lighting technology and is developing a novel, cost-effective multi-modal sensor system for the non-destructive assessment of plant health and quality using the example of grapevines. A technological basis and strategy for the on-site assessment of quality characteristics and thus improved decision-making support in agriculture and breeding will be created. The aim of the project is to merge these approaches and AI-based data evaluation into a multi-modal sensor system for efficient phenotyping in grapevine breeding. For this purpose, VIS & SWIR wavelengths (<1600nm) are to be used for the first time to record characteristics of different dimensions with a single single-shot measurement:Grape geometry (including berry size, grape weight) as a yield parameter and indicator for Botrytis bunch rotBackscatter properties and spectral signature of the berry skin to derive physical properties, its wax layer and phenolic content. The mobile sensor system to be developed (IOF & LEJ) will enable multi-dimensional data acquisition, initially in the laboratory and later in the field. Ground truth data based on established methods from phenotyping and analytics (JKI) will serve as training and validation data for the development of AI-supported data evaluation. With such a system, for example, the effects of (a)biotic stressors can be detected at an early stage in the short term and resistant, climate-adapted varieties can be selected more effectively in the long term. In this way, the project contributes directly to increasing the sustainability and resilience of viticulture against the backdrop of current climate change.Sub-project (JKI): Reference data and evaluation The specifications of the sensor system are defined in cooperation with the partners as the basis for the development work. Generation of extensive reference data in variable breeding material using established methods (hyperspectral, 3D, impedance, etc.). Validation of the prediction accuracy in combination with meta-information (e.g. metabolome data). Supporting software development and carrying out test measurements in mapping populations and varieties to evaluate their suitability for use in breeding and viticulture.Sub-project (IOF): multispectral imaging Optical design of a miniaturized multi-aperture camera for simultaneous image acquisition in several discrete spectral bands. Optimization of the relative brightness (corresponding source selection), mastering and replication of the micro-optical components, opto-mechanical camera design based on a commercial industrial camera, integration of the sub-system and support for system integration and application.Sub-project (IOF): 3D sensor technology System design for the integration of a classic and a multispectral camera as well as lighting into an overall concept in order to generate both 3D and spectral data in sufficient resolution by data fusion of the optical channels. Algorithm development and training together with JKI. Implementation of the hardware in a robust system with a customized user interface.Sub-project LEJ: Light innovation Selection and evaluation of LED sources according to the selected spectral ranges for multispectral sensor technology and imaging. Opto-mechanical design of the light source as a subsystem, construction of the corresponding components (printed circuit boards, cooling components and illumination optics) and integration of these sources into a light source for imaging. Development of customized driver electronics for controllable illumination. Integration of the sensor system together with IOF.
Federal Ministry of Research, Technology and Space
