S18 - Session P1 - Development of a vision and image analysis system to evaluate the natural regulation of crop pests

Authors: Jean-Michel Ricard *, Gilles Sentenac, Anne Guerin, Paul Masquin, Alain Ferre, Alicia Fougère, Véronique Tosser, Jonathan Marks-Perreau , Thierry Corbière, Gwenael Duclos, Antoine Gardarin, Laeticia Girerd

Natural pest regulation is still difficult to evaluate because of methodological difficulties, especially in quantifying predation. The lack of an indicator for this ecosystem service leads to its underutilization for reducing plant protection products in all types of crops. The Mirage project (2019-2023) aims to remove this methodological barrier by combining a macro camera and image analysis. A prototype camera, called Beecam®, has been designed to produce high-definition videos and photos (8MPix), while the overall footage is dramatically reduced by using real-time "in-image motion detection". Two synchronized sensors allow close observation of up to a few cm2. The sensitivity is very high so that sub-millimeter insects are detected. To process all this data, a neural network was developed for automatic classification, initially of 8 taxa of natural enemies from a database of 8000 photos, some of which were obtained with the camera. Improving the performance of the software, called Harmony, is ongoing as well as adding to the number of identifiable taxa. The performance of this tool is currently being tested on different stages of the main pests of annual and perennial crops, which are filmed in the field during the immobile phase. The first results show that it offers interesting perspectives to identify new trophic relationships and to better quantify them. The transfer potential of this tool will be evaluated in teaching, in experimentation (effect of agroecological practices and developments on natural regulation) and in crop management.