A new app for smartphones to count the number of flowers per inflorescence under field conditions

Abstract

The number of flowers per inflorescence is an interesting porameter forplan! phenotyping and early yield forecast. Unfortunately, o manual count of flowersis unviable as il is time and labor demanding. This work presents o new, easy-to-use,smartphone application for the automated actual flower estimation in grapevineinflorescences in the field, bosed on artificial vision techniques ond mothemoticolmodelling. The applicotion, called VitisFlower®, offering a friendly interface, firstlyguides the user to appropriately take an inflorescence photo (when inflorescencesare at BBCH 55) using the smartphone's camera. Then, by means of imoge analysis,the flowers in the image are detected ond counted. Finally, the actual number offlowers in the inflorescence is inferred from the detected flower number using amathemalical model, and both, detected and actual values, are provided to the user.The process con be performed as mony tímes as needed. VitisFlower®, al first, isoffered, free of charge, for devices powered by Android®, being the 10S® versionunder curren! developmenl. VitisFlower® is an easy to use, relioble, non-invasivetool to automate flower counting of grapevine inflorescences in the vineyord, whichcan be a valuable information for yield forecasl and precision viticulture strategiesi m plementation.