|
|
Italian researchers have developed a method to predict the yield of vineyards based on aerial images. They demonstrated that it is possible to quickly and reliably determine the harvest quantity before picking using drones. This study published on the research platform Oeno-one is the first on the use of drones for yield estimation in viticulture. The team consisted of scientists from the National Research Council of Italy and the University of Modena and Reggio Emilia. They utilized the fact that the external characteristics of agricultural products can be well captured with digital image processing. Crucial for the evaluation of the digital aerial images are the image pixels that come from grapes.
The investigations took place in 2021 and 2022 in several Sangiovese vineyards near Siena. There, the researchers placed colored reference markers to identify the grapes in the aerial images and automate the image processing. Drones equipped with cameras then flew over the vineyards and photographed the vines. The geometric distortions that aerial images bring were corrected by the scientists using a control point tool. For grape detection, they used color thresholds and image filters. From the number of grape pixels determined in this way, the researchers then calculated the grape yield per vine using a linear regression model. After the harvest in both years, the actual yield was established. The result of the automatic prediction proved to be satisfactory: the relevant error metric was within the tolerance range.So far, winemakers often estimate vineyard yield by randomly cutting and weighing grapes. The projected values derived from this then yield the expected harvest quantity. This procedure takes time and is particularly cumbersome in steep locations. The new method is significantly faster, independent of topography and weather, and avoids interventions in the vineyard.The research team now aims to further improve image processing and utilize deep learning tools to recognize and segment grape pixels. Other goals include extending the approach to other agricultural areas and developing user-friendly tools for application, for example, with smartphones.(cs / oeno-one.eu)More on the topic: