A Real-time Approach System for Vineyards Intra-row Weed Detection

Abstract

With the incorporation of autonomous robotic platforms in various areas (industry, agriculture, etc.), numerous mundane operations have become fully automated. The highly demanding working environment of Agriculture let the development of techniques and machineries that could cope with each case. New technologies (from high performance motors to optimization algorithms) have been implemented and tested in this field. Every cultivation season, there are several operations that contribute to crop development and have to take place at least once. One of these operations is the weeding. In every crop, there are plants that are not part of it. These plants, in most cases have a negative impact on the crop and had to be removed. In the past the weeding was taken place either by hand (smaller fields) or by the use of herbicides (larger fields). In the second case, the dosage and the time are pre-defined, and they are not taking into consideration the growth percentage and the weed allocation within the field. In this work, a novel approach for intra-row weed detection in vineyards is developed and presented. All the experiments both for data collection and algorithm testing took place in a high value vineyard which produce numerous wine varieties. The aim of this work is to implement an accurate real-time robotic system for weed detection and segmentation using a deep learning algorithm in order to optimize the weeding procedure. This approach consists of two essential sub-systems. The first one is the robotic platform that embeds all the necessary sensors and the required computational power for the detection algorithm. The second one is the developed algorithm. From all the developed models, the selected one performed accurately in the training procedure and in the unknown datasets. In order to properly validate the algorithm, the unknown datasets were acquired in different time periods with variations in both camera angle and wine varieties. © 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0)