Image-Based Measurement of Grape Inflorescence Length for Automatic Inflorescence Trimming

In the production of table grapes, inflorescence trimming is essential for producing high-quality grapes by resolving nutrient competition and suppressing the occurrence of diseases. However, the appropriate timing for inflorescence trimming is short for farmers, causing a significant burden. Therefore, the automation of the trimming task is highly expected. In previous research, a technology was proposed to estimate inflorescence length using computer vision and assist in inflorescence trimming with augmented reality (AR). However, there was a challenge in the estimation accuracy being influenced by the shape of the inflorescence. In this study, we propose a new method for accurately detecting the inflorescence axis and estimating the inflorescence length regardless of its shape, aiming to automate inflorescence trimming in grape cultivation using a robot. The experimental results demonstrate that the proposed method outperforms the previous method in all evaluation metrics, proving its capability to accurately detect the inflorescence axis and estimate the inflorescence length, irrespective of the shape.