Within-row spacing sensing of maize plants using 3D computer vision

Within-row plant spacing plays an important role in uniform distribution of water and nutrients among plants which affects the final crop yield. While manual in-field measurements of within-row plant spacing is time and labour intensive, little work has been done on an alternative automated process. We have attempted to develop an automatic system making use of a state-of-the-art 3D vision sensor that accurately measures within-row maize plant spacing. Misidentification of plants caused by low hanging canopies and doubles were reduced by processing multiple consecutive images at a time and selecting the best inter-plant distance calculated. Based on several small scale experiments in real fields, our system has been proven to measure the within-row maize plant spacing with a mean and standard deviation error of 1.60 cm and 2.19 cm, and a root mean squared error of 2.54 cm, respectively.