Abstract

The increase of technological resources and the automation of agricultural processes have contributed to the development of new technologies that allow for characterizing and evaluating the phenotype of plants. Specifically, in a climate change scenario, plant phenotyping technologies are fundamental to accelerating breeding programs of essential crops and contribute to the selection process for the development and subsequent development of new varieties and cultivars. In this sense, reconstructing using 3D images of plants and acquiring their spatial information is an effective alternative to solve these problems. In this study, using aerial RGB imagery, two methods (k-means and Normalized Red-Green Difference Index) were implemented for estimating the covered area by watermelon plants. The methods were contrasted using the linear regression slope (roe) and coefficient of determination (R2), as well as the index of agreement (d). The results indicated a medium agreement between the methods when the plants were small (roe =0.68; R2=0.75; d=0.55). Whereas when plants are larger there was a better agreement between models (roe =0.89; R2=0.91; d=0.89).