Citrus orchards under formation evaluated by UAV-Based RGB Imagery

• Published in: Scientia agricola Vol. 79; no. 5
• Main Authors: Oliveira, Willer Fagundes de, Santos, Silvânio Rodrigues dos, Struiving, Tiago Barbosa, Silva, Lucas Alves da
• Format: Journal Article
• Language: English
• Published: Escola Superior de Agricultura "Luiz de Queiroz" 01-01-2022; Universidade de São Paulo
• Subjects: AGRICULTURE, MULTIDISCIPLINARY; citrus trees; image segmentation; precision agriculture; Python; precision agriculture; image segmentation; citrus trees; Python
• ISSN: 1678-992X; 1678-992X
• DOI: 10.1590/1678-992x-2021-0052

Few studies have investigated the biometric attributes of citrus orchards under formation that use RGB sensors on board unmanned aerial vehicles (UAV) and the challenges are great. This study aimed to develop and validate a method of using aerial UAV images by automated routines to evaluate the biometric attributes of a crop of ‘Tahiti’ acid lime under formation. We used a multirotor UAV, programmed to capture images at three different map scales, with a frontal and side overlap of 80 %. Geoprocessing was carried out both with and without ground control points on each scale. An automated routine was developed in an open-source environment, consisting of three processing phases: i) Estimation of the plant biometric attributes, ii) Statistical analysis, and iii) Statistical Report Map (SRM). The use of the developed routine allowed to delimit and estimate the crown projection area with an accuracy of more than 95 % as well as identify and quantify the plants with an accuracy of over 97 %. The use of ground control points during the processing stage does not increase accuracy in estimating the biometric attributes under evaluation. On the other hand, map scale is strongly correlated with the quality of the estimates, especially plant height. The results allowed to define a method for the acquisition and analysis of aerophotogrammetric data using a UAV, which can be used to measure the plant biometric attributes under analysis and the method can be easily adapted to perennial crops.