Efficiency analysis of the use of model matching algorithm for plant counting

• Autor(a) principal: Arantes, Bruno Henrique Tondato
• Data de Publicação: 2020
• Outros Autores: Arantes, Leticia Tondato, Santos, Janyne Moura dos, Ventura, Matheus Vinicius Abadia, Gomes, Luiz Fernando
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/4576
• topic: Arboles frutales; Detección remota; Drone; Agricultura de precisión; Árvores frutíferas; Sensoriamento remoto; Drone; Agricultura de precisão.; Fruit trees; Remote sensing; Drone; Precision agriculture.

Often the producer does not know the exact number of fruit trees on his property or is unaware over the years due to the death of many plants. As a result, in order to avoid the need for a field trip for manual counting, this research aimed to use a model matching algorithm in parallel with the use of a low-cost drone to assess its efficiency in automatic counting of spaced canopy plants and joints. The red, green and blue bands captured by the Phantom 4 Advanced were used, and the red band with linear enhancement for the cut option, to facilitate the distinction of the orchard and the rest of the targets in the image and to obtain a better result in the detection of fruit trees. The flight was performed at a height of 80 meters with an overlap between bands of 70% and in the same range of 80%. As a result, 97.98% of fruit trees were detected in plants with well-spaced crowns and 88.52% were identified in plants with crowns together. The numbers of false positives found were small for all situations tested, these false positives being weeds. It is concluded that the technique is efficient for counting plants with fair and spaced crowns, and detection can be improved when there is a good contrast between what you want to detect and the targets that are not of interest.