Estimación de la carga frutal en manzano cv. " Fuji" a través del análisis de imágenes digitales
Loading...
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de Concepción
Abstract
La determinación de carga frutal es clave para la gestión de prácticas de poda y raleo en huertos de manzano. El objetivo de este trabajo fue evaluar el potencial uso del análisis de imágenes digitales utilizando métodos en el espacio de color RGB para estimar carga frutal en cultivos de manzanos cv. ‘Fuji’. Para ello se establecieron tres niveles de carga frutal (alta, media y baja) a través de tres intensidades de raleo manual: 0 %, 50 % y 75 %. Se contabilizó manualmente la carga frutal y se tomaron capturas fotográficas a cada árbol, para posteriormente identificar los frutos utilizando diferencias de canales y umbrales en el espacio de color RGB. Las imágenes fueron sometidas a cuatro métodos: i) conteo digital de los frutos en la imagen a través del software ImageJ; ii) Obtención del área de la máscara binaria; iii) Conteo digital del número de frutos por cuadrante; y iv) cálculo del índice área de fruta por área de cuadrante (AF/AC). El método RGB fue efectivo para discriminar la cantidad de frutos rojos y verdes en la imagen. El método de conteo digital de número de frutos por cuadrantes y del índice AF/AC presentaron la mayor sensibilidad estadística (valor p=0,0004 y 0,0005, respectivamente) para discriminar diferencias de carga frutal en manzanos, mientras que el método de número de frutos por cuadrante presentó la mayor capacidad predictiva (R2= 0,91: p=0,0001).
Determining crop load is crucial for managing of pruning and thinning practices in apple orchards. The study aimed to assess the effectiveness of digital image analysis using methods in red, green, and blue (RGB) color space methods in estimating crop load in apple cv. ‘Fuji’. Three crop load levels (high, medium, and low) were established based on manual thinning intensities of 0%, 50% and 75%, respectively. The crop load was manually-estimated and photographic captures were taken for each tree. Fruits were identified using channel differences and thresholds in the RGB color space. Four methods were applied to the images: i) digital counting of fruits using ImageJ software; ii) calculation of the area of binary mask; iii) digital counting of fruits per quadrant; and iv) calculation of the fruit area index per quadrant area (AF/AC). The RGB method effectively discriminated red and green fruits in the images. Digital counting of fruits per quadrant and the AF/AC index presented the highest statistical sensitivity (p value = 0.0004 and 0.0005, respectively) in discriminating crop load differences in apple trees. Moreover, the number of fruits per quadrant presented the strongest predictive capacity (R2=0.91; p=0.0001). These results suggest that quadrant-based RGB image analysis is a promising method for estimating crop load differences in apple trees. However, further refinement is needed in future research work.
Determining crop load is crucial for managing of pruning and thinning practices in apple orchards. The study aimed to assess the effectiveness of digital image analysis using methods in red, green, and blue (RGB) color space methods in estimating crop load in apple cv. ‘Fuji’. Three crop load levels (high, medium, and low) were established based on manual thinning intensities of 0%, 50% and 75%, respectively. The crop load was manually-estimated and photographic captures were taken for each tree. Fruits were identified using channel differences and thresholds in the RGB color space. Four methods were applied to the images: i) digital counting of fruits using ImageJ software; ii) calculation of the area of binary mask; iii) digital counting of fruits per quadrant; and iv) calculation of the fruit area index per quadrant area (AF/AC). The RGB method effectively discriminated red and green fruits in the images. Digital counting of fruits per quadrant and the AF/AC index presented the highest statistical sensitivity (p value = 0.0004 and 0.0005, respectively) in discriminating crop load differences in apple trees. Moreover, the number of fruits per quadrant presented the strongest predictive capacity (R2=0.91; p=0.0001). These results suggest that quadrant-based RGB image analysis is a promising method for estimating crop load differences in apple trees. However, further refinement is needed in future research work.
Description
Tesis presentada para optar al título de Ingeniera Agrónoma
Keywords
Manzano - Rendimiento, Manzano - Control biológico, Manzanas - Calidad