Volumentric model evaluation using digital image processing of unbroken corn kernels
In this article, three models are compared to estimate the volume of corn (Zea mays) grains from values obtained using digital image processing techniques on images of the sample. In the first model, the volume of each grain is assumed to be proportional to its cubed length. In the second model, the...
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Artículo revista |
| Lenguaje: | Español |
| Publicado: |
Instituto Agrotécnico "Pedro M. Fuentes Godo" - Facultad de Ciencias Agrarias (UNNE)
2025
|
| Materias: | |
| Acceso en línea: | https://revistas.unne.edu.ar/index.php/agr/article/view/8548 |
| Aporte de: |
| Sumario: | In this article, three models are compared to estimate the volume of corn (Zea mays) grains from values obtained using digital image processing techniques on images of the sample. In the first model, the volume of each grain is assumed to be proportional to its cubed length. In the second model, the volume is proportional to the square of the width, multiplied by its length. In the third model, the volume is proportional to the projected area multiplied by the length. The proportionality constants are determined experimentally using the toluene displacement method. Ten samples, each containing one hundred grains were prepared. Five samples were used to determine the proportionality constants needed in the models, while the remaining five were used to compare the volumes obtained by the three models. Images of the samples were obtained using a desktop scanner at a resolution of 300 dpi. The operations on the images were carried out using the ImageJ® software. The length, width, and area of each grain were determined by finding the ellipse that best fits the projected area of each grain in the red channel of each binarized image. The volumes of the control samples were determined using the values of the proportionality constants, and then compared with the experimental ones. The average percentage relative deviation (RPD) was calculated for each model using these five samples. The first model had a DRP of 3.9 % while the second and third models had DRP of 4.9 % and 2.5 %, respectively. The low margin of error and simple requirements for application make this methodology easily adaptable for determining volume using non-destructive methods. |
|---|