Development of a portable system for detection of leaf area in plants

Abstract

Recently, the determination of leaf area and growth rates in plants through portable devices has become popular due to the advantages offered by artificial vision systems to identify and classify objects using images. The purpose of the research was to develop a portable system for measuring leaf area through image recognition. The methodology includes 3D designs, acrylic device building, determining the area in pixels by the technique of radial basis neural networks (RBFN) directly in the RGB color space and pixel conversion to cm². The results obtained by analysis of variance for each species showed that the p-value was greater than 0.86 for each class of plants. Moreover, the system obtained coefficients of determination higher to 0.90 for Orange leaves, coefficients of determination higher to 0.95 for Patevaca leaves and coefficients of determination higher to 0.99 for Chirimoya leaves from a set of 30 leaves belonging to 3 species of plants with different sizes and areas compared to manual analysis. This system is a useful tool for the objective determination of leaf area in plants and becomes a nationwide alternative compared to existing expensive systems. Furthermore, the system is reprogrammable, flexible, not destructive and low cost.