Resumen

GARCIA-CRUZ, Edgar et al. Identification with probabilistical neuronal networks of deficiences of iron and manganese by using digital images from bean leaves (Phaseolus vulgaris L.). Agrociencia [online]. 2015, vol.49, n.4, pp.395-409. ISSN 2521-9766.

The visual symptomatology of nutriment deficiencies, like iron (Fe) and manganese (Mn) in plant leafs is similar in their coloration and the kind of leaf they present on. A method based on the analysis of digital images of the leaves, capable to discriminate the differences of such deficiencies is required. The aim of this research was to analyze digital images of common bean (Phaseolus vulgaris L. var. Cacahuate), in order to identify differences in the Fe and Mn lesions in the initial development stage, when it is possible to revert damages with fertilization. To do so, we used a classifier created with probabilistic neuronal networks. The experimental treatments were: 1) partial deficiency (DP) of Fe (50 %); 2) DP of Mn (50 %); 3) total deficiency (DT) of Fe (0 %); 4) DT of Mn (0 %); 5) Fe/Mn interaction (0 % Fe, 0 % Mn); 6) control (100 % Fe, 100 % Mn), with 10 repetitions; Steiner solution was used as reference. The mean values of eight color and three texture variables from digital images of six common bean leaf samples were obtained; these were of 100X100 pixels (360 total samples) in 74 dds. These mean values were used as entry variables to generate the classifiers with a cascade correlation algorithm of the Fe and Mn deficiency treatments. The classifiers that only considered textural characteristics had correct global classification of symptoms less or equal to 44 %. In contrast, the highest percentage of correct global classification of the classifiers in the test was of 76.6 % with six variables, which included texture and color characteristics, and six exit classes of difference treatments. The reduction of the number of classes did not increase the percentage of correct classification in the test.

Palabras llave : chroma; entropy; RGB color space; local homogeneity; hue; second angular momentum.

        · resumen en Español     · texto en Español     · Español ( pdf )