PROCESSAMENTO DIGITAL DE IMAGENS COMO ESTIMATIVA DE TEOR DE N FOLIAR NA CULTURA DO MILHO

Abstract

Nitrogen (N) is a nutrient that is directly related to the plants development and proper nitrogen fertilization can provide harvest in less time and increase productivity. The N have been handled incorrectly and often non-satisfactory supplied to the land. Proper management depends on constant monitoring of nutritional status that can be accomplished by predicting the leaf nitrogen content. This work aims to evaluate digital image processing algorithms, in order to establish a valid method to estimate the nutritional stress of N (nitrogen) on maize, and the feasibility of its implementation on mobile devices using photos sampled in the field under natural lighting environment. It was found that the thresholding process should use the HSV factor, more suitable under natural light if the sampling in the field was conducted between 11:15 am and 14:55. Among the algorithms we highlight the results of the DGCI (Dark Green Color Index) obtained from factor HSV (Hue, Saturation, Vibration), with significant correlation to leaf analysis of corn (R² = 0.92) and the adoption of NDVI (Normalized Difference Vegetation Index) obtained from the RGB histogram (Red, Green, Blue). This is the same method studied considering the FieldScout CM 1000 and GreenSeeker chlorophyll meters used in this research. No significant correlation among the readings was found when considering values from chlorophyll meters obtained in any moment of the day or from different light conditions. However, there is a positive correlation among the FieldScout CM 1000 and GreenSeeker, corresponding to sampling among 11h16m and 14h32m (R² = 0.75) and among 12h15m and 14h53m (R² = 0.92), as well is possible to estimate the IRC of maize based on the DGCI to V5 to V7 phenological stages directly at crop field using a mobile device.