ABSTRACT Laboratory chemical analysis of leaf samples can be costly and time-consuming, making it impractical for assessing crop variability. To address this challenge, researchers have focused on developing non-invasive tools that aid nitrogen (N) management, maximizing profits, minimizing environmental impact, and meeting market demands. This study aimed to develop a computer vision-based classifier system for assessing the N status in bean crops. An experiment was conducted in a greenhouse, involving five treatments (0%, 50%, 100%, 150%, and 200% N of the recommended dose) with six replications, totaling 30 pots containing six seedlings of Phaseolus vulgaris L. beans in four different phenological phases (V4, R5, R6, and R7). Digital RGB images of the bean canopies were captured using a camera at four-week intervals (30, 37, 44, and 51 days after emergence - DAE). The images were manually labeled to create an image database based on N status. Four different computational N status classifiers were developed by training a Convolutional Neural Network (CNN), one for each DAE. The classifiers were evaluated using confusion matrix metrics (accuracy, precision, and recall), resulting in an overall accuracy of about 80% when evaluating nitrogen status at five levels. Improved results were achieved by grouping the saturation classes of the 150% and 200% treatments with the 100% class (>=100% class), yielding an accuracy of 97% for 30 and 44 DAE. Promising results aside, this method opens new possibilities for improvement and application to other treatments, electromagnetic spectrum bands, and crops.

ABSTRACT Direct seeding systems are characterized by soil disturbance only occurring in the sowing line and the maintenance of the remaining area is achieved by covering with crop residues. Without total area soil mobilization as recommended by the no-tillage system, associated with the increase in the degree of soil compaction caused by the traffic of agricultural machinery, there can be a decrease in crop productivity. This study aimed to evaluate machine parameters, soil mobilization, and corn yield as a function of different furrow opener working depths (0.065, 0.095, 0.125, and 0.150 m) and evaluate the direct seeding of winter ground cover plants (black oat or forage turnip) in compacted Oxisol. The experiment was arranged in strips, in a randomized block design, with eight treatments and four replications. The deepening of the furrow opener shank increased the strength and power required at the drawbar in the seeding operation, as well as increased the soil mobilization in the seeding line. However, it did not prove to be an efficient management practice to increase corn productivity.

Este trabalho teve como objetivo avaliar o uso de índices espectrais, retirados de imagens digitais, para discriminar diferentes doses de N no feijoeiro. O trabalho, conduzido em vasos de 8 dm³, teve cinco tratamentos (0; 50; 100; 150 e 200 kg de N ha-1), com dez repetições. As imagens foram adquiridas aos 30; 40 e 50 dias após a emergência. Foram desenvolvidas funções discriminantes quadráticas, tendo como vetores de entrada as médias dos "pixels" de diferentes combinações dos quatro índices espectrais testados. Três diferentes tamanhos de blocos de imagem foram testados 9 x 9; 20 x 20 e 40 x 40 "pixels". Os melhores resultados foram alcançados pelos blocos de 9 x 9 e 20 x 20 "pixels", apresentando classificação 94; 96 e 96% superior à classificação ao acaso para os blocos 9 x 9 "pixels" e 92; 94 e 94% para os blocos 20x20 "pixels" aos 30; 40 e 50 dias após a emergência, respectivamente.

The objective of this work was to evaluate image spectral indices to detect nitrogen levels applied to bean plants. The experiment was in pots with 8 dm³ of soil, five treatments (0; 50; 100; 150 and 200 kg of N ha-1) with ten repetitions. The images were taken at 30; 40 and 50 days after emergence. Quadratic discriminant functions were developed, and the input vector was the average of the pixels from different combinations of four spectral indices. Three different sizes of image blocks were tested: 9 x 9, 20 x 20 and 40 x 40 pixels. The best results were obtained when using the 9 x 9 and 20 x 20 pixels blocks, with classification performance of 94; 96 e 96% better than a random classification for the blocks 9 x 9 pixels and 92; 94 and 94% for the blocks 20 x 20 pixels at 30; 40 e 50 days after emergency, respectively.