Coffee ripeness monitoring is a key indicator for defining the moment of starting the harvest, especially because the coffee quality is related to the fruit ripeness degree. The most used method to define the start of harvesting is by visual inspection, which is time-consuming, labor-intensive, and does not provide information on the entire area. There is a lack of new techniques or alternative methodologies to provide faster measurements that can support harvest planning. Based on that, this study aimed at developing a vegetation index (VI) for coffee ripeness monitoring using aerial imagery. For this, an experiment was set up in five arabica coffee fields in Minas Gerais State, Brazil. During the coffee ripeness stage, four flights were carried out to acquire spectral information on the crop canopy using two quadcopters, one equipped with a five-band multispectral camera and another with an RGB (Red, Green, Blue) camera. Prior to the flights, manual counts of the percentage of unripe fruits were carried out using irregular sampling grids on each day for validation purposes. After image acquisition, the coffee ripeness index (CRI) and other five VIs were obtained. The CRI was developed combining reflectance from the red band and from a ground-based red target placed on the study area. The effectiveness of the CRI was compared under different analyses with traditional VIs. The CRI showed a higher sensitivity to discriminate coffee plants ready for harvest from not-ready for harvest in all coffee fields. Furthermore, the highest R² and lowest RMSE values for estimating the coffee ripeness were also presented by the CRI (R²: 0.70; 12.42%), whereas the other VIs showed R² and RMSE values ranging from 0.22 to 0.67 and from 13.28 to 16.50, respectively. Finally, the study demonstrated that the time-consuming fieldwork can be replaced by the methodology based on VIs.

SEGMENTAÇÃO DE IMAGENS RGB USANDO DIFERENTES ÍNDICES DE VEGETAÇÃO E MÉTODOS DE LIMIARIZAÇÃO A segmentação é um dos aspectos fundamentais envolvidos no processamento de imagens, que geralmente consiste na discriminação de objetos de interesse e fundo da imagem. O presente estudo objetivou avaliar o efeito de diferentes índices de vegetação (IV) (ExG, ExGR e NDI) no desempenho de três métodos de limiarização (Otsu, Ridler e Triângulo) em termos de precisão e tempo de processamento na segmentação de imagens. Para tal, foram utilizadas 30 imagens advindas de área cultivada com milho sob diferentes tipos de cobertura do solo (plantio convencional, casca de café e palhada). O processamento das imagens foi realizado através de algoritmos desenvolvidos com base nos IV e métodos de limiarização. A acurácia das imagens resultantes foi avaliada com a verdade de campo obtida pelo algoritmo K-means. Os resultados demonstraram desempenho superior para o método do triângulo quando precedido dos índices NDI (90,7%) e ExGR (90,23%) e dos métodos de Otsu e Ridler quando precedidos pelo NDI com 89,06% e 89,03% de acurácia, respectivamente. O tempo de processamento foi estatisticamente igual entre os métodos avaliados. De modo geral, a abordagem combinada de IV e métodos de limiarização foram capazes de separar com alta acurácia a cultura do milho do objeto de fundo. Palavras-chave: processamento de imagens, imagens digitais, método do triângulo. ABSTRACT: Image Segmentation is one of the fundamental aspects involved in image processing, which generally consists of discriminating objects of interest from its background. Thus, the objective of this study was to evaluate the effect of vegetation indices (VI) (ExG, ExGR, and NDI) on the performance of three automated thresholding methods (Otsu, Ridler, and Triangle) in terms of accuracy and processing time on image segmentation. A set of 30 images from an area cultivated with maize under different types of soil cover (conventional planting, no-tillage with coffee husk, and straw residue) were selected and processed. The images were processed through algorithms developed based on VI and thresholding methods. Then, the accuracy of the resulting images was evaluated through the ground truth image obtained by the K-means algorithm. The results demonstrated superior performance for the triangle method when preceded by the NDI (90.7%) and ExGR (90.23%) indices and the Otsu and Ridler methods when preceded by the NDI with 89.06% and 89.03% accuracy, respectively. The processing time was statistically equal among the evaluated methods. In general, the combined approach of VI and thresholding based methods were capable of separating with high accuracy the maize crop from the background. Keywords: image processing, digital images, triangle method.

Delineation of site-specific nutrient management zones (MZ) provides a basis for practical and cost-effective management of spatial soil fertility in precision agriculture. Therefore, the objective of this study was the delineation of MZs in a soybean field using geostatistics, principal component analysis (PCA), and the fuzzy k-means algorithm. The study was carried out in a field with 204 ha located in São Desidério, western Bahia state, Brazil (12 ° 25 ‘ 12” S, 45 ° 29ʹ 46” W). To do so, samples of soil attributes (0–20 cm), soybean yield, electrical conductivity (EC) at 0.20 m (EC02), 0.50 m (EC05), 1.00 m (EC1), 2.00 m (EC2) soil profile depth, and the Normalized Difference Vegetation Index (NDVI) were obtained in 204 points (100 x 100 m grid). After soil sampling and laboratory analyzes, the data were submitted to descriptive statistics and a Spearman correlation analysis was performed to select those attributes related to soybean yield. Then, the spatial variability of these attributes was assessed and spatial distribution maps were constructed using geostatistical tools. Next, PCA and fuzzy k-means algorithm were then performed to delineate MZs. Finally, the agreement between the MZs maps obtained from the PCA and soybean yield was assessed using the Kappa index. Results showed that the optimal number of MZs was two, which resulted in a Kappa index of 0.61 (very good). Moreover, the analysis of variance indicated heterogeneity between all attributes analyzed in the MZs. Finally, the defined MZs provide a basis of information for site-specific nutrient management.

The increasing market interest in coffee beverage, lead coffee growers around the world to adopt more efficient methods to select the best-quality coffee beans. Currently, coffee beans selection is carried out either manually, which is a costly and unreliable process, or using electronic sorting machines, which are often inefficient because some coffee beans defects, such as sour and immature beans, have similar spectral response patterns. In this sense, the present work aimed to assess coffee beans quality using both computer vision and machine learning techniques, such as Support Vector Machine (SVM), Deep Neural Network (DNN) and Random Forest (RF). For this purpose, an algorithm written in Python language was developed to extract shape and color features from coffee beans images. The obtained dataset was then used as input to the machine learning algorithms. The data reported in this study pointed to the importance of color descriptors for classifying coffee beans defects. Among the variables used, the components from RGB (Red, Green and Blue) and HSV (Hue, Saturation and Value) color spaces presented the most relevant contribution for the classification models. Also, the results reported in this study provides evidence that computer vision along with machine learning algorithms can be used to identify and classify coffee beans with a very high accuracy (> 90%). Key words: Deep neural network; classification; artificial intelligence; image processing; granulometry.

Spray nozzles can be considered the most important items of sprayers, and the ideal nozzle-pressure combination plays a major role in the effectiveness of pest control in the field. Spray nozzle manufacturers often develop, and make new alternatives available on the market; however, studies are always important to verify the behavior of tríese new technologies concerning the spectrum of the droplets produced. Therefore, the present study aimed to evaluate the spectrum of droplets generated by four spray nozzles (FC3D 100-02; VP 110-02; TR 110-02 and TT 110-02) under different working pressures (138, 207, 276, 345, and 414 kPa). For this, a factorial scheme in a completely randomized design with four replications was used. The parameters evaluated were volumetric median diameter (VMD), relative amplitude (Span), and the percentage of droplets with a diameter smaller than 100 um (<100 um). The spray nozzles models and the working pressure had a direct effect on the droplet size. Furthermore, there were no significant differences between the TR 110-02 and VP 110-02 models for the percentage of spray volume containing droplets <100 um neither for the VMD. Lastly, the increase of pressure resulted in an increasing linear effect on the Span index for the FC3D 100-02 nozzle.

Spatial interpolation methods are frequently used to characterize soil attributes’ spatial variability. However, inconclusive results, about the comparative performance of these methods, have been reported in the literature. Therefore, the present study aimed to analyze the efficiency of ordinary kriging (OK) and inverse distance weighting (IDW) methods in estimating the soil penetration resistance (SPR), soil bulk density (SBD), and soil moisture content (SM) using two distinct sampling grids. The soil sampling was performed on a 5.7 ha area in Southeast Brazil. For data collection, a regular grid with 145 points (20 x 20 m) was created. Soil samples were taken at a 0.20 m layer depth. In order to compare the accuracy of OK and IDW, another grid was created from the initial grid (A), by eliminating one interspersed line, which resulted in a grid with 41 sampled points (40 x 40 m). Results showed that sampling grid A presented less errors than B, proving that the more sampling points, the lower the errors that are associated with both methods will be. Overall, the OK was less biased than IDW only for SBD (A) and SM (B) maps, whereas IDW outperformed OK for the other attributes for both sampling grids.