João Célio Luna de Carvalho

The present study was conducted at the Federal Rural University of Rio de Janeiro, in the years 2024 and 2025, to evaluate the performance of three supervised classifiers - maximum likelihood, random forest, and support vector machine - for spectral discrimination between maize (Zea mays L.) and the weed (Cyperus rotundus), considering different phenological stages and two cropping seasons (main and second crop), using multispectral images acquired by remotely piloted aircraft. Overall accuracy, as well as class metrics of precision, recall, and F1-score, were used to evaluate the performance of the algorithms in classifying four targets: maize, weed, soil, and shadow. The performance of the classifiers improved as the phenological stage progressed. In the first year (second crop), the V8 phenological stage showed the greatest differentiation between the plants, and the highest-performing classifier was maximum likelihood, with an overall accuracy of 89%. For the maize class, precision was 0.88, recall was 0.81, and F1-score was 0.85, and for the weed class, precision was 0.68, recall was 0.89, and F1-score was 0.77. In the second year (main crop), the best differentiation occurred at the V6 stage, and the maximum likelihood classifier again presented the best performance, with an overall accuracy of 88%, precision of 1.00, recall of 0.78, and F1-score of 0.88 for the maize class, and precision of 0.52, recall of 1.00, and F1-score of 0.68 for the weed class. Thus, across the years analyzed, the maximum likelihood classifier performed best.

Computer vision systems have been successfully applied to automate processes related to the selection and classification of food based on quality standards. In this context, this research aimed to develop a computer vision system comprising an image acquisition device and software capable of estimating the physical parameters of tomato fruits from digital images. The software was developed in Python and implemented on a Raspberry Pi microprocessor, enabling both image capture and the calculation of fruit physical parameters (volume, sphericity, and surface area). The physical parameters of 45 tomatoes from three classes (Saladette, Italian, and Cherry) were estimated using the proposed system and compared with reference values obtained through traditional methods. The comparison indices revealed a high correlation and agreement between actual and estimated values for all physical parameters when analyzing fruits across all classes. The models generated from the total set of fruits (without class distinction) achieved R2 values greater than 0.95 for all physical parameter estimates. The computer vision system enabled the measurement of tomato physical parameters, presenting approximate estimates to the reference values obtained by traditional methods. This approach represents a viable alternative for estimating the sphericity, surface area, and volume of tomato fruits.

Computer vision systems can be an alternative to traditional methods of analyzing the quality of forage crops, allowing the instantaneous, non-destructive monitoring of the crop, with cost reduction. This study aimed to evaluate the quality parameters of Tifton 85 (Cynodon spp.) using digital images, relating spectral indices to the quality parameters of this forage. In the experimental area, four levels of nitrogen fertilization were applied and the analyses were made at different times after the standardization cut (14, 28, 42, and 56 days). The quality parameters evaluated were mineral matter, crude protein, and neutral detergent fiber. From images obtained in the visible (RGB) and near-infrared (RGNIR) spectral regions, spectral indices were generated. Principal component analysis was applied to summarize the information obtained by spectral indices into a single principal component (PCI). PCI associated with spectral indices was related to forage quality parameters for each cutting time using simple quadratic regression models. The relationships between mineral matter and spectral indices were variable over time. Crude protein and neutral detergent fiber showed the highest relationships with the spectral indices obtained by RGNIR images already at the initial times. Thus, although the RGB images have shown satisfactory results to obtain information about the quality of Tifton 85, the NIR band tends to increase the reliability of the relationships at early times.