Tag Archives: computer vision

306–318 S. Mancini, G.A.S. Ferraz, F.M. de Oliveira, S.V. Rubio, D. Cecchin, G.M. Reis, L. Conti, V. Becciolini and P.F.P. Ferraz
Prediction of body mass in dairy cows using LiDAR technology
Abstract |

Prediction of body mass in dairy cows using LiDAR technology

S. Mancini¹, G.A.S. Ferraz¹, F.M. de Oliveira¹, S.V. Rubio¹, D. Cecchin², G.M. Reis³, L. Conti⁴, V. Becciolini⁴ and P.F.P. Ferraz¹*

¹Federal University of Lavras, School of Engineering, Department of Agricultural Engineering, University Campus, PO Box 3037 - CEP 37200-000 Lavras, Minas Gerais, Brazil
²Fluminense Federal University, Department of Agricultural and Environmental Engineering, Rua Passo da Pátria, 156, PO Box, 21065-230, Niterói, Brazil
³Florida International University, Knight Foundation School of Computing and Information Sciences, Miami, Florida, USA
⁴University of Florence, Department of Agriculture, Food, Environment and Forestry, IT50145 Florence, Italy
*Correspondence: patricia.ponciano@ufla.br

Abstract:

Precision livestock farming has driven the search for technological solutions that enable remote and non-invasive monitoring of animal health and zootechnical performance. This study aimed to investigate the use of computer vision for predicting body mass (BM) in a herd of Holstein Friesian dairy cows using three-dimensional images acquired via LiDAR (Light Detection and Ranging). From these images, the individual body volume (BV) of each animal was digitally estimated. Subsequently, the correlation between LiDAR-derived BV and body mass measured using the conventional weighing method was evaluated. Based on this relationship, statistical regression models were fitted to predict BM from BV. The results showed that BV measured using LiDAR achieved a coefficient of determination (R²) of 0.7861 and a mean absolute percentage error (MAPE) of 4.43%. These findings demonstrate the feasibility of LiDAR technology as an effective, non-invasive alternative tool for estimating bovine body mass. The use of LiDAR enabled detailed recording of morphological characteristics and measurements, allowing continuous monitoring of body development without the need for direct animal handling. In conclusion, the incorporation of computer vision systems into dairy cattle management can optimize operational efficiency, reduce labor requirements, and promote a more sustainable production system, while enhancing animal welfare through automated monitoring.

Key words:

computer vision, dairy farming, regression model

164-179 F.M. Oliveira, P.F.P. Ferraz, G.A.S. Ferraz, D. Cecchin, A.F.S. Stopatto, V. Becciolini and M. Barbari
Evaluation of chest circumference in 3D lateral images of dairy cattle farming for body mass prediction
Abstract |

Full text PDF (1 MB)

Evaluation of chest circumference in 3D lateral images of dairy cattle farming for body mass prediction

F.M. Oliveira¹, P.F.P. Ferraz¹*, G.A.S. Ferraz¹, D. Cecchin², A.F.S. Stopatto³, V. Becciolini⁴ and M. Barbari⁴

¹Federal University of Lavras, Department of Agricultural Engineering,
University Campus, PO Box 3037 - CEP 37200-000 Lavras, Minas Gerais, Brazil
²Fluminense Federal University, Department of Agricultural and Environmental Engineering, Rua Passo da Pátria, 156, PO Box, 21065-230, Niterói, Brazil
³Federal University of Lavras, Department of Animal Science, University Campus, PO Box 3037 – CEP 37200-000 Lavras, Minas Gerais, Brazil
⁴University of Florence, Department of Agriculture, Food, Environment and Forestry, IT50145 Florence, Italy
*Correspondence: patrícia.ponciano@ufla.br

Abstract:

The advancement of precision livestock farming has underscored the importance of developing innovative and non-invasive methods for monitoring animal health and productivity. In this context, this study evaluated the application of computer vision to estimate the body mass (BM) of Holstein-Friesian dairy cows using 3D images captured laterally with the Intel RealSense D435i depth camera. The methodology involved correlating chest circumference (CC) measurements obtained in the field with those derived from lateral 3D images. A total of 250 animals were analyzed, with BM ranging from 420 to 855 kg, and the relationship between CC and BM was modeled using regression techniques. The results indicated a coefficient of determination (R² = 0.88) and a mean absolute percentage error (MAPE) of 3.94% for CC measured in the field. For CC derived from 3D images, R² was 0.847, with an MAPE of 5.29%. Although the 3D image-based method showed a slight reduction in accuracy, it demonstrated significant potential as a non-invasive and efficient alternative for estimating BM in dairy cows. Furthermore, the study highlights the role of 3D imaging technologies in acquiring detailed morphological data, enabling a more comprehensive understanding of body composition dynamics over time. These findings reinforce the potential of integrating digital technologies into dairy farming, promoting sustainable, precise, and labor-efficient management practices.

Key words:

3D images, computer vision, dairy cattle farming, non-invasive body mass estimation

574–585 S. Polyanskikh, I. Arinicheva, I. Arinichev and G. Volkova
Autoencoders for semantic segmentation of rice fungal diseases
Abstract |

Full text PDF (937 KB)

Autoencoders for semantic segmentation of rice fungal diseases

S. Polyanskikh¹, I. Arinicheva², I. Arinichev³* and G. Volkova⁴

¹Plarium Inc., 75/1 Uralskaya Str., RU350001 Krasnodar, Russia
²Kuban State Agrarian University named after I.T. Trubilin, 13 Kalinina Str., RU350044 Krasnodar, Russia
³Kuban State University, 149 Stavropolskaya Str., RU350040 Krasnodar, Russia
⁴All-Russian Research Institute of Biological Plant Protection, 1 VNIIBZR Str., RU350039, Krasnodar, Russia
*Correspondence: iarinichev@gmail.com

Abstract:

In the article, the authors examine the possibility of automatic localization of rice fungal infections using modern methods of computer vision. The authors consider a new approach based on the use of autoencoders – special neural network architectures. This approach makes it possible to detect areas on rice leaves affected by a particular disease. The authors demonstrate that the autoencoder can be trained to remove affected areas from the image. In some cases, this allows one to clearly highlight the affected area by comparing the resulting image with the original one. Therefore, modern architectures of convolutional autoencoders provide quite acceptable visual quality of detection.

Key words:

autoencoder, computer vision, fungal diseases, neural network