05 May 2021
14:00 Master's Defense Fully distance
Theme
Detection and tracking of chickens: a new computational methodology for monitoring patterns in commercial poultry
Student
Renan Silva Ramalho Vilas Novas
Advisor / Teacher
Fabio Luiz Usberti
Brief summary
This work aims to develop a framework, based on computer vision techniques, for detecting and monitoring chickens in commercial broiler houses. From the proposed framework, it was possible to extract descriptive attributes related to the spatial distribution and movement of the chickens. These attributes can be used in warning systems, to detect anomalous events and situations of thermal discomfort in the aviaries. The identification and correction of these events in the aviaries, when carried out quickly, have the potential to guarantee the growth and welfare of the chickens. The main contribution of the proposed framework consists in transforming the event detection problem in commercial aviaries into a particle dynamics problem. From the proposed framework, it was possible to analyze the dynamics of distribution and movement associated with different types of events in the broiler houses. The computational methodology proposed in this work was validated through analysis of correlation between attributes obtained by the framework and indexes of distribution and activity in the literature. Through the analysis of peaks and valleys of the extracted attributes, it was possible to identify periods of intense activity and heterogeneous dispersion of the chickens, such as, for example, occasions when birds were fleeing humans walking in the aviary.
Examination Board
Headlines:
| Fábio Luiz Usberti | IC / UNICAMP |
| Daniella Jorge de Moura | FEAGRI / UNICAMP |
| Hélio Pedrini | IC / UNICAMP |
Substitutes:
| Leandro Aparecido Villas | IC / UNICAMP |
| Mário César San Felice | DC / UFSCar |