Uma metodologia probabilística para combinação de detectores de pessoas
Ano de defesa: | 2015 |
---|---|
Autor(a) principal: | |
Orientador(a): | |
Banca de defesa: | |
Tipo de documento: | Tese |
Tipo de acesso: | Acesso aberto |
Idioma: | por |
Instituição de defesa: |
Universidade Federal de Minas Gerais
UFMG |
Programa de Pós-Graduação: |
Não Informado pela instituição
|
Departamento: |
Não Informado pela instituição
|
País: |
Não Informado pela instituição
|
Palavras-chave em Português: | |
Link de acesso: | http://hdl.handle.net/1843/BUBD-A7NK8B |
Resumo: | The coexistence of robotic systems and human beings is increasing every day. Because of that, these systems should be designed to interact safely with people, being essential the presence of an advanced perception of the environment and the ability to detect peopleeciently. Once the detection of people can be challenging in environments containing other objects, modern people detectors are not completely reliable since they usually fail to detect people and also make false detections. In this context, this doctoral thesis proposesa methodology to combine high-level information from several people detectors using probabilistic techniques. The objective is to exploit the individual advantages of several people detectors yielding in a more accurate and complete information than the one given by a single detector alone. Thus, when more than one detector nd a person at the same position, the condence of detection is increased. The detectors rely on information from one or more sensors, such as cameras and laser rangenders. The detectors' combination allows the prediction of the position of the persons inside the sensors' elds of view and, in some situations, outside them. Also, the fusion of the detector's output can make people detection more robust to failures and occlusions. The proposed methodology is based on a recursive Bayes Filter, whose prediction and update models are specied in function of the detectors used, taking into account a measure of condence assigned to them, which is obtained experimentally. The concepts of prediction and update are used in the steps of the methodology, making use of temporal information to improve the quality of detection. Experiments were executed with a mobile robot that collects real data in a dynamic environment containing people while moving autonomously. The implementation of the methodology uses a local semantic grid to represent the robot's local workspace, which contains probability values related to the presence of people in specic regions of the workspace. The obtained results indicate the improvements brought by the approachin relation to a single detector alone and show that it is possible to obtain a larger number of detections of the people keeping the number of false detections low |