Detalhes bibliográficos
| Ano de defesa: |
2019 |
| Autor(a) principal: |
Bravo, Daniel Trevisan
 |
| Orientador(a): |
Araújo, Sidnei Alves de |
| Banca de defesa: |
Araújo, Sidnei Alves de,
Pamboukian, Sergio Vicente Denser,
Quaresma, Cristiano Capellani,
Belan, Peterson Adriano,
Alves, Wonder Alexandre Luz |
| Tipo de documento: |
Tese
|
| Tipo de acesso: |
Acesso aberto |
| Idioma: |
por |
| Instituição de defesa: |
Universidade Nove de Julho
|
| Programa de Pós-Graduação: |
Programa de Pós-Graduação em Informática e Gestão do Conhecimento
|
| Departamento: |
Informática
|
| País: |
Brasil
|
| Palavras-chave em Português: |
|
| Palavras-chave em Inglês: |
|
| Área do conhecimento CNPq: |
|
| Link de acesso: |
http://bibliotecatede.uninove.br/handle/tede/2570
|
Resumo: |
The current panorama of diseases caused by the Aedes aegypti mosquito in Brazil and worldwide has motivated numerous research efforts in various areas of knowledge. In addition to health prevention campaigns, the technology proves to be a great ally, using unmanned aerial vehicles (UAVs) to acquire aerial images, facilitating the work of health surveillance teams. However, such images are usually analyzed manually (visually) and may require a lot of time from health agents. This work proposes a computer vision approach for the automatic identification of objects and scenarios that represent potential breeding sites of the Aedes aegypti mosquito, from aerial images of urban areas acquired by UAVs. The proposed approach includes 4 steps: composition of orthomosaics, identification of suspicious objects and scenarios, detection of small portions of water and generation of annotated orthomosaics and reports. To detect suspicious objects and scenarios, two techniques were explored: convolutional neural networks RNC and Bag of Visual Words BoVW combined with the Support Vector Machine classifier SVM (BoVW + SVM), and the results obtained were measured using the mean Average Precision mAP-50. In object detection using a YOLOv3 model RNC, we obtained the rate of 0.9610 for mAP-50, while in the scenario detection task, we compared the results of tiny-YOLOv3 RNC and BoVW + SVM, the respective rates of 0.9028 and 0.6453 were obtained. These results suggest that the RNCs are sufficient to identify potential breeding sites since together they led to the average rate of 0.9319 for mAP-50. Regarding the detection of small portions of water, the experiments conducted obtained the value of 0.9757 for the measure of similarity Structural Similarity Index SSIM. The results obtained in the experiments involving the 4 steps showed that the proposed approach can make significant contributions to the implementation of computer systems aimed at assisting health agents in the planning and execution of activities to combat Aedes aegypti mosquito with the use of UAVs. |
