Defesa de TCC: Thiago Estrêla Kalid
- Curso: Engenharia Elétrica (Campus Curitiba)
- Aluno: Thiago Estrêla Kalid
- Orientador: Prof. Thiago Alberto Rigo Passarin (DAELT-CT)
- Título: DESENVOLVIMENTO DE MÉTODO DE ODOMETRIA VISUAL PARA INSPEÇÕES POR IMERSÃO
- Quando: 12/fev/2025 às 09:30
- Local: Sala de Multimeios Didáticos (CI-003)
- Link: https://meet.google.com/rrh-yqpw-roi
- Banca:
- Prof. Jean Carlos Cardozo da Silva (DAELT-CT/UTFPR)
- Prof. Daniel Rodrigues Pipa (DAELN-CT/UTFPR)
- Prof. Thiago Alberto Rigo Passarin (DAELT-CT/UTFPR)
Abstract: Non-destructive testing ultrasound inspections are commonly used for structural health assessment of vital parts of the infrastructure of the oil & gas industry. Subsea pipelines are especially prone to inner and outer corrosion, as they usually operate in harsh and corrosive environments. During an inspection, moving the ultrasonic transducer is common practice as it increases the probability of detecting a discontinuity or flaw. This routine is called a sweep and requires information about the performed trajectory to properly trigger the acquisition and reconstruct spatial (3D) models of the specimen. This role is performed by an odometer, a device capable of estimating displacement based on some sort of measurement. The most common type is mechanical odometers, but commercially available alternatives for underwater 2D displacements are too bulky and complex. In this work it was proposed a visual odometer — a new category of odometer for nondestructive testing that, based on image processing, estimates 2D and 3D displacements during an inspection. The proposed solution employs a monocular camera pointed toward the specimen encased in a waterproof vessel. A Raspberry Pi 4 was used for image processing and communication to the US acquisition system. Experiments were made to test the proposed solution under controlled and freehand trajectories. Through controlled well-known trajectories, the visual odometer in a standalone operation estimated 2D and 3D trajectories with cumulative error ranging from 0.05 % to 2.8 % of the traveled distance. The uncontrolled routine simulated the conditions found in a real-world pipeline inspection along with the US acquisition system. It demonstrated the system’s ability to be part of a broader inspection apparatus. Overall, the experiments helped to sustain the idea that the proposed solution could be part of a reliable odometry system. If noted improvements are made, the visual odometer offers good potential for use in subsea pipeline inspections.
Publicação relacionada:
- Kalid, Thiago E., et al. “Virtual encoder: a two-dimension visual odometer for NDT.” in: Proceedings of the 13th European Conference on Nondestructive Testing (ECNDT 2023).