Trabalho em anais de eventoOpen Access
Exploiting domain knowledge for content-agnostic template matching in structured environments
RODRIGUES, RÔMULO T.; SOUZA, MARCELO; MOURA, DAVIDSON ROBERT; VARELLA, JERRY; BESSA, IURY; VINENTE, KENNY
Resumo
Resumo
Mobile robots operating in structured environments can exploit the semantic information embedded in alphanumeric identifiers to improve navigation and task execution. Traditional computer vision methods and deep learning approaches struggle to detect identifiers under viewpoint variability and onboard computational constraints. We propose a perception algorithm that leverages domain knowledge and 3D geometric reconstruction to locate and decode alphanumeric identifiers. The method projects the planar surface into a synthetic orthographic view using relative pose estimation and homography constraints, achieving invariance to perspective, rotation, scale, and translation. It then uses a content-agnostic template capturing only character layout for efficient detection. Results using an off-the-shelf OCR engine demonstrate that our approach detects the identifier and recovers its alphanumeric string even under severe viewpoint distortion.
Autoria
Autores do grupo (2)
A lista completa de autores está no topo da página. Abaixo, apenas os pesquisadores do e-Controls vinculados a esta publicação.
palavras-chave
Palavras-chave
template matchingdomain knowledgestructured environmentsobject recognition
Status
Published
links
Links e arquivos
Trabalho em anais de evento