Détection de défauts de surface sur le bois par capteur tactile capacitif

Abstract

The rise in popularity of collaborative robots is creating new automation opportunities that were not previously viable for many small and medium-sized businesses. This is the case for the inspection of sanding and polishing tasks on wood products. In order to perform this inspection, a robot must be equipped with sensors that allow it to feel the surface for defects. In this thesis, we explored the possibility of using a capacitive tactile sensor to detect the presence of surface defects on wood. Initially, we used the four modalities of the CoRo tactile sensor to train a variety of artificial neural networks to detect the presence of three types of surface defects : scratches, indentations, and tearouts. None of the modalities were able to detect the presence of scratches given their very small dimensions. That being said, the static modality, which measures the pressure pattern on the sensor’s surface, performed best for indentations and tearouts. We then tested whether the fusion of modalities improves the detection of defects. Contrary to what we had anticipated, the results show that modality fusion does not necessarily lead to better performance than using the individual modalities. Nevertheless, we confirmed that the static modality, used alone or in combination with the dynamic modality, is the most useful. In sum, we have demonstrated that the CoRo capacitive tactile sensor represents a promising solution for the detection of surface defects in wood. However, it should be redesigned with this task in mind to achieve maximum performance.

Date	5 Jan 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Vincent Duchaine (Supervisor)