A Real-Time Acquisition and Segmentation System for Thermographic Images

D.F. García, R. Usamentiaga, I. Marín, and J.A. González (Spain)


Thermographic image, segmentation, merging method.


This article shows the techniques involved in the real-time acquisition and segmentation of thermographic images taken of very long flat products. In particular, these techniques have been applied to steel strips, with average dimensions of 1,2m in width by 10.000m in length, and with a temperature range from 100ºC to 200ºC. A novel description technique adapted to this kind of thermographic images is proposed in this work. The image is described as a sequence of sections, each of them composed of several regions with the average same spatial and thermal properties. The thermographic sensor used acquires a thermal 1D image (scan) while the strip passes under it. Enhanced scans are obtained by applying new filtering techniques based on oversampling. The aggregation of the successive enhanced scans generates the 2D-image of the strip while it is being acquired. Also, new methods of thermal and spatial calibrations are presented in this article. To obtain the desired description, each new enhanced scan is segmented using a merging method based on fuzzy decision criteria. Next, if the regions of the segmented scan match the regions of the current section, it is aggregated to the section. Otherwise, it starts a new candidate section. The techniques developed allow the representation of the thermographic images with maximum simplicity while retaining a level of detail which allows the representation to be used, for further purposes, such as inspection or defect detection.

Important Links:

Go Back