Title :
Image processing technique for segmenting microstructural porosity of laser-welded thermoplastics
Author :
Leboeuf, Karl ; Makaremi, Iman ; Muscedere, Roberto ; Ahmadi, Majid
Author_Institution :
Dept. of Electr. & Comput. Eng., Univ. of Windsor, Windsor, ON, Canada
Abstract :
Plastics are used in a truly vast number of applications, and research is continously carried out to improve every aspect of the plastics industry. A recent study of laser transmission welding [1] required cross-sectional images of the weld´s microstructure to be analyzed for the presence of pores, which are tiny bubbles that may form during the weld process. It is believed that the number and size of pores may be indicative of the weld strength [1]. The current state of the art for detecting these pores involves manually drawing a contour around each one; a laborious process given that a typical sample may have hundreds-to-thousands of pores. This paper presents a segmentation system for classifying the pixels of a microstructural image of a thermoplastic laser weld as either belonging to a pore or the background. The algorithm is robust in terms of dealing with noise from flbreglass strands, cloudy pores, and varying exposure time. On average, it is estimated that the proposed algorithm is able to correctly classify pores at a rate of approximately 90% without requiring any user intervention.
Keywords :
image classification; image resolution; image segmentation; laser beam welding; plastics; porosity; production engineering computing; cross sectional images; image classification; image pixels; image processing technique; image segmentation; laser transmission welding; microstructural porosity; plastics industry; thermoplastics; weld strength; Artificial neural networks; Humans; Image segmentation; Neurons; Noise; Training; Welding; Artificial Neural Networks; Image Processing; Laser Welding; Thermoplastics;
Conference_Titel :
Electronics, Circuits and Systems (ICECS), 2011 18th IEEE International Conference on
Conference_Location :
Beirut
Print_ISBN :
978-1-4577-1845-8
Electronic_ISBN :
978-1-4577-1844-1
DOI :
10.1109/ICECS.2011.6122385