Title: Line Detection by Centre and Width Estimation

Year of Publication: Nov - 2014
Page Numbers: 87-98
Authors: Wei Hua Chen
Conference Name: The International Conference on Artificial Intelligence and Pattern Recognition (AIPR2014)
- Malaysia


Lines, thin regions of approximately-uniform width and colour, are important in the visual interpretation of information. Lines are a primary feature of many objects, and are used to convey semantic information in the human world, e.g. road markings. In order to interpret such information, an efficient method is desired for the detection of lines under this basic definition, independent of the colour and shape of the lines to be detected. A method of obtaining the centre-line and width of arbitrarily-coloured and -curving salient lines is described in this paper. However, this method alone is shown to be inadequate for its originally-intended purpose, the detection of wires in the disassembly of end-of-life electronics, due to complications such as twisted wires and the large number of other lines present in images taken during assembly. Examples are also provided of the application of the described method to other contexts.