MACHINE VISION FOR ANALYSING THE POSITION OF FASTENING NAILS ON WOODEN RAILWAY SLEEPERS

摘要

Wooden railway sleeper inspections in Sweden and to arnlarge extent elsewhere are carried out manually by arnhuman operator; visual inspection being the mostrncommon approach. Manually inspecting railway sleepersrnis slow and time consuming. Machine vision algorithmsrninvestigating surface cracks on the sleeper and sinking ofrnthe metal plate have been studied for the purpose ofrnautomating the task. In this particular article, informationrnconcerning how far the fastening nail has lifted out ofrnposition is investigated with an aim of using suchrninformation while assessing the condition of the sleeper.rnLaser beams incident on the sleeper have been used tornhighlight the geometrical form of the sleeper/plateailrncomplex. Digital images of the nail were acquired mimicrnhuman visual capabilities. Appropriate image analysisrntechniques were applied to further process the images andrnnecessary features were extracted. Results ofrnunsupervised learning, achieved in the current workrnindicate that expectation maximization algorithmrnproduced reliable results.