In the last years the detection and classification of surface defects of material is assuming great importance. Visual inspection can help to increase the product quality and, in particular context, the maintenance of products. The railway infrastructure is a particular field in which the periodical surface inspection of rolling plane can help an operator to prevent critical situation. A defect on rolling surface appears generally as a grey level variation useful for its classification. Main idea is to utilize the image processing to help a human operator in the detection of defects on the rolling surface. The prototype realised uses two Dalsa line scanner camera SP-12 to acquire the left and right rail image with a sampling rate of 2 mm per line. An encoder connected to the axel box with 2 mm resolution generates the line acquisition trigger for the cameras. The left and right images are processed to extract the rolling surface strip by image and to classify defects. We use neural network to tracking the rolling surface in the image. This method is able to track the rail also in the switch, cross level, and so on. The detection of defect uses a gradient oriented approach to emphasis the image regions with grey level variation. Four directions 0°, 45°, 90°, and 135° will be considered as defect principal direction. The union of all four normalised histograms is used as input sample for a neural network classifier. A test phase has been performed on a real trolley.
展开▼
