Study On The Edge Detection With Improved Canny And Three-dimensional Reconstruction For The Damaged Blade

Blade is the heart of the engine, in order to save the cost of renewal, it has significantly value to repair the blade which is worth to fix up. According to statistics, the blades which need to weld are more than half of the total to repair. At present, the engine blades of China’s civil aviation are all sent abroad to repair, which is a waste of time and seriously hampers to the development of the civil maintenance industry. Therefore, it is significant to research welding equipment with our own intellectual property. In order to locate the contours to be welded accurately and restore the site topography of defects quickly, it is necessary to establish a vision detection system for positioning.The primary objective to establish vision system is to restore the three-dimensional contours of the blade, which is helpful to layer slices in the process of welding. Most of the current recovery mode is offline measurement, which gets the three-dimensional contours according to the points cloud by three-dimensional measurement. The method is cumbersome and consuming seriously, which need to locate the edge once again before welding. A new method was proposed to solve those problems. The method adopted online vision detection system which based on recovery of the blade edge extended contour. A trinocular vision system was built to get information of the twisted blade in multi dimensions. This system restored three-dimensional morphology of the edge defects through three-dimensional mosaic and extension of the key points in each view point, which is beneficial to get reason for layering in the process of welding. Simulation for the blade which tip part existed defects showed that the precision of restoration reached to 0.2mm, therefore the new method can meet the requirements of welding and lift efficiency significantly.In order to ensure the accuracy of trajectory planning, it is required to accurately detect the image edge in each field once again after the blade is clamped. Proposing a way based on extreme value suppression to overcome the Canny shortage in the process of detection for discontinuous edge. Specifically, the continuity factor of edge was brought in to improve the traditional way which refrains points of edge exceedingly. At the same time, adaptive filtering and the Otsu method was adopted to reduce error coming from artificial selection, which increased the adaptability of the algorithm. Experiments show that the method has a great advantage in edge detection especially for continuous edge like blade.