Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Τεχνικές εξαγωγής σημείων ενδιαφέροντος σε τούνελ από σκυρόδεμα

Davaris Konstantinos

Full record

---

Summary

Recently, the detection of structural defects (cracks and spallings) has been of great interest. Several tunnels have been in operation for many years, while the cost of a new tunnel construction is very high. Occasionally, there have been many accidents, which require the immediate inspection and detection of their structural state.In this diploma thesis, we developed an algorithm, which uses and assesses five different detectors from the Computer Vision Matlab toolbox. Our ultimate goal is to detect cracks and spallings in the inner concrete lining of a tunnel. The data used are 37 real-time images of the tunnel V-S-H in Switzerland. Each one of the 37 images, has its respective annotation image, which includes the exact coordinates of the defective points of the concrete lining. We use these real-time images combined with their respective annotation images, so as to find and separate the defective points (cracks and spallings), from the structurally non-defective points. These points are used to assess the detectors selected, thus to reach a useful conclusion. In conclusion, SURF descriptor yields the most robust results as regards the total defects of concrete tunnel. On the contrary, Harris presents the best performance in detecting spalling defects while BRISK as for the cracks.

Available Files

Services