Fast training of support vector machines using sequential minimal optimization
Advances in kernel methods
| Hi-index | 0.00 |
The paper describes a new module of the developed robotic sewerage inspection system. The sewerage pipe is inspected by a remotely controlled inspection tractor equipped by a camera head able to rotate and zoom. This contribution describes a method and a software solution which allows to align the new inspection video and the archived video of the same pipe section (typically captured ten years ago). The aim of the analysis is to see how the pipe defects develop in time. The alignment of videos based on correspondences sought in images is overambitious. We have chosen the pragmatic approach. The text information from odometer which is superimposed in the video is automatically located and recognized using Optical Character Recognition (OCR) technique. The recognized distance from man-hole of the pipe allows to align both videos easily. The sewerage rehabilitation expert can then use only one remote control of the VCR for video positioning. This contribution describes the proposed solution, briefly mentions its implementation and demonstrate its function on practical sewerage inspection videos. However, our indexing approach can be used with any videos with superimposed text.