MCLAS: An Image Segmentation-based Alarm System for Manhole Cover Loss

摘要

Alarm of the loss of sewer manhole covers is very important in municipal administration. An integrated distributed system named MCLAS (Alarm System for Manhole Cover Loss) is designed for automatically discovering and notifying the loss of sewer manhole covers. The system consists of three modules: Camera Calibration Module, Alarm Module, and Distributed System Module. We have to locate the manhole cover's position which can be regarded as ellipse in image in Camera Calibration Module. It is the main topic of MCLAS. In the field of pattern recognition, HT (Hough Transform) is a powerful tool for finding predefined features in digital image and is always applied to locate shape which has clear analysis geometry feature. However, HT is very memory and time consuming. An improved HT based on contour tracking is proposed in MCLAS, whose required computational storage space and time consumption are cheaper than the classical HT. Experimental results on typical scenarios show that the correct locating rate is up to 86%.