Manhole Detection using Image Processing on Google Street View imagery


Manholes are an essential part of the maintenance of both sewer lines and stormwater drains which are an integral part of rainwater management during monsoons. Open and broken manholes have been a cause of loss of life and injury for many years especially during rains when flooded roads cause low visibility. The positioning of said manholes according to other architecture such as metro stations, buildings and bridges shall endorse careful infrastructure management and help build efficient utility networks. The current methods used for keeping track of public utility networks lack technologies that can effectively eradicate human error while data management such as data-logging, database duplicity managers and de-seasonalization of manhole status data. Present methods involve heavy fieldwork to keep the map and status of manholes updated. Google street view images can be used to detect these manholes. This paper presents a method to detect manholes on a location basis. Its implementation can be supplemented with avoidance of the path of the manhole in times of emergencies and adversities. The areas explored in this paper are image pre-processing, detection methodology and correlation model. For pre-processing the images bilateral filtering, equalized histogram and Canny edge detection methods were used. After pre-processing the images, contour detection with convex hull method was deployed. For further detection, a method that uses pixel-wise iteration to eliminate undesirable contours was used. Finally, a regression model was created for finding a relation between the distance of manhole from the point of image capture and the area of the manhole in the image. This process can be further extrapolated for obtaining the GPS coordinates of the manhole.



Software And Hardware

• Hardware: Processor: i3 ,i5 RAM: 4GB Hard disk: 16 GB • Software: operating System : Windws2000/XP/7/8/10 Anaconda,jupyter,spyder,flask Frontend :-python Backend:- MYSQL