Recognition of inside pipeline geometry by using monocular camera and PSD sensors

Abstract

To deploy an untethered robot inside pipelines autonomously, it is prerequisite to recognize pipeline elements such as straight pipeline, elbow, T-branch and miter. This paper presents an algorithm for recognizing pipeline elements using a monocular camera and PSD (Position Sensitive Device) sensors. Previously, we have presented a method that uses special features of the pipelines (such as elbows or branches) as landmarks by recognizing the shadows of the elbows and branches. Though it is effective, it has limitations and we propose a method of using PSD sensors as a supplementary sensor to overcome the weakness. The method is easy to implement, but helps us collect accurate information necessary for navigating inside pipelines without heavy computation. The method is composed of two parts, which is the method for distinguishing T-branch and miter and searching the direction in T-branch and elbow respectively. The design for the sensing module is presented and the algorithm for recognition procedures is addressed. The sensing module is implemented in an in-pipe robot, called MRINSPECT VI, and its performances are validated.