A Novel Centralization Method for Pipe Image Stitching

Abstract

The creation of unwrapped stitched images of pipework internal surfaces is being increasingly used to augment routine visual inspection. A significant challenge to the creation of these stitched images is the need to estimate the pose and position of the camera for each frame, which is often alleviated through the use of a mechanical centralizer to ensure the camera is held in the center of the pipe. This article proposes a novel method for image centralization and pose estimation, which is particularly beneficial to circumstances where mechanical centralization is impractical. The approach involves post-inspection centralization of the captured video, by first estimating the probe's position relative to the pipe, using an integrated laser ring projector combined with the camera sensor, and then using this position to unwrap the image, so it produces an undistorted view of the pipe interior (equivalent to unwrapping a centralized view). These unwrapped images are then stacked to produce a stitched image of the pipe interior. In this paper pose estimation was successfully demonstrated to have a 90% confidence interval of ±0.5 mm and ±0.5° in translation and rotation changes. This pose estimation is then used to create stitched images for both a visual test card image mounted inside a pipe and an aluminum pipe sample with artificial defects, in both cases demonstrating near equivalent results to those obtained using traditional mechanical centralization.

A Simultaneous Pipe-Attribute and PIG-Pose Estimation (SPPE) Using 3-D Point Cloud in Compressible Gas Pipelines

“…To mitigate this problem, two exteroceptive sensing approaches have been investigated in the literature: laser image processing and visual odometry. The laser image processing approach estimates the robot pose or pipe diameter by analyzing the shape of the reflected laser image patterns projected onto the camera image plane [ 13 , 14 , 15 , 16 , 17 , 18 ]. In [ 13 ], Kim et al propose a laser system consisting of four point lasers, a hyperbolic mirror, and an omni-directional camera, and present an algorithm that estimates the rotation angles yielding a specific light pattern on the image plane of the omni-directional camera.…”

“…In [ 13 ], Kim et al propose a laser system consisting of four point lasers, a hyperbolic mirror, and an omni-directional camera, and present an algorithm that estimates the rotation angles yielding a specific light pattern on the image plane of the omni-directional camera. A conical laser system is also proposed in [ 14 , 15 , 16 , 17 , 18 ], where a conical laser beam is radiated to the inner wall of the downstream pipe and its reflected light on the camera image plane is analyzed for the estimation of the robot pose or pipe attributes. The robot pose is estimated by the matching pose from the feature database in [ 14 ], or computed by the non-linear optimization formulations in [ 15 , 16 ].…”

“…Two major approaches to exteroceptive sensing have received significant attention from academia and industry: laser image processing and visual odometry . The laser image processing approach irradiates a laser beam pattern on the inner wall of the pipeline and estimates the pipe attributes or robot pose through the mathematical modeling of the reflected light captured at a camera image [ 13 , 14 , 15 , 16 , 17 , 18 ]. On the other hand, the visual odometry approach extracts visual features from a camera image, and estimates the relative robot pose by associating them with the corresponding features in the subsequent camera images [ 19 , 20 , 21 , 22 ].…”

“…Although both exteroceptive sensing approaches are suitable for tethered or self-propelled in-pipe robots, they still face three common limitations in the long-distance ILI of compressible gas pipelines: First, the projection of the inner pipe wall surface onto the camera image plane results in an intrinsic loss of depth information that needs to be recovered from either the complex mathematical model of laser image processing or the association of visual features over subsequent frames. Second, their estimation is incomplete in the sense that they require prior information such as pipe radius with an assumption of a perfectly round cross section in [ 13 , 14 , 15 , 16 , 19 , 20 , 21 ], which may cause additional estimation errors to the pipes with different nominal thicknesses or high ovality due to deformation. Third, a high variation of PIG speed can significantly deteriorate the estimation quality of the inner wall geometry, either due to a single plane observation of the pipe surface in [ 13 , 14 , 15 , 16 ] or an erroneous association of blurred visual features in [ 19 , 20 , 21 , 22 ].…”

“…Second, their estimation is incomplete in the sense that they require prior information such as pipe radius with an assumption of a perfectly round cross section in [ 13 , 14 , 15 , 16 , 19 , 20 , 21 ], which may cause additional estimation errors to the pipes with different nominal thicknesses or high ovality due to deformation. Third, a high variation of PIG speed can significantly deteriorate the estimation quality of the inner wall geometry, either due to a single plane observation of the pipe surface in [ 13 , 14 , 15 , 16 ] or an erroneous association of blurred visual features in [ 19 , 20 , 21 , 22 ].…”