Contour Primitive of Interest Extraction Network Based on Dual-Metric One-Shot Learning for Vision Measurement

Although many existing vision measurement systems have achieved high performances, they are object-specific and have limitations in flexibility. Toward intelligent vision measurement that can be conveniently reused for novel objects, this article focuses on the image geometric feature extraction with one-shot learning ability. We propose a contour primitive of interest (CPI) extraction network with dual metric (CPieNet-DM), which can obtain a designated CPI in a query image of a novel object under the guidance of only one annotated support image. First, the dual-metric learning mechanism is proposed, which not only utilizes inter-image similarity as guidance but also leverages the intra-image coherency of CPI pixels to facilitate the inference. Second, a neural network is designed to infer the CPI map based on the dual metric, which also predicts the CPI's geometric parameters. Moreover, the dual context aggregator is plugged in to provide the awareness of both images’ contexts. Third, the network training is jointly supervised by the multiple tasks of dual-metric learning, geometric parameters regression, and CPI extraction. The online hard example mining is utilized to improve the training outcome. The effectiveness of the proposed methods is validated with a series of experiments.

I. Introduction

Image-based vision measurement is a noncontact and real time intelligent perception technology focusing on the structure and poses of a target object, which is widely applied in industrial sensing and robotic automation [1], [2]. After capturing one or more object images, first, a set of geometric features that are of interest to the measurement task should be precisely extracted. Second, the geometric calculation is conducted using these geometric features to obtain the measurement results. For example, to monitor the bolt looseness in flanges, the bolt angles were measured with the line and ellipse contours in the bolt image [3]. The front face contour and spline corner features were extracted to calculate the pose of large gears based on monocular vision [4]. For the robotic peg-in-hole assembly task, the hole pose was measured based on circle features and binocular vision [5]. Line segment features were used to indicate the pose errors of two precise components in image spaces [6]. However, the above-mentioned methods are object-specific, which cannot be conveniently reused for a novel object.