Robust Vision-Based Target Outline Reconstruction and In-Plane Trajectory Measurement

Abstract

Noncontact vision-based in-plane displacement and posture trajectory measurement is essential in structural health monitoring, aerospace, and related fields. However, current techniques such as optical flow and digital image correlation (DIC), remain challenges under intensity-varying and noise-interfered test scenes. To this end, by leveraging on binary segmentation and new line fitting method, this article introduces a novel target outline reconstruction (TOR) method for dynamic target tracking, creating an approach for robust vision-based in-plane displacement and posture parameter measurements. The reconstruction process begins with target-background separation, applying a binarization algorithm combined with edge extraction to obtain the initial target outline. In parallel, image processing with convolution forms the gradient value coordinate, filtering the densest regions to net the qualified points for edge line reconstruction. The TOR method was evaluated in various test scenes with background light shadowing. Experimental results demonstrate that target displacement and posture parameters are well monitored with high accuracy and robustness.