Texture-and-Shape Based Active Contour Model for Insulator Segmentation

Yajie Yu, Hui Cao, Zhuzhu Wang, Yuqiao Li, Kang Li, Sheng Quan XieIEEE Access 2019 · DOI

Abstract

Insulator segmentation is a critical step for automatic insulator fault diagnosis in high voltage transmission systems. Existing methods fail to segment insulators when they have a low contrast with the surroundings. Considering the unique shape and texture characteristics of insulators, a texture-and-shape based active contour model is proposed for insulator segmentation. The segmentation is achieved by evolving a curve iteratively by the texture features and shape priors. In the texture-driven curve evolution, a semi-local region descriptor is used to extract the texture features of insulators and a new convex energy functional is defined based on the extracted features with the topology-preserving term. In the shape-driven curve evolution, the shape context descriptor is used to align the shape template with the current curve. The proposed method is applied to a set of images, and the experimental results confirm the efficacy and effectiveness of the proposed method for segmenting insulators in cluttered backgrounds.

Problem Statement

Automating the inspection of high-voltage transmission lines is essential for power grid reliability, but it starts with a difficult computer vision task: insulator segmentation. Insulators often have low contrast against their surroundings (like grey stone or dry grass) and appear in highly cluttered aerial backgrounds. Existing methods generally fall into two camps: patch-labeling, which requires massive amounts of manually labeled training data, and traditional curve evolution (Active Contours), which often suffers from "leakage" into background textures or splitting apart in noisy images.

Methodology

The authors propose a Texture-and-Shape Based Active Contour (TS-CV) model. The core idea is an iterative, two-stage evolution of a curve that leverages both local texture and global shape constraints.

Texture-Driven Evolution

The model uses a semi-local region descriptor within the Beltrami framework to map image textures into a higher-dimensional space. It defines a new convex energy functional to pull the curve toward the insulator's boundaries. Crucially, a topology-preserving term is included to prevent the curve from splitting or merging, ensuring it remains a single, continuous object as dictated by the shape template.

Texture feature map extracted using a semi-local region descriptor to identify insulator boundaries.

Shape-Driven Evolution

To maintain geometric consistency, the model uses Shape Context Descriptors to align a known shape template with the current evolving curve. It employs Procrustes analysis—a statistical method for shape comparison—to calculate the necessary transformation (scaling, rotation, translation) to nudge the curve back toward the expected geometry.

Datasets and Experimental Setup

The study utilized an insulator image dataset consisting of 100 aerial images representing real-world "in the wild" conditions with varying backgrounds and lighting. The authors used specific shape templates (Shape 1 and Shape 2) as priors to guide the segmentation of different insulator types.

Example of a 3-disk insulator shape template used as a prior in the shape-driven evolution stage.

Key Results

The proposed TS-CV model was compared against three other variations: the standard Chan-Vese (CV) model, a texture-only model (T-CV), and a shape-only model (S-CV).

Method Error Rate (ER%) Computation Time (s)
Chan-Vese (CV) 33.21% 1.15
Texture-based (T-CV) 20.79% 2.01
Shape-based (S-CV) 8.45% 10.14
Proposed (TS-CV) 4.39% 10.99

The results show a significant improvement in accuracy. While the standard CV model failed significantly (33.21% error), the hybrid TS-CV model achieved a highly precise 4.39% error rate, proving it can distinguish insulators from complex backgrounds where other models see only noise.

Conclusion and Future Work

The TS-CV model successfully addresses the challenges of low-contrast and cluttered backgrounds by combining texture features with rigid shape priors. However, the primary drawback is computational speed. At nearly 11 seconds per image, the model is currently too slow for real-time processing on a drone. Future work will focus on optimizing the shape-driven evolution process to enable real-time application for autonomous power line maintenance.

Summary updated

Abstract

Insulator segmentation is a critical step for automatic insulator fault diagnosis in high voltage transmission systems. Existing methods fail to segment insulators when they have a low contrast with the surroundings. Considering the unique shape and texture characteristics of insulators, a texture-and-shape based active contour model is proposed for insulator segmentation. The segmentation is achieved by evolving a curve iteratively by the texture features and shape priors. In the texture-driven curve evolution, a semi-local region descriptor is used to extract the texture features of insulators and a new convex energy functional is defined based on the extracted features with the topology-preserving term. The topology-preserving term keeps the curve's topology unchanged as the curve topology is determined by the shape template. In the shape-driven curve evolution, the shape context descriptor is used to align the shape template with the current curve. The semantic transformation between the shape template and the current curve is obtained by Procrustes analysis and then adopted to update the current curve to resemble the shape prior. The proposed method is applied to a set of images, and the experimental results confirm the efficacy and effectiveness of the proposed method for segmenting insulators in cluttered backgrounds.
BibTeX
@inproceedings{yu2019texture,
  author = {Yajie Yu and Hui Cao and Zhuzhu Wang and Yuqiao Li and Kang Li and Sheng Quan Xie},
  title = {{Texture-and-Shape Based Active Contour Model for Insulator Segmentation}},
  booktitle = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2922257},
}
Figure