Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Textureless Object Recognition: An Edge-based Approach

Frincy Clement, Kirtan Shah, Dhara Pancholi, Gabriel Lugo Bustillo, Irene Cheng

TL;DR

This work tackles textureless object recognition by leveraging edge-centric features and extensive data augmentation to compensate for limited discriminative texture. It builds 15 large datasets (≈340,000 images each) that explore edge-only, edge-combination, and edge-enhanced RGB representations, and evaluates four classifiers to determine the most effective configuration. The key finding is that RGB images enhanced with a combination of all three edge detectors (Canny, HED, Prewitt) provide the best overall accuracy, with HED edges offering the strongest single-feature signal. The study also reveals robustness with white-background test images but reduced performance when backgrounds vary, underscoring the need for background-aware or clutter-robust approaches in practical settings.

Abstract

Textureless object recognition has become a significant task in Computer Vision with the advent of Robotics and its applications in manufacturing sector. It has been challenging to obtain good accuracy in real time because of its lack of discriminative features and reflectance properties which makes the techniques for textured object recognition insufficient for textureless objects. A lot of work has been done in the last 20 years, especially in the recent 5 years after the TLess and other textureless dataset were introduced. In this project, by applying image processing techniques we created a robust augmented dataset from initial imbalanced smaller dataset. We extracted edge features, feature combinations and RGB images enhanced with feature/feature combinations to create 15 datasets, each with a size of ~340,000. We then trained four classifiers on these 15 datasets to arrive at a conclusion as to which dataset performs the best overall and whether edge features are important for textureless objects. Based on our experiments and analysis, RGB images enhanced with combination of 3 edge features performed the best compared to all others. Model performance on dataset with HED edges performed comparatively better than other edge detectors like Canny or Prewitt.

Textureless Object Recognition: An Edge-based Approach

TL;DR

This work tackles textureless object recognition by leveraging edge-centric features and extensive data augmentation to compensate for limited discriminative texture. It builds 15 large datasets (≈340,000 images each) that explore edge-only, edge-combination, and edge-enhanced RGB representations, and evaluates four classifiers to determine the most effective configuration. The key finding is that RGB images enhanced with a combination of all three edge detectors (Canny, HED, Prewitt) provide the best overall accuracy, with HED edges offering the strongest single-feature signal. The study also reveals robustness with white-background test images but reduced performance when backgrounds vary, underscoring the need for background-aware or clutter-robust approaches in practical settings.

Abstract

Textureless object recognition has become a significant task in Computer Vision with the advent of Robotics and its applications in manufacturing sector. It has been challenging to obtain good accuracy in real time because of its lack of discriminative features and reflectance properties which makes the techniques for textured object recognition insufficient for textureless objects. A lot of work has been done in the last 20 years, especially in the recent 5 years after the TLess and other textureless dataset were introduced. In this project, by applying image processing techniques we created a robust augmented dataset from initial imbalanced smaller dataset. We extracted edge features, feature combinations and RGB images enhanced with feature/feature combinations to create 15 datasets, each with a size of ~340,000. We then trained four classifiers on these 15 datasets to arrive at a conclusion as to which dataset performs the best overall and whether edge features are important for textureless objects. Based on our experiments and analysis, RGB images enhanced with combination of 3 edge features performed the best compared to all others. Model performance on dataset with HED edges performed comparatively better than other edge detectors like Canny or Prewitt.
Paper Structure (15 sections, 15 figures)

This paper contains 15 sections, 15 figures.

Table of Contents

  1. Introduction
  2. View-based
  3. Feature based method
  4. Shape Based Approach
  5. Related Work
  6. Proposed Method
  7. Image Data Acquisition
  8. Obtaining Ground Truth Data
  9. Data Balancing & Augmentation
  10. Summary of Results
  11. Conclusion and Future Work
  12. Acknowledgement
  13. Resources
  14. Author Biography
  15. Appendix

Figures (15)

  • Figure 1: Project Workflow
  • Figure 2: Data Preprocessing to obtain ground truth objects from initial acquired 2D images
  • Figure 3: Distribution of data across 30 classes
  • Figure 4: Results of Manual Augmentation
  • Figure 5: Results of Automatic Augmentation using Augmentor API
  • ...and 10 more figures