Visual Locating of Reactor in an Industrial Environment Using the Composite Method

Area of Science:

• Robotics and Automation
• Computer Vision
• Industrial Engineering

Background:

• Automated unloading in sherardizing requires precise reactor pose and position detection.
• Traditional image processing and deep learning methods have limitations in industrial environments with variable lighting and dust.

Purpose of the Study:

• To develop a robust method for locating reactors for automatic unloading.
• To overcome the challenges posed by industrial environmental conditions like luminance and dust.

Main Methods:

• Analysis of defects in classic image processing and deep learning for reactor localization.
• Application of an improved You Only Look Once (YOLO) model to identify handling holes.
• Development of a handling hole corner detection method using image morphology and Hough transform.
• 3D handling hole model creation based on binocular stereo vision principles.

Main Results:

• The improved YOLO model effectively identifies the region of interest for handling holes.
• The proposed method achieves high precision in position recognition (within 4.64 mm) and pose estimation (within 1.68 °).
• The system meets operational requirements at a distance of approximately 5 meters from the reactor.

Conclusions:

• The developed method provides an effective solution for accurate reactor localization in challenging industrial environments.
• The approach enables precise pose and position determination, facilitating automated processes like reactor unloading.
• The findings demonstrate the feasibility and effectiveness of the proposed computer vision system for industrial automation.