Advanced Image Analytics for Mobile Robot-Based Condition Monitoring in Hazardous Environments: A Comprehensive...

Area of Science:

• Robotics and Automation
• Artificial Intelligence
• Data Science

Background:

• Mobile robots are vital for condition monitoring (CM) in hazardous environments like mines, collecting extensive image data.
• Processing large volumes of image data and noise presents challenges for identifying thermal anomalies.
• Existing industrial big data analytics struggle with mobile robot-generated image datasets.

Purpose of the Study:

• To develop an integrated approach for efficient processing and visualization of thermal anomaly data from mobile inspection robots.
• To address the limitations of big data analytics in handling complex image datasets from hazardous industrial sites.
• To enhance condition monitoring processes through advanced data analysis techniques.

Main Methods:

• A novel dimension reduction procedure combining semantic segmentation (VGG16 CNN) for feature selection.
• Application of Random Forest (RF) and Extreme Gradient Boosting (XGBoost) classifiers for pixel class label prediction.
• Exploration of unsupervised learning with PCA-K-means for dimension reduction and classification of unlabeled thermal defects.

Main Results:

• The proposed methodology effectively handles image-based CM tasks in hazardous environments.
• The approach demonstrated robust performance in processing and visualizing thermal data from real-world mobile robot inspections.
• Successful identification and classification of thermal anomalies and defects were achieved.

Conclusions:

• The integrated approach significantly enhances the efficiency of condition monitoring processes in hazardous industrial settings.
• The methodology proves effective in managing and interpreting large-scale thermal image data collected by autonomous systems.
• This work validates the practical application of advanced AI and data reduction techniques for industrial inspection robots.