Industrial equipment detection algorithm under complex working conditions based on ROMS R-CNN.

Area of Science:

• Computer Vision
• Machine Learning
• Industrial Automation

Background:

• Industrial equipment detection is crucial for automation and safety.
• Existing methods struggle with complex conditions like rotation, occlusion, and multi-scale variations.
• Accurate detection under these challenging scenarios remains an open problem.

Purpose of the Study:

• To propose a novel deep learning algorithm, ROMS R-CNN (Rotation Occlusion Multi-Scale Region-CNN), for robust industrial equipment detection.
• To enhance detection accuracy and reduce errors in complex industrial environments.
• To provide a more reliable solution for automated industrial monitoring and control.

Main Methods:

• Utilized MobileNetV2 as a feature pyramid network to integrate multi-scale information.
• Developed a rotation anchor mechanism using k-means clustering for aspect ratio and angle, addressing tilted equipment.
• Implemented a Non-Maximum Suppression algorithm with penalty factors to handle overlapping objects.

Main Results:

• The proposed ROMS R-CNN algorithm demonstrated superior performance compared to existing methods in industrial equipment detection.
• Significantly reduced instances of missed detections and false positives.
• Achieved a mean Average Precision (mAP) of 0.939 on common industrial equipment datasets.

Conclusions:

• ROMS R-CNN offers a highly effective deep learning solution for accurate industrial equipment detection.
• The algorithm successfully overcomes limitations posed by rotation, occlusion, and multi-scale variations.
• This advancement contributes to improved reliability and efficiency in industrial automation systems.