Target detection of helicopter electric power inspection based on the feature embedding convolution model

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Aerospace Engineering

Background:

• Traditional helicopter electric power inspection faces limitations in scope and real-time capability.
• Need for advanced models to process aerial imagery for efficient power infrastructure monitoring.

Purpose of the Study:

• To enhance helicopter electric power inspection using an improved Feature Embedding Convolution (FEC) model.
• To address challenges of limited inspection scope and poor real-time performance.
• To improve the accuracy and efficiency of electric power target identification and fault detection.

Main Methods:

• Simulation experiments and model analysis to determine optimal keyframes and flight trajectories.
• Development and application of an improved Feature Embedding Convolution (FEC) model for real-time, large-range feature extraction from aerial images.
• Optimization of the FEC model using reinforcement learning for enhanced fault detection capabilities across diverse environments.

Main Results:

• The improved FEC model demonstrated a 30% increase in accuracy for electric power circuit and equipment detection compared to traditional algorithms.
• Inspection range was expanded by 26% using the proposed FEC model.
• Reinforcement learning optimization led to over a 36% performance increase in fault detection.

Conclusions:

• The proposed FEC model significantly improves the accuracy and scope of helicopter electric power inspections.
• The optimized model offers a more scientific and efficient strategy for electric power inspection and fault detection.
• This advancement contributes to ensuring the reliability and efficiency of power infrastructure monitoring.