Engineering Vehicles Detection Based on Modified Faster R-CNN for Power Grid Surveillance

Area of Science:

• Computer Science
• Electrical Engineering
• Artificial Intelligence

Background:

• Power grid security is crucial for operational stability.
• Engineering vehicle intrusion poses a significant threat to power grid infrastructure.
• Existing surveillance methods lack accuracy in complex environments.

Purpose of the Study:

• To develop an intelligent surveillance method for detecting engineering vehicles near power grids.
• To improve the accuracy and reduce false alarms in intrusion detection systems.
• To enhance the security and operational integrity of power grid infrastructure.

Main Methods:

• Proposed an intelligent surveillance method based on the Faster R-CNN framework.
• Modified the Faster R-CNN network by adjusting ROI pooling layer position and adding convolutional layers.
• Utilized transfer learning on a custom dataset of engineering vehicle images from transmission line corridors.

Main Results:

• The modified Faster R-CNN model demonstrated high accuracy in detecting engineering vehicles.
• The system achieved a low false alarm rate in realistic, complex monitoring scenes.
• Increased depth in the feature classification part proved effective for detection.

Conclusions:

• The proposed intelligent surveillance method is effective for early warning of power grid intrusions.
• The system enhances power grid security by accurately identifying unauthorized engineering vehicles.
• The method offers a reliable solution for monitoring critical infrastructure against external threats.