Mapping Electrical Poles Using DL
Keywords: YOLO, CUDA, Photogrammetry, Google Street View, Deep Learning, GPU
Maintaining the electric grid requires accurate maps of utility infrastructure. Traditional methods are labor-intensive. Leveraging Google Street View, deep learning, and GPUs, we developed an application to automate the detection and localization of electrical poles.
This application automates the labor-intensive process of mapping utility poles, providing a scalable and efficient solution. Accurate maps enhance planning, operational efficiency, and rapid damage assessment post-storm, supporting the maintenance and resilience of the electric grid.
Technical Approach:
• Detection Model: Utilized the YOLO (You Only Look Once) model for real-time detection of electrical poles in Google Street View images. YOLO's efficiency and accuracy make it ideal for this task.
• Processing Speed: Implemented CUDA for accelerated processing, enabling the handling of large datasets and real-time analysis.
• Localization Accuracy: Enhanced the location accuracy of detected poles using photogrammetry techniques. By analyzing multiple images and their spatial relationships, we refined the positional data of the poles.
• Integration: Combined these technologies into a cohesive system that processes Google Street View images, detects electrical poles, and accurately maps their locations.
WDVA Information
Certification Number
WDVAARKS23
ARKSOFT INC