Rotorcraft Landing Sites Identification: Scaling and Generalization of the AI Model

Rotorcraft Landing Sites Identification: Scaling and Generalization of the AI Model

Rotorcraft Landing Sites Identification: Scaling and Generalization of the AI Model

Status: Completed

Funding Agency: U.S. Department of Transportation

Abstract: 

The Federal Aviation Administration and the Department of Transportation need accurate records of where rotorcraft landing sites are and what type they are, but acquiring, verifying, and updating that information is difficult. This project built an autonomous system that authenticates the coordinates in the FAA master landing site database and searches for helipads across large designated areas. The approach uses a convolutional neural network that learns helipad features from aerial imagery. The team assembled a database by checking FAA and other records against Google Earth satellite imagery, then trained and validated several CNN architectures to select a model.

Investigators: Dr. Nidhal Carla Bouaynaya, PhD., Dr. Mohammad Jalayer, PhD., Dr. Ghulam Rasool, PhD.

Links to Publications:

Rotorcraft Landing Sites Identification: Scaling and Generalization of the AI Model, CAIT-UTC-REG54, June 2022