What is Aimee?
Aimee is Searidge Technologies’ advanced neural network framework for development of Artificial Intelligence (AI) based solutions for Air Traffic Control and airport efficiency.
Searidge has been working with artificial intelligence for several years with its’ vision processing/remote tower technology. Leveraging this experience into the development of Aimee and extending the platform to include new functional areas has laid the foundation for an exciting new era in ATM technology. Aimee has been developed to greatly simplify the configuration and training of neural networks with large and complex data sets; to allow the continuous evaluation and testing of output, and most importantly, to predict and certify performance within a safety critical context.
Aimee has four key functional areas: Computer Vision Processing (Remote/Digital Tower Video); Natural Language Processing (Controller-Pilot Radio Telephony); Flight Data/Surveillance Processing and Analysis; Weather Processing. Aimee allows ATC and airport stakeholders to leverage deep learning in each of these functional areas to build robust and safe AI-powered applications.
The current revolutionary advantage of AI technology is to recognize patterns in enormous amounts of data at speeds that greatly surpass human capabilities. The benefit of the pattern recognition capabilities of AI become even more apparent when data from different sources such as video sensors, ATC radio, ADS-B, and airport operations systems are processed simultaneously. Then the neural network is able to find data patterns across all input sources which results in the ultimate sensor fusion across the entire ATC systems environment. Aimee would then, for example, be able to alert ATC stakeholders about potential incidents before they happen by comparing data from all sources to data patterns from previous analyzed and learned incidents.
The application possibilities in ATC and airport operations environments are virtually limitless. As airports and aerodromes continue to grow and have more and more capacity, AI can provide the required assistance to controllers and operators to control an ever increasing number of movements safely and efficiently. Searidge Technologies has always placed a great value on innovation in the ATC industry and the Aimee AI platform is designed to enable stakeholders from all levels of ATC to solve problems which previously seemed unsolvable.
An Example of How Aimee Works
Aimee is able to segment out aircraft in airport scenes by comparing the live images to an aircraft model, which it previously learned by processing thousands of annotated training images. The training image set used by Searidge to train Aimee is very extensive and includes images from various airports that feature different environments, weather, traffic levels, and aircraft types. The extensive training data set allows the neural network to learn an abstract model of aircraft which allows the recognition of never before seen aircraft models. Using this abstract aircraft model, the neural network achieved revolutionary detection performance at diverse deployment sites.
The Aimee AI platform presents a great improvement over the previous state of the art aircraft detection technologies. Algorithms such as feature detection and background subtraction have numerous limitations which include not being easily adaptable to different environmental and weather conditions and different aircraft sizes. The neural network aircraft detection featured in Aimee is able to overcome many of these challenges to provide among other things excellent probability of detection in all environmental and weather conditions, aircraft occlusion resolution, classification of detected objects, and detection of aircraft that are only partially visible in the video sensor field of view. Additionally, the neural network is robust to camera vibration and shaking which commonly occur for mast installations of camera sensors. Object detection and classification are however only the tip of the iceberg of the enormous potential AI has to increase the safety and efficiency of air traffic control and airport operations.
Aimee in Action: Fort Lauderdale-Hollywood International Airport
Fort Lauderdale-Hollywood International Airport (FLL) is located in the Greater Miami area, Florida and with two runways is one of the top 30 busiest (by passenger count) airports in the United States. According to the Federal Aviation Administration (FAA), FLL had approximately 290 thousand aircraft operations and served just over 29.2 million passengers in 2016. FLL serves as the hub for many airlines and is also home to the headquarters of Silver Airways.
Searidge was selected by FLL to provide its intelligent video platform, as a key technology in their Virtual Ramp Control System (VRCS). Previous to the VRCS, the FLL ramp traffic was controlled from a small structure adjacent to the ramp area. The VRCS achieved the airport’s goal to manage the ramp traffic remotely from the airport operations center in a safe and efficient manner. Not only does the VRCS provide a more cost efficient solution, but it also ensures that the ramp traffic can be controlled with minimal line of sight limitations as the airport continues to grow.
Searidge provided the ramp controllers at FLL with real time video footage from an array of video sensors, covering nearly the entire airport surface. Aimee, the Searidge AI platform, then used advanced object detection and recognition artificial neural network technology to detect aircraft on the airport surface by processing the video streams from 31 thermal video sensors installed across the airport. The system then uses these aircraft detections and their proximity to the airport gates to provide FLL staff with gate statuses for almost all gates. The processing of video information to provide gate statuses is commonly referred to as a Gate Video Processing (GVP) system. AI-based Gate Video Processing systems like FLL, enable among other things automatic gate time billing and the collection and analysis of gate usage statistics.