Runway Incursion Detection Using Correlated Vehicle and Video Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in generating sufficient and accurate training data to effectively train machine learning models for detecting runway incursions, which are critical for ensuring airport safety.
Innovation Solution
The proposed airport object location system combines vehicle location units, an electro-optical sensor system, and a model generator to correlate vehicle location information with video frames from an aircraft, creating a dataset that enhances the accuracy of machine learning models for incursion detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors and data sources are integrated to improve detection accuracy, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent combines multiple data sources including electro-optical sensor data, vehicle location unit data, and reference vehicle data into a unified dataset. This merging of heterogeneous data sources improves measurement precision for incursion detection while managing system complexity through integrated processing.
Solution Approach 2:
The system uses multi-functional components that can process different types of data (video frames, location information, timestamps) through a unified correlation framework. The model generator serves multiple purposes by training on diverse data types to detect various incursion scenarios, reducing the need for separate specialized systems.
2Loss of information
If comprehensive vehicle location information and video data are collected to improve dataset quality, then information completeness improves, but loss of time increases
Solution Approach 1:
The system performs preliminary correlation of vehicle location information with video frames using timestamps and location data before model training. This preliminary organization of comprehensive data reduces the time required for actual detection by having the dataset pre-processed and structured in advance.
Solution Approach 2:
The patent uses reference vehicles with sensors that create copies of the operational environment from different perspectives. These reference datasets serve as training data that captures comprehensive information without requiring simultaneous collection from all sources during actual detection, reducing time loss.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An airport object location system comprising a number of vehicle location units, a sensor system, and a model generator. The number of vehicle location units is connected to a number of vehicles. The number of vehicle location units generate vehicle location information for the number of vehicles in an area including an operations surface at an airport and vehicle timestamps for the vehicle location information. The sensor system is connected to a reference vehicle. The sensor system is configured to generate sensor data for the area, wherein reference timestamps and reference location information are associated with the sensor data. The model generator is configured to correlate the vehicle location information for the vehicles with the sensor data using the vehicle timestamps, the reference location information, and the reference timestamps to form a dataset.