FPGA implementation of computer vision techniques for taxiway centreline detection

Abstract

This project deals with manoeuvring problems encountered during the aircraft taxiing phase. In adverse weather conditions and at night, taxiway and runway markings are less visible posing a greater challenge to the pilots and increasing the chance of an accident. The work focused on the research and development of imagining systems and image processing techniques which can be used to provide taxi guidance in all weather conditions. The system processes images of taxiways and runways and through feature extraction techniques, detects the edges and other select features, such as a taxiway centerline, to identify the path the aircraft should take to remain on the taxiway. A polynomial fit on the centerline is used to identify the ideal path, the heading error and cross-track error. A Kalman filter is implemented to track these parameters in order to reduce the effect of noise. A prototype of the system was developed in the Matlab environment and was tested on real and synthetic imagery. The tests validated the performance of the system in different operating conditions, including low visibility and at night. Tests were also performed on infrared video capture in live trials at Malta International Airport. The developed algorithms were finally implemented on an embedded device (FPGA). These had to be optimized for the implementation so that t hey could provide accurate results in real-time without consuming excessive FPGA space in order to allow efficient utilization of the electronic resources available. The FPGA design was tested with images that were used for testing the prototype and the results were compared. Differences were small and acceptable, thus verifying the operation of the FPGA implementation.