SIRT: Machine-Learning-based Selective Intensity Range Thresholding for Aircraft Visual Docking Guidance Refinement and Interpretation

We propose an automated Visual Docking Guidance System (VDGS) message interpretation service towards a smart and safe aviation endeavor. Owing to the priority of aviation safety criticality and the simultaneous need for smart airport, ground marshallers are getting replaced by VDGS to automatically detect obstacles, probable wingtip collisions and provide suitable assisted parking guidance to pilots. Nevertheless, the discrete presence of light-emitting diodes (LED) in VDGS dot-matrix display coupled with adversarial climatic conditions and far, low-light visibility creates the barrier in automated message detection and interpretation services. In this paper, we propose a novel Selective Intensity Range Thresholding (SIRT) for learning degraded LED pixel intensities and generate a refined synthetic display image. Furthermore, we propose an end-to-end pipeline for automatic recognition of VDGS alphanumeric texts and interpretation of symbologies to infer docking messages. The proposed solution aims to provide automated notification to the pilots about interpreted situational awareness messages from VDGS and record VDGS LED malfunction for rectification.