The Lenaeroproekt Institute has integrated an AI-powered photogrammetry service developed by Cloud X to analyze airport runway imagery. The system automatically detects and classifies surface defects, reducing the processing time for a single runway from ten working days to one.
The AI model, trained on the CX AI Platform, performs pixel-level segmentation to identify cracks and joints. It processes high-resolution images of 10,000 by 50,000 pixels, with individual image files starting at 1 GB in size. The automated processing of a single image takes approximately five minutes.
Engineers retain control over the final assessment, using a browser-based interface to review, edit, or add defect markings. The service generates detailed reports, calculates total defect lengths, and tracks historical data to compare runway conditions over time, with all manual adjustments preserved for future analysis.
The project involved a collaborative training process where experts from Lenaeroproekt and Cloud X prepared datasets containing labeled cracks and joints. The model was specifically trained to distinguish between actual damage and surface seams, incorporating rare defect types to improve classification accuracy.
