This article explores the special matching technique for infrared and visible light image coordinates using the Depth Anything V2 AI model. The model estimates the depth of both image types, obtaining pixel depth information. It then matches spatial coordinate points via coordinate transformation and algorithms. The matched images are fused using the RFN-Nest fusion network, combining infrared thermal radiation and visible image details. The fused image enhances visible image texture details while retaining infrared target prominence, improving clarity and information content for better image analysis and target identification.

Шао Ю. (науч. рук. Васильев А.С.) Spatial matching of infrared and visible image coordinates based on depth anything v2 ai model // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2025]. URL: https://kmu.itmo.ru/digests/article/14397