“In response to the problems of low feature matching accuracy, noise in predicted depth maps, and incomplete edge reconstruction in multi view stereo tasks of remote sensing images, the research team proposes a diffusion constrained multi view stereo network called DiffusionMVS. By designing a multi-scale feature extraction module MFE-FPN, an adaptive feature aggregation module AFA, and a diffusion constrained optimization model DCM, the average absolute error index was improved by 28.11% and 3.37% respectively in the WHU-TLC and LuoJia MVS datasets, providing a new solution for high-precision remote sensing 3D reconstruction.”