Deep learning-based standardized lattice field transformation change detection method

Abstract

The invention discloses a standardized lattice field transformation change detection method based on deep learning. The method comprises the following steps: step 1, obtaining a remote sensing image; 2, drawing vector data; 3, making a data set; step 4, model training; 5, applying the model; according to the method, the Transs-SGSLN model is adopted, the capability of capturing the long-distance dependency relationship of the model is increased by utilizing the Transform, the change area in the dual-tense image is effectively captured, the detection precision of the complex field is high, and large-range drawing can be realized through model generalization, so that the full-process automatic processing of lattice field transformation drawing is realized; according to the method, the problems of intensive labor and high time and economic cost of a traditional field investigation or manual interpretation mode are solved, so that the detection efficiency is greatly improved, the technical blank of current standardized grid field transformation in the drawing method is filled, and reliable technical support is provided for land protection, agricultural management and policy making.

Images