Improved Forest Fire Smoke Image Segmentation Model Based on GVF Model

This article proposes an improved fire smoke image segmentation model (UGVFOF) based on the GVF model. The two components of the UGVFOF model and the image gradient are coupled through the use of Laplace operators on the manifold during the generation process. Due to the coupling term taking into account more image information during the diffusion process, UGVFOF snake outperforms traditional image based fire and smoke detection methods in some aspects, such as weak edge protection, target separation, and converging to deeper concave surfaces while retaining certain advantages such as expanding capture range and insensitivity to initialization. The extensive image experiments in this article demonstrate its advantages.