Hostile environment inside blast furnace (BF) and nonuniform fluidization characteristics of burden surface bring challenges to the extraction of burden line, which is an important factor affecting the smelting efficiency. In this study, based on the imaging principle of Synthetic Aperture Radar (SAR), a mechanical swing radar was designed to capture the high-density radar echo signals. By analyzing the characteristics of radar spectrum, the entropy weight method was used to complete the coordinate transformation from the nonuniform coordinate point cloud map to the grayscale image in real space to visualize the smelting states. Then, the burden surface features were enhanced by gamma correction, and the adaptive threshold segmentation was used to extract the surface transitional belt in the image. Finally, the burden line points were extracted by the energy centrobaric correction method to fit the burden line. Compared with traditional algorithm, experiments on industrial data indicate its feasibility and effectiveness.