The real-time Blast Furnace (BF) condition is of great significance to guide optimizing the charging operation and improving the gas flow distribution. The iron and steel enterprises can reduce the energy consumption and carbon emission, and also ensure the safe, green and efficient production of BF. However, because of the high-temperature, dusty and dark environment inside the furnace, it is difficult to obtain various sensing data, including temperature, images and pressure et al. To overcome the limitations, the industrial endoscope provides clear and fully-covered images inside the furnace with the flame information, but requires high-quality post-processing. Therefore, this paper proposes a new prediction method of the furnace condition. Experiments demonstrate that the proposed method can obtain abundant flame morphology information, which are especially essential for the condition prediction.