A novel detection method for ore-coke ratio of blast furnace based on structure-texture entropy of images

The ore-coke ratio is crucial for regulating gas flow distribution, enhancing permeability, and ensuring smooth operation of the blast furnace. Direct detection of the ore-coke ratio, leveraging the structure and texture differences in ore and coke images, is a widely accepted method. The harsh environment of the blast furnace results in unclear burden surface images, making it challenging to differentiate between ore and coke. This complicates image-based direct detection for ore-coke ratio, with few studies addressing this issue. This paper firstly introduces a structure-texture entropy, grounded in information entropy, to effectively differentiate between ore and coke. Then, a structure-texture enhancement algorithm based on image extreme dark features is proposed to optimize the accuracy of the structure-texture entropy. Finally, based on the high-accuracy structure-texture entropy, a novel algorithm for burden surface area extraction is presented. This method uses feature vector clustering and functional optimization for real-time ore-coke ratio detection. Industrial experiments demonstrate that the proposed method achieves high-accuracy ore-coke ratio detection, meeting the requirements for long-term stability of the blast furnace.