Skip to Main Content
Accurate recognition of burning state is critical in sintering process control of rotary kiln. Recently, flame image-based burning state recognition has received much attention. However, most of the existing methods demand accurate image segmentation, which is quite challenging due to poor image quality caused by smoke and dust inside the kiln. In this study, we develop a more reliable method for burning state recognition without image segmentation issue. From the experience of operators, more discriminable flame and material zones will facilitate the subsequent feature extraction and burning state recognition. Motivated by this knowledge, we propose to include texture analysis based on Gabor filter as a pre-processing to improve the recognition result further and then extract global features based on eigen-flame images decomposition, and finally recognize the burning state using pattern classifier. The advantages of our method are threefold. Firstly, our Gabor filter selection approach can generate a compact filter bank to distinguish ROIs much more and offers help to facilitate the sequel. Secondly, the eigen-flame image method provides global features of an image with large class separability and hence can lead to more accurate classification performance. Thirdly, the new method is more robust and reliable than image segmentation-based methods and temperature-based method. Experimental studies show the effectiveness of our method.