An prediction approach of sintering state in rotary kiln based on image analysis

A new prediction method based on image analysis is presented to overcome the measurement difficulty of some heavy equipment with instruments. The method is used in the prediction of sintering states in the rotary kiln. Firstly, the regions of interesting (ROIs) was segmented from sintering images based on an improved dual-fast marching method, which includes the regions of materials and enough burning, etc. Then the characteristics were extracted from ROIs, moreover the SVM pre-treatment classification model of the sintering state was constructed based on one-versus-another method to classify the sintering states into multi-kinds. Secondly, the distribution of wrong sample point in the pre-treatment classification was studied, the easily confused samples were classified as one kind and multi-lay SVM classification was designed. Finally, the dynamic model of sintering state with sintering images was established based on the multi-lay SVM method to predict the developing tendency of sintering states. The experiment results show that the new method is fast, accurate and has extensive application.