Title :
Eigen-flame image-based robust recognition of burning states for sintering process control of rotary kiln
Author :
Li, Weitao ; Mao, Kezhi ; Zhou, Xiaojie ; Chai, Tianyou ; Zhang, Hong
Author_Institution :
Key Lab. of Integrated Autom. of Process Ind., Northeastern Univ., Shenyang, China
Abstract :
Recognition of burning states is a critical issue in sintering process control of rotary kiln, and the recognition is usually based on the temperatures of burning zone. Recently, flame image-based burning states recognition has received much attention. However, most of the methods reported involve image segmentation, which is a very challenging problem due to the poor quality of flame images of rotary kiln, and the recognition accuracy is hard to guarantee. In this study, we propose a new method for burning states recognition without demanding image segmentation. The basic idea of this new method is to extract features from eigen-flame images to classify these flame images using probabilistic neural network. The advantages of the new method are threefold. Firstly, the new method is computationally more efficient than image segmentation-based methods. Once the eigen-flame images are obtained off-line, feature extraction is simply a process of correlating an image with each of the eigen-flame images. Secondly, the eigen-flame image-based features capture the global characteristics of an image and hence can lead to more accurate recognition. Thirdly, the new method is robust and is able to deal with images with poor quality. The proposed new method is validated through extensive experimental studies, to show that the new method improves recognition accuracy substantially compared with image segmentation-based methods and temperature-based methods. It is expected that more consistent product quality, higher yield and lower energy consumption can be achieved if the new burning state recognizer is incorporated into our previously developed hybrid control system for rotary kiln.
Keywords :
feature extraction; flames; image classification; image recognition; kilns; neural nets; probability; process control; production engineering computing; sintering; burning state recognition; eigen-flame image-based robust recognition; feature extraction; probabilistic neural network; rotary kiln; sintering process control; Character recognition; Feature extraction; Fires; Image recognition; Image segmentation; Kilns; Neural networks; Process control; Robust control; Robustness;
Conference_Titel :
Decision and Control, 2009 held jointly with the 2009 28th Chinese Control Conference. CDC/CCC 2009. Proceedings of the 48th IEEE Conference on
Conference_Location :
Shanghai
Print_ISBN :
978-1-4244-3871-6
Electronic_ISBN :
0191-2216
DOI :
10.1109/CDC.2009.5400123