Optical Flow Algorithm And Research On Its Application To Rotary Kiln Status Recognition

Rotary kiln is one important thermo-device which is wildly adopted in cement, paper and metallurgy industry etc. However the tranditional control scheme can not be applied because of the features of rotary kiln such as large time lag, nonlinearity and difficulty to measure the key parameters. In this article, image processing techniques are applied to extract image characteristics from the rotary kiln images which reflect material sintering status and buring zone combustion status to recognize the rotary kiln status by pattern recognition techniques in the future.First, the production process of alumina material sintering rotary kiln is introduced, the feasibility of introducing material sintering status and burning zone combustion status to rotary kiln control is analyzed, and the image characteristics to be extracted are proposed. Among these characteristics, the granule size is more difficult to measure because of the noise in the image and the material trait. In this article, one method which utilizes motion segmentation to calculate the granule size is proposed.Optical flow is applied for motion segmentation in this article. The principal assumption, algorithm categories, etc. are outlined. The tensor-based optical flow algorithm is applied which is one hot topic of research on optical flow these years. It has merits such as high accuracy and obtaining dense optical flow field. Motion segmentation which has much effect upon the algorithm accuracy is one important part of the tensor-based optical flow algorithm.To improve the motion segmentation accuracy, research on optical flow is conducted and two revised algorithms are put forward. One is the accurate optical (low algorithm along motion boundary. The other is an optical flow algorithm which uses mean shift for motion segmentation. The later algorithm obtains motion segmentation while resolving optical flow field and incorporates other sources of information such as location and gray level to improve the segmentation accuracy.Then, the research on the application of optical flow algorithm to practical rotary kiln image sequence is discussed. Because the severe noise and abrupt brightness change in the rotary kiln image, the image shouble be processed. The algorithm procedure is introduced, and the contents e.g. region extraction, brightness compensation and image preprocessing are formulated.Finally, the development of rotary kiln status recognition system is introduced. The diagram of system software and hardware is formulated; the key modules such as extraction of image characteristics, characteristic sequences filtering and rotary kiln status recognition are analyzed. And the experimental results of industrial images are given.