Feature Extraction Of A Burner's Jet Images Based On Visual Technology

The distribution of flow fields in burners is of vital importance for the safety and economic performance of PC-fired boilers. However, the only effective approach in approving the overall flow-field structure in the furnaces seems to be the adjustment of the flow field near the burner zone. Therefore, monitoring and studying the flow field in the burner zone is of great significance.In this thesis, a system for the visualization of cold flow-field in furnace is constructed, and visualization methods are applied to obtain image of the cold-state jet flow at the burner's outlet; a software for the processing of the jet images is developed and algorithms are designed and optimized so that digital image processing techniques are used to analyze the obtained image and at the same time extract geometric parameters such as the edges of the jet. The procedure of the algorithms steps as follow:①discrete-points noise denoising based on weighted mean filter algorithm;②image segmentation based on region growing algorithm;③acquiring edge pixels based on adaptive edge detection algorithm;④linear model based on generalized least square method;⑤locating the centerline position of jet and extracting other geometric parameters such as the angle between two linear edges and angle information of the jet geometric centerline, and these parameters can be used for the quantitative analysis of the distribution characteristics of the cold-state flow at the burner zone, also more accurately predict how reasonable the structural organization of the flow field inside the furnace is under heat-state conditions. Research results show the relationship between time-consuming and data precision is optimum when a 10×10 filter operator is chosen and a 5% tolerance of image dynamic is set between the seed value and the edge pixel value according to the growing rule. Time-consuming is about 2~4 seconds to complete the all image processing procedure in a personal computer equipping an AMD 3000+ CPU and a 512MB memory. Considering the shortcutting in data-processing and the stability and reliability of the algorithm, the approach has a potential meaning for real-time analysis of flow field of a burner's jet in furnaces.The introduction of visualization techniques make real-time monitoring of the flow possible, and digital image processing techniques provides sufficient data to support computerized analysis of the flow field. Compared with traditional research methods, the novel method of visualization combined with digital image processing integrates photoelectric information, computer vision, image processing techniques and other cutting-edge research results. It not only enables direct visualization of the flow field, paying attention to the qualitative geometric characteristics the qualitative geometric characteristics, but also allows quantitative description of the flow characteristics by using the geometric parameters obtained from the jet. Fast data processing, stable and reliable feature extraction algorithm of this method is bound to give it broad practical application prospects.This thesis studies the visualization of flow fields inside boilers in power stations. It provides the topic with a new research idea, and a new feasible research approach for the computer-aided analysis of flow field characteristics.