Automated Measurement of Coke Size by Computer Image Processing
Download citation file:
- Ris (Zotero)
- Reference Manager
The stability, production and coke consumption rate of the blast furnace may be greatly influenced by the uniformity of coke size. Therefore an automated system to measure and estimate the coke size is needed. In this paper, a computer-based vision system was established to acquire coke image by using a CCD camera. A digital scheme that combines various digital image processing techniques was developed to detect the edge of individual coke properly from the coke image. Then a novel approach was proposed to find the minimum width (coke size) of the detected coke shape by using the area moment of inertia principle with rotational search. The system developed was tested on various size of real cokes which include large (>70mm), middle (35∼70mm) and small (<35mm) size coke. The results show that average error of the coke size detected from the coke image was less than 10%. However, the average error of coke size between the detected coke and the real three-dimensional coke could be higher to 20∼30%. Further modification of the measured data by using statistics of the average difference of coke sizes reduces the error to about 10%. The feasibility and usefulness of the system is demonstrated.