Abstract: Deciding the explosion nature of coal dust needs examining its combustion instantaneous flame length. The currently-used large tubular coal dust explosion identification system, due to using naked-eye observation as the main detecting means, has many problems, such as poor measurement accuracy, big errors of repeatability, excessive human disturbance, and so on. Therefore, a new method bringing the digital image processing technology into the identification system of coal dust explosion was presented. This method can pretreat and segment the images of the coal dust flame, and make the edge detection through the digital image processing technology. And thus, the characteristic parameters of the maximum flame length can be obtained, and the judgment basis for the coal dust explosion can be deduced out.