Skip to Main Content
In this paper, an image-based flame pixel detection method is proposed. Based on detailed observation of the color distribution in RGB space, an algorithm to extract the candidate flame pixels is proposed. After removing the falsely extracted fire-like pixels, we restore the mistakenly removed flame faults and holes and reconstruct the flame region. The experimental results are compared with other methods. It shows a higher accuracy and robustness of fire detection by our proposed method. A high accuracy is necessary and important for the next step to detect fire dynamic features.