Low-Light Forest Flame Image Segmentation Based on Color Features

Yuanbin Wang,Jieying Ren
DOI: https://doi.org/10.1088/1742-6596/1069/1/012165
2018-08-01
Journal of Physics: Conference Series
Abstract:For some forest fire images with low light intensity, the traditional segment algorithm is not very good. To solve this problem, this paper proposes a low-light forest fire image segmentation method based on color features. In order to increase the brightness of the image, equalization is performed on the V channel in HSV color space. The distribution of flame pixels in RGB color space and YCbCr color space are analyzed. Based on this characteristic, the flame is segmented and the whole flame is extracted. Experiment results showed that the proposed algorithm in this paper had higher detection rate of flame pixels, better segmentation effect, and good anti-interference performance.
What problem does this paper attempt to address?