An algorithm for detecting forest-fire was proposed

which was based on multiple features fusion of smoke and spatial accuracy compensation.Firstly the detection results derived separately from motion analysis and wavelet analysis were fused

and then the fusion results were cumulated using smoke color and motion cues to realize the smoke detection.The smoke profiles were extracted with spatial accuracy compensation based on the scene horizon.We have tested our algorithm on a number of image sequences

and the results show that our algorithm can provide significant improvements over accuracy of forest-fire detection

decrease false alarm rate and enhance the robustness.This approach has important application value in forest-fire surveillance system.