Fig. 1. Flow chart of video smoke detection algorithm

Fig. 2. Motion object detection. (a) Original images; (b) process results based on GMM; (c) process results based on morphology

Fig. 3. CNN architecture of video smoke detection

Fig. 4. Data set. (a) Examples of positive samples; (b) examples of negative samples

Fig. 5. Comparison of smoke recognition effects. (a)(b)(c) Before setting probability threshold; (d)(e)(f) after setting probability threshold

Fig. 6. Video set. (a) Examples of smoke videos; (b) examples of non-smoke videos