Presentation is loading. Please wait.

Presentation is loading. Please wait.

Low False Reject Rate and False Accept Rate Multi-step Fire Detection Method Zhu, Xingjie; Xu, Yong; Chen, Huaiyou; Liu, Yan; Wen, Jiajun; Zhu, Qi Optik.

Similar presentations


Presentation on theme: "Low False Reject Rate and False Accept Rate Multi-step Fire Detection Method Zhu, Xingjie; Xu, Yong; Chen, Huaiyou; Liu, Yan; Wen, Jiajun; Zhu, Qi Optik."— Presentation transcript:

1 Low False Reject Rate and False Accept Rate Multi-step Fire Detection Method Zhu, Xingjie; Xu, Yong; Chen, Huaiyou; Liu, Yan; Wen, Jiajun; Zhu, Qi Optik - International Journal for Light and Electron Optics, vol. 124, pp. 6636-6641, 12// 2013. Reporter : Yi-Zheng, Lin 1

2 Outline Introduction Fire Detection Method ◦ Motion Region Detection ◦ Color-based Fire Detection ◦ Fire Region Refinement Experiments Conclusion Personal Remark 2

3 Introduction Fire detection has become an important issue because it is closely related to the safety of people and property. There have been a number of visual fire detection algorithms in the literature. However, most previous fire-detection techniques can only be used in particular conditions and situations and might be not suitable for indoor small-scale fire detection. 3

4 Introduction The fire has the characteristic that fire regions are moving when the flame is burning. The motion history image(MHI) has been used to detect moving objects. However, it cannot detect the moving objects very accurately. A number of proposed to use Gaussian Mixture Model(GMM) to obtain the mov- ing objects. 4

5 Introduction The color of fire is also a very important feature for fire detection and has been widely studied. Different researchers have different ideas about how to use the color to get the fire regions. We note that there are be some false fire detections when we get the fire regions using each of the color methods. 5

6 Introduction In this paper, we propose a three-step fire detection method to intelligently monitor the small-scale fire. The experimental results on 77 fire vi- deos and 249 no-fire videos show that our method obtains a false reject rate (FRR) and false accept rate(FAR) of 10.4% and 6.02% respectively. 6

7 Fire Detection Method 7

8 Motion Region Detection As the fire always appears as a motion region, we can first exploit the foreground detections method to extract the motion region from the video. In the case where the camera is fixed, GMM is suitable to extract the back- ground. 8

9 Fire Detection Method Motion Region Detection 9

10 Fire Detection Method Motion Region Detection 10

11 Fire Detection Method Color-based Fire Detection 11

12 Fire Detection Method Color-based Fire Detection 12

13 Fire Detection Method Color-based Fire Detection Video frame RGB-based HSI-based RGB+HSI 13

14 Fire Detection Method Fire Region Refinement 14 D is smaller than the threshold Dt, the fire region will be regarded as true fire region.

15 Fire Detection Method Fire Region Refinement 15 d = frame difference

16 Fire Detection Method Fire Region Refinement 16 d = frame difference

17 Experiments TotalVideo lengthimage resolution Fire video77One min/per each352 x 288 Non-fire video249Ten min/per each355 x 288 17

18 Experiments 18 The method in [17] exploits the YCbCr color. The method in [6] were also tested for com-parison.

19 Experiments 19 The method in [17] exploits the YCbCr color. The method in [6] were also tested for com-parison.

20 Experiments 20

21 Conclusion Small-scale fires. Low FAR. Identifying the majority fire detection. Distinguishing the objects that have similar characteristic as fires from the fire. 21

22 Personal Remark Execution time Hardware Actual using 22


Download ppt "Low False Reject Rate and False Accept Rate Multi-step Fire Detection Method Zhu, Xingjie; Xu, Yong; Chen, Huaiyou; Liu, Yan; Wen, Jiajun; Zhu, Qi Optik."

Similar presentations


Ads by Google