¹School of Information Science and Engineering, Central South University, Changsha, China
²College of Information Science and Engineering, Hunan University, Changsha, China
³Dept of Computer Engineering and Information Technology, Kigali Institute of Science and Technology, Kigali, Rwanda

---

Abstract—Fire hazards occurring recently in the world lead to the need of designing accurate fire detection systems in order to save human lives. The newest innovations continue to use cameras and computer algorithms to analyze the visible effects of fire and its motion in their applications like the adaboost classifier which is well known for its strength in rigid objects detection from images. This paper presents a Fire Detection System (FDS) with an algorithm that works side by side with the adaboost classifier to determine the presence of fire in an image taken by a normal web camera (webcam), in order to decrease the false alarms in an indoor scene. The images are first preprocessed and their selected discrete cosine coefficients are kept for analysis to get better coefficients that will be fed to a neural network for classification and results are compared to a statistical approach used in combination with binary background mask (BBM) and a wavelet-based model of fire’s frequency signature(WMF) to test its accuracy.

Index Terms—Computer vision, fire detection, neural network, Discrete Cosine Transform, adaboost classifier

[PDF]

Cite: Jean Paul Dukuzumuremyi, Beiji Zou, Carine Pierrette Mukamakuza, Damien Hanyurwimfura, Emmanuel Masabo, "Discrete Cosine Coefficients as Images Features for Fire Detection based on Computer Vision," Journal of Computers vol. 9, no. 2, pp. 295-300, 2014.