Fire Detection System Based On Image

Fire is very harmful to human being’s life, especially accompanied with the highdeveloping speed of economy. It is getting more serious, so the researchers pay theirattentions to how to prevent and reduce fires. The fire detection based on digital imageis a new method of detection during the recently years. Support Vector Machine isstudied and used to complete the detection system.It is supposed that the video frame obtained by camera were separated and to formimages sequence firstly. Using gaussian mixture model for image segmentation of themoving area, and using the flame inner flame color information as a threshold value,eliminate invalid area, get the flame suspicious area. Then the circularity, the amount ofsharp angles, area proportion of red and green components, area variance rate, relatedcoefficient and flicker frequency in the suspected region were calculated and analyzed.At last, the model parameters of support vector machine were optimized by improvedartificial fish-swarm algorithm，choosing the optimal parameters by cross validation. Onthis foundation, the support vector machine classifier was built，using labeled datatraining Support Vector Machine.The extracted flame features were assembled as theinput vectors of the SVM classifier and the feature data can be classified and recognizedby SVM. Finally, designed and completed a fire detection alarm system based on PC.The experiment results show that the algorithm has high positive alarm rate, theseparating hyper-plane can be generated automatically by training SVM and it canseparate the sample data accurately. After verification, the system can complete the firealarm function.