Research And Implement Of Flame Detection Technology Based On Bag Of Features

Flame detection based on computer vision is a new type of flame detection methodwhich based on image processing and pattern recognition technology, it will not besubject to the complex factors of the space size or the bad environment. Compared withconventional flame detection technology, its application scenarios more wider and itsaccuracy more higher. Based on image processing technique and statistical patternrecognition method, in this paper, a flame detection method based on Bag ofFeatures(BOF) is proposed.Firstly, this paper studied the color space model, the visual characteristics of flameand pattern recognition classification method, for image feature extraction andBOF algorithm model provided the theoretical foundation. In the second place, studiedBOF algorithm model and its build process. In the end, after experimental verification,trained flame discrimination model based on BOF model and decision-making methodwas given, and determined whether there is a flame area according tothe decision-making method. The specific implementation process is divided into threeparts: first of all, collected a large number of fire image samples and extracted the colorspace features, clustered these color space features to generate a visual codebook; Inaddition, collected a large number of fire images and sectioned them to build flamesamples, extracted the color space features of flame samples as the feature set of trainingsamples, so that flame feature model is generated; Finally, built test image samples anddetected theirs flame areas, then the experimental results are given and analyzed.Experimental results show that, flame detection method based on Bag of Features canaccurately identify flame areas, it has good reliability, accuracy, and environmentaladaptability.