Fire detection method based on joint features and QLBC

The invention discloses a fire detection method based on joint features and QLBC, and the method comprises the steps: joint feature selection: a joint feature set is composed of transformation features and texture features, and specifically comprises a gray-level co-occurrence matrix-based feature GLCM, a discrete wavelet transform-based feature DWT and a discrete cosine transform-based feature DCT; using a forward feature selection method SFS to sequentially select feature vectors; according to fire image classification based on a quadratic Logit-Boost classifier QLBC, a group of sub-classifiers based on quadratic Logit-Boost is established, a likelihood function between an extraction feature and a test feature is calculated, all sub-classification results are combined into one classifier based on a likelihood result, logit loss is calculated by using a gradient descent function, and weight is adjusted according to a loss function. Compared with the prior art, the method can provide a fire classification res