Image smoke and fire smoke feature extraction method based on neural network

The invention relates to the technical field of image recognition and analysis, in particular to an image smoke and fire smoke feature extraction method based on a neural network, and the method comprises the following steps: S1, obtaining and processing an on-site picture of a power transmission line; s2, marking smoke and fire smoke characteristics in the field picture of the power transmission line; s3, training a feature extraction neural network model by using a fast-RCNN (Regional Convolutional Neural Network); s4, acquiring smoke and fire smoke feature images in different states for training a smoke and fire smoke feature extraction model; s5, performing an HOG feature extraction method; and S6, obtaining a trained smoke and fire smoke feature extraction neural network model. According to the method, the smoke and fire smoke characteristics contained in the image are extracted in combination with the deep neural network, potential dangers can be eliminated earlier, and losses are avoided. 本发明涉及图像识别分析技术