Flame target detection method based on digital image and convolution features

Abstract

Because the generalization of a flame detection model based on image features is not strong, and the requirement of a deep neural network model for the number of training samples is high, the invention provides a flame target detection method based on digital images and convolution features, and the method comprises the steps: firstly making a data set comprising video dynamic features; replacingthe standard convolution of the VGG16 in the classic Faster R-CNN with the depth separable convolution, and reducing the number of convolution layers; cutting 256 image blocks from the original imageaccording to a candidate box generated by the RPN, and extracting LBP features of each image block; reducing the size of an output feature map of the ROI pooling layer and the number of neurons of a full connection layer through convolution, and further reducing network parameters; and finally, combining the extracted LBP features, the dynamic features in the data set and the pooled tiled featurevectors, and sending the combined feature vectors to a full connection layer for classification and regression. The flame target detection model constructed by the patent has relatively high detectionprecision, is convenient to improve for overcoming the defects of a test result, and is high in flexibility.

Description

technical field [0001] The invention relates to a flame detection method. Background technique [0002] The rapid detection of flame is of great significance to the early warning and timely treatment of fire. Fire monitoring video system is an important means of fire prevention, and flame is an important visual sign of fire. hotspot. Flame recognition methods based on image features are mainly divided into two categories: flame static features and dynamic features. The static features of flame mainly include edge and color space information, etc. Qiao Jianqiang et al. proposed a flame detection method based on edge features, which extracts edge feature information from images for flame recognition; Chen Tianyan et al. proposed a threshold segmentation method based on YCbCr color space, using flame Feature extraction flame regions in color space. Flame dynamic features include mixed Gaussian background model, frame difference, etc. For example, Li Qinghui et al. used an ad...

AI Technical Summary

Problems solved by technology

[0004] The training of the deep convolutional neural network has high requirements on the capacity of the data set. Although migration learning can sometimes achieve good results on small and medium data sets, the similarity between the flame and the objects in the current public data set is very small, and the migration The effect of learning is negligible

Images