A flame image sequence classification method and device using convolutional neural network

Abstract

The invention discloses a flame image sequence classification method and device using a convolutional neural network. The method includes: obtaining a foreground image sequence, obtaining a comprehensive optical flow diagram corresponding to the foreground image sequence; obtaining a first training set and a first test set ; Obtain whether the comprehensive optical flow map corresponds to the category label of the image sequence of the real flame; train the preset first convolutional neural network model; test the trained first convolutional neural network model, obtain the first test result; judge the first Whether the test result is greater than the first preset threshold; if so, use the trained first convolutional neural network model as the target first convolutional neural network model; if not, adjust the training parameters of the trained first convolutional neural network model , and return to execute the training preset first convolutional neural network model; use the target first convolutional neural network to classify the integrated optical flow graph to be classified. By applying the embodiments of the present invention, the false detection rate of flame region detection can be reduced.

Description

technical field [0001] The present invention relates to the technical field of flame detection, and more specifically relates to a flame image sequence classification method and device using a convolutional neural network. Background technique [0002] The use of fire can be seen everywhere in daily life, but accidental fire or improper use are likely to cause fire, which will cause huge damage to the natural environment or human life and property. Timely and effective detection of flames can reduce or even avoid these injuries to a large extent. [0003] At present, the commonly used fire detection methods include fire detection methods based on temperature sensors or smoke sensors. However, this method has low timeliness and limited use scenarios. In order to improve the timeliness of detection, flame detection methods based on image processing have been widely used. The principle of the flame detection method based on image processing is: first artificially design featu...

AI Technical Summary

Problems solved by technology

[0004] However, in practical applications, if there are a large number of moving objects in the environment monitored by the image monitoring equipment, the images obtained by the image monitoring equipment will contain a large number of moving pixels, and too many moving pixels may affect the quality of the image monitoring equipment. Interference in the feature comparison process will lower the accuracy of flame area recognition, leading to false detections

Images