A Fire Image Recognition Method Based on Deep Learning

An image recognition and deep learning technology, applied in the field of fire image recognition based on deep learning, can solve the problems that the accuracy is difficult to meet the requirements, the detection effect is not ideal, and the features are not obvious enough, so as to improve the accuracy, be easy to extract, and speed up the simulation. effect of speed

CN109522819BActive Publication Date: 2020-08-18XI AN JIAOTONG UNIV
16 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Publication Date
2020-08-18

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention belongs to the technical field of image information processing, and discloses a fire image recognition method based on deep learning, including: collecting smoke pictures and normal pictures in the early stage of fire as the training set and test set of convolutional neural network; The dark channel images constitute the final training set and test set; construct a convolutional neural model that can detect smoke; train the neural network to obtain a smoke detection model, and test and evaluate the performance of the model. Compared with the prior art, the present invention improves the correct rate of smoke detection in a single image by using dark channel images and deep learning methods, and at the same time increases the detection speed, and can be actually applied to fire detection in cities or forests.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the technical field of image information processing, and in particular relates to a fire image recognition method based on deep learning. Background technique

[0002] Fire detection has always been an important field of image information processing technology. How to apply image information processing technology to effectively control fire and prevent fire spread has attracted the attention of many researchers and has become one of the research hotspots in the field of computer vision.

[0003] Generally speaking, the evolution of a fire can be divided into four stages: the invisible stage, the visible smoke stage, the open flame stage and the spreading stage. In order to minimize the loss caused by fire, fire early warning work is usually concentrated in the first two stages. Traditional fire detection mainly uses sensors such as temperature sensors, gas sensors, and humidity sensors to analyze parameters such as ambient tem...

Examples

Embodiment

[0096] The fire image recognition method based on deep learning of the present embodiment, comprises the steps:

[0097] Step 1: Construct neural network sample training set and test set:

[0098] In this example, the pictures containing the smoke in the early stage of the fire and the normal pictures without the fire are collected for training the convolutional neural network. Specifically, the pictures containing the smoke in the early stage of the fire were collected through fire video extraction and small-scale open flame experiment shooting, and the normal pictures without fire were collected through shooting in daily life. The training set has a total of 10,712 photos, including 2,201 smoke pictures, which are labeled as 1, and 8,501 non-smoking pictures, which are labeled as 0;