Fire alarm based on image semantic description
Download as PDF
DOI: 10.25236/systca.18.088
Author(s)
Xuan Zhao, Dengyin Zhang, Huanhuan Bao
Corresponding Author
Dengyin Zhang
Abstract
The complexity of outdoor smoke scenes often leads to low accuracy of outdoor fire alarms. We propose a fire alarm method of automatic generation of image semantic description to improve the accuracy of early warning in outdoor scenes. By establishing a semantic context model based on Bayesian network, the convolutional neural network is used to extract feature values from the labelled samples and train them to realize object detection. Then we train the context semantic model of the target object in the segmentation region to realize the semantic description of the scene to distinguish whether the smoke is caused by fire. The experimental results on the image datasets of five types of smoke scenes show that the method we propose in this paper has higher accuracy and lower recall rate than the existing methods.
Keywords
CNN, Object detection, Semantic description, Fire alarm