摘要： 传统的火灾检测方法是感烟、感温、感光探测器以及红外对射探测，对于装有监控设备的大空间建筑物，该文提出了一种基于图像视觉特征的火灾检测方法。通过对摄像头摄取的图像进行处理，提取其色彩特征和小波特征，然后利用神经网络通过有监督的学习建立图像特征向量与三维火灾向量的联系，达到利用图像进行火灾识别的目的。

关键词: 视觉特征, 图像识别, 直方图, 小波变换, 神经网络

Abstract: Traditional methods for fire detection are smoke detection, temperature detection, light detection & infrared detection. For large space building with monitor and control , an image visual feature based method for fire detection is brought forward. Processing the image which is taken by camera, people can get its color feature & wavelet feature, then, using neural network & supervise learning, people can establish relations between image feature vector and three dimension fire vector, achieve the purpose of using image to detect fire.

中图分类号: 

• TP391.4