摘要
火灾是影响公众安全和社会发展的主要灾害之一,为了减小危害,需要准确的识别火灾的发生。本文利用图像处理技术,对火灾图像进行预处理:包括降噪、灰度化以及二值化,提取预处理后图像中的特征值:圆形度、形体变化率和角点数。然后使用图像样本对BP神经网络进行训练,最后将训练得到的BP神经网络用于火灾图像识别。我们设置火灾和蜡烛火焰两个实验场景,对图像样本进行分析,能够正确的区分火灾图像和蜡烛图像。实验表明:此方法对于火灾的准确识别具有重要意义。
Fire is one of the major disasters affecting public safety and social development. In order to reduce the harm, it is necessary to accurately identify the occurrence of a fire. In this paper, image processing technology is used to pre-process fire images: including noise reduction, grayscale and binarization. The eigenvalues in the preprocessed images are extracted: circularity, rate of change of body shape and number of corners. Then the image samples are used to train the BP neural network. Finally, the trained BP neural network is used for the fire image recognition. We set fire and candle flame two experimental scenes, the analysis of image samples, the right to distinguish between fire images and candle images. Experiments show that this method is of great significance for the accurate identification of fire.
出处
《数码设计》
2017年第8期22-23,共2页
Peak Data Science
