摘要
为实现低照度室内火灾的早期预警,提出一种计算机视觉检测方法。通过摄像头获取图像,采用自适应中值滤波对图像进行去噪,采用直方图均衡化进行增强,并采用基于改进学习率的混合高斯模型提取烟雾区域,对所提取烟雾的纹理特征、运动方向特征以及面积特征进行融合,采用基于支持向量机的识别算法对烟雾与干扰物进行分类检测。实验结果表明,该算法可以有效区分烟雾与干扰物,有较高的识别率与鲁棒性。
In order to achieve the early fire warning of low illumination indoors environment,a computer vision detection method was proposed.Images of low illumination indoors were obtained from the camera.After image denoising by adaptive median filter,and image enhancement by histogram equalization,smoke area was extracted by method based on Gaussian mixture model which is improved on the vector.By integrating features of the texture characteristics,movement direction feature and the area feature of smoke,smoke and distractions were finally classified based on recognition algorithm of support vector machine.The experimental results showed that the smoke algorithm not only can effectively distinguish between smoke and distractions,but also has higher recognition rate and robustness.
出处
《消防科学与技术》
CAS
北大核心
2017年第4期493-496,共4页
Fire Science and Technology
基金
国家自然科学基金资助项目(51277149)
陕西省教育厅自然科学专项课题(14JKk1467)
关键词
低照度室内火灾
烟雾
支持向量机
火灾探测
low illumination indoor fire
smoke
support vector machine
fire detection
