基于 HSV 颜色空间特征的视频烟雾检测-《计算机技术与发展》

文章信息/Info

作者:: 曹灿灿¹ ; 龚声蓉²; 周立凡² ; 钟珊²; 1. 东北石油大学计算机与信息技术学院,黑龙江大庆 163318;
2. 常熟理工学院计算机科学与工程学院,江苏常熟 215500

Author(s):: CAO Can-can1 ; GONG Sheng-rong2 ; ZHOU Li-fan2 ; ZHONG Shan2; 1. School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China;
2. School of Computer Science and Engineering,Changshu Institute of Technology,Changshu 215500,China

Keywords:: smoke detection; HSV color space features; convolutional neural networks; conv-12; smoke alert

摘要:: 由于烟雾具有形状不规则、扩散缓慢的特性,导致传统烟雾识别方法对烟雾检测存在一定的缺陷,如烟雾检测准确率低、烟雾警报响应时间长等问题。为了满足野外空旷场景下烟雾检测的准确性和实时性,提出了基于 HSV( Hue,Saturation,Brightness,色调,饱和,明亮) 颜色空间特征和卷积神经将网络相结合的视频烟雾检测的识别方法。通过将烟雾图像的 RGB 颜色空间特征映射到 HSV 颜色空间特征后提取烟雾候选区域,提取到的烟雾候选区域经过高斯混合模型进行运动判断,然后将符合运动特征的烟雾候选区域图像送入到训练好的卷积神经网络中进行烟雾识别。针对传统烟雾检测效率问题,设计了卷积神经网络 conv-12 用于烟雾识别。实验结果表明,基于 HSV 颜色空间特征和卷积神经网络 conv-12 相结合的视频烟雾识别方法对视频烟雾检测的准确率为 96. 45% ,烟雾检测率为 93. 3% ,烟雾报警平均响应时间为 0. 9s。相较于其他方法,在烟雾检测准确率、烟雾检测率、烟雾警报响应时间都有一定的提升。

Abstract:: Due to the irregular shape and slow diffusion of smoke,traditional smoke identification methods have some defects in smoke detection,such as low accuracy, long smoke alarm response time and so on. In order to satisfy the accuracy and real time of smoke detection in open field,we propose an identification method of video smoke detection based on combining HSV( Saturation,Brightness,Tone) color space features and convolution neural networks. The smoke candidate region is extracted by mapping the RGB color space features of the smoke images to the? HSV color space features,then the smoke candidate region image is fed into the trained convolution neural network for smoke recognition. For? the traditional smoke detection efficiency,a convolutional neural network conv-12 is designed for smoke recognition. Experimental results show that the accuracy of video smoke recognition method based on HSV color space features and convolution neural network conv-12 is 96. 45% ,? the smoke detection rate is 93. 3% ,average response time of smoke alarmis 0. 9 s. Compared to other methods,it has a certain improvement in smoke detection accuracy,smoke detection rate and smoke alarm response time.