摘要
在进行煤矿火灾烟雾识别时,图像分割阈值设置不当导致早期烟雾欠敏感,火灾烟雾识别准确率不高。为此,提出一种基于深度学习的煤矿火灾烟雾智能识别方法。首先,在构建粗糙度直方图中选择最优的波谷值作为图像分割阈值,对煤矿火灾图像展开粗分割。然后,利用帧差法提取图像中的运动区域,在RGB颜色空间内,构建基于煤矿火灾烟雾的颜色模型,用于删除运动区域中的非烟雾干扰。在烟雾区域内筛选最佳种子点,同时基于粗分割结果,运用区域生长算法展开精细化分割,获取最终的煤矿火灾烟雾分割结果。最后,引入深度学习中的双层级联卷积网络展开煤矿火灾烟雾智能识别。仿真分析表明,所提方法对煤矿火灾烟雾智能识别准确率较高,对于保证人员生命安全具有重要意义。
In the process of coal mine fire smoke recognition,improper setting of the image segmentation threshold results in less sensitive early smoke and lower accuracy of fire smoke recognition.Therefore,an intelligent recognition method of coal mine fire smoke based on deep learning is proposed.First of all,the optimal wave trough value is selected as the image segmentation threshold in the construction of the roughness histogram,and the coarse segmentation of the coal mine fire image is carried out.Then,the moving area in the image is extracted by the frame difference method.In the RGB color space,a color model based on coal mine fire smoke is constructed to remove non-smoke interference in the moving area.The best seed points are selected in the smoke region,and based on the coarse segmentation results,the refined segmentation is carried out using the region growth algorithm to obtain the final segmentation results of coal mine fire smoke.Finally,the two-layer cascade convolutional network in deep learning is introduced to carry out intelligent recognition of coal mine fire smoke.Simulation analysis shows that the proposed method has a high accuracy rate for intelligent recognition of coal mine fire smoke,and is of great significance for ensuring the safety of personnel.
作者
韩振国
刘清臣
马金盼
邵旭东
HAN Zhen-guo;LIU Qing-chen;MA Jin-pan;SHAO Xu-dong(Jalainur Coal Industry Co.,Ltd.,Hulunbuir Inner Mongolia 021410,China;China Shenyang Research Institute Co.,Ltd.,Shenfu Liaoning 113122,China)
出处
《计算机仿真》
2025年第10期405-409,433,共6页
Computer Simulation
基金
矿井火灾智能抑控系统研究(HNKJ21-H56-12)。
关键词
深度学习
煤矿火灾烟雾
智能识别
Deep learning
Coal mine fire smoke
Intelligent recognition
