基于图像顺滑度的矿井外因火灾识别及抗干扰方法研究  

Research on identification and anti-interference method of mine external fire based on image smoothness

作　　者：孙继平[1] 李小伟 SUN Jiping;LI Xiaowei(School of Artificial Intelligence,China University of Mining and Technology(Beijing),Beijing 100083,China)

机构地区：[1]中国矿业大学(北京)人工智能学院,北京100083

出　　处：《中国矿业大学学报》2025年第1期215-226,共12页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(52364017);国家重点研发计划项目(2016YFC0801800)。

摘　　要：图像监测是矿井火灾火焰主要感知方法,但受矿井光源干扰.图像圆形度和矩形度方法受摄像机安装位置、拍摄目标图像角度影响大,难以排除矿井光源的干扰.本文揭示了圆形灯、正方形灯和长方形灯等矿井光源图像实际边界周长近似等于其等面积矩形图像边界周长,火焰图像实际边界周长明显大于其等面积矩形图像边界周长的特点.提出图像顺滑度计算方法,用形状与外接矩形图像相似、面积与目标图像面积相等的相似矩形周长,除以图像实际边界周长的比值,表示图像顺滑度.目标图像顺滑度数值越小,说明图像边界越不顺滑、凸凹越严重.提出基于图像顺滑度的矿井外因火灾识别及抗干扰方法,计算目标图像顺滑度,根据矿井光源图像顺滑度数值较大,而火焰图像顺滑度数值较小,区分火焰与矿井光源.本方法不受矿井光源形状、摄像机距检测目标距离和图像大小、摄像机安装位置及其拍摄检测目标的角度等影响,适应范围广,识别准确率高.试验研究表明,顺滑度识别火灾火焰图像准确率为98.1%,召回率为97.9%;矩形度识别准确率为81%,召回率为78.9%;圆形度识别准确率仅为33.3%,召回率为28.9%.Image monitoring is the main sensing method for mine fire,but it is interfered by the light source in the mine.Image roundness and rectangularity methods are greatly affected by the camera installation position and the angle of the target image,and it is difficult to exclude the interference of the light source in the mine.It is revealed that the perimeter of the actual boundary of the image of the mine light source,such as circular lamp,square lamp and rectangular lamp,is approximately equal to the perimeter of the boundary of the rectangular image of the same area,while the perimeter of the actual boundary of the flame image is obviously larger than that of the boundary of the rectangular image of the same area,and other characteristics.The proposed image smoothness calculation method,with the shape and external rectangular image similar to the area of the target image area equal to the perimeter of the similar rectangle,divided by the image of the actual boundary perimeter of the ratio of the smoothness of the image.The smaller the value of the target image smoothness,the less smooth the image boundary is,and the more serious the convexity and concavity are.An image smoothness-based identification and anti-interference method for mine external fires is proposed,which calculates the smoothness of the target image and distinguishes the flame from the mine light source based on the large smoothness value of the mine light source and the small smoothness value of the flame image.This method is not affected by the shape of the mine light source,the distance of the camera from the detection target and the image size,the installation position of the camera and the angle of the camera and the detection target,etc.It has a wide range of adaptability and high recognition accuracy.The experimental study shows that the accuracy of recognizing the fire flame image by smoothness is 98.1%,and the recall rate is 97.9%;the accuracy of recognizing rectangularity is 81%,and the recall rate is 78.9%;and the accuracy of recognizin

关 键 词：矿井火灾 图像顺滑度 图像边界 火灾监测 图像识别 

分 类 号：TD752.3[矿业工程—矿井通风与安全]