矿山应急救援红外图像SVM分割算法  

SVM segmentation algorithm for infrared image of mine emergency rescue

作　　者：杨伟 陈益能 童鑫 YANG Wei;CHEN Yineng;TONG Xin(Yibin Rescue and Fire Brigade,Sichuan Chuanmei Huarong Energy Co. ,Ltd. ,Yibin 644500,China;School of Safety,Xi’an University of Science and Technology,Xi’an 710054,China)

机构地区：[1]四川川煤华荣能源有限责任公司宜宾救护消防大队,四川宜宾644500 [2]西安科技大学安全学院,陕西西安710054

出　　处：《陕西煤炭》2022年第1期82-87,91,共7页Shaanxi Coal

摘　　要：受红外热成像固有特性限制及灾后巷道复杂环境的影响,有效的红外图像分割方法是完成应急救援的必要保证。为实现矿山钻孔救援生命信息红外图像人体目标识别,确定被困矿工位置,需对红外图像进行精确分割,为此,提出了基于SVM理论的红外图像分割方法。将XKQY-Ⅰ型红外生命信息探测仪侦测的视频分为正样本、负样本及测试样本,利用高低帽变换相结合的方法增强红外图像,采用交叉验证法选择最佳参数C与G,结合RBF核函数对数据集进行训练,获取SVM分类器,实现了测试样本的红外图像分割。并将图像分割效果与边缘检测算法、Otsu阈值分割法、K-mean聚类法、形态学分水岭法等传统图像分割方法进行了对比分析。结果表明,基于SVM的红外图像分割方法不需先验知识和优选阈值等预处理程序;运算时间为0.190 s,是QGA算法的24.33%;错分率为0.06,是QGA算法的55.05%;抗噪能力强。该算法可有效应用于矿山钻孔救援红外图像的分割,为后续人体目标识别提供有力支撑。Due to the limitation of the inherent characteristics of infrared thermal imaging and the complex environment of roadways after disaster,effective infrared image segmentation is a necessary guarantee for emergency rescue.In order to realize the human target recognition of mine drilling rescue life information infrared image and determine the location of trapped miners,it is necessary to accurately segment the infrared image.Therefore,an infrared image segmentation method based on SVM theory is proposed.The video detected by XKQY-Ⅰinfrared life information detector is divided into positive samples,negative samples and test samples.The infrared image is enhanced by the combination of high and low hat transformation.The best parameters C and G are selected by cross validation method,and the data set is trained based on RBF kernel function to obtain SVM classifier,so as to realize the infrared image segmentation of test samples.The image segmentation effect is compared with traditional image segmentation methods such as edge detection algorithm,Otsu threshold segmentation method,K-mean clustering method and morphological watershed method.The results show that the infrared image segmentation method based on SVM does not need preprocessing programs such as prior knowledge and optimal threshold;the operation time is 0.190 s,which is 24.33%of QGA algorithm;the misclassification rate is 0.06,which is 55.05%of QGA algorithm;it has strong anti noise ability.The algorithm can be effectively used for infrared image segmentation of mine drilling rescue,and provide strong support for subsequent human target recognition.

关 键 词：钻孔救援 生命信息 红外图像 支持向量机 图像分割 

分 类 号：TD77[矿业工程—矿井通风与安全] TN219[电子电信—物理电子学]