基于红外成像的电力施工视频监控图像目标识别算法  

Target recognition algorithm of electric power construction video monitoring image based on infrared imaging

作　　者：陈莉波 周小华 杨晒 蒯雷 万俊 梁勇 CHEN Libo;ZHOU Xiaohua;YANG Shai;KUAI Lei;WAN Jun;LIANG Yong(State Grid Chongqing Electric Power Company,Chongqing 400000,China;State Grid ChongqingQi爷nan Power Supply Branch,Chongqing 400000,China)

机构地区：[1]国网重庆市电力公司,重庆400000 [2]国网重庆市电力公司綦南供电分公司,重庆400000

出　　处：《激光杂志》2021年第9期129-133,共5页Laser Journal

摘　　要：为高精度、全面识别电力施工视频监控图像目标,研究一种基于红外成像的电力施工视频监控图像目标识别算法。采用基于局部线性映射方法的红外成像彩色化处理算法,通过色彩空间变换、色彩传递、伪彩色编码突出目标三步骤,实现监控图像彩色化,突出基于红外成像的电力施工视频监控图像目标;针对突出的基于红外成像的电力施工视频监控图像目标,采用基于支持向量机的目标识别算法,通过最佳超平面预测目标种类,实现目标识别。测试结果表明:不论在室外或是室内的电力施工视频监控图像中,所提算法均具备较好的电力施工视频监控图像目标识别效果。该算法在不同监控环境中,漏识率与误识率均低于3%,可高精度、全面识别电力施工视频监控图像目标。In order to identify the image target of electric power construction video monitoring with high precision and overall,a target recognition algorithm based on infrared imaging for electric power construction video monitoring image is studied. The color processing algorithm of infrared imaging based on local linear mapping method is adopted to realize the color of monitoring image and highlight the image target of electric power construction video monitoring based on infrared imaging through three steps of color space transformation,color transfer and pseudo color coding;for the outstanding electric power construction video monitoring image target based on infrared imaging,the support direction is adopted. The target recognition algorithm of measuring machine can predict the target type through the best hyperplane to realize the target recognition. The test results show that the proposed algorithm has good target recognition effect in both outdoor and indoor electric power construction video monitoring images. In different monitoring environments,the rate of missed recognition and false recognition is less than 3%,which can identify the video monitoring image target of electric power construction with high precision and comprehensive.

关 键 词：红外成像 电力施工 视频监控 图像 目标 识别 

分 类 号：TN219[电子电信—物理电子学]