基于HSV空间融合Retinex算法的全天候运动目标检测  

All⁃Weather Moving Target Detection by Retinex Algorithm Based on HSV Space Conversion

作　　者：陈卡 黄俊杰 包嘉琪 秦亚光[3] CHEN Ka;HUANG Junjie;BAO Jiaqi;QIN Yaguang(Guangxi Zhongjin Lingnan Mining Co Ltd,Laibin 546100,Guangxi,China;School of Computer Science and Artificial Intelligence,Wuhan Textile University,Wuhan 430200,Hubei,China;School of Resources and Safety Engineering,Central South University,Changsha 410083,Hunan,China)

机构地区：[1]广西中金岭南矿业有限责任公司,广西来宾546100 [2]武汉纺织大学计算机与人工智能学院,湖北武汉430200 [3]中南大学资源与安全工程学院,湖南长沙410083

出　　处：《矿冶工程》2023年第5期17-21,25,共6页Mining and Metallurgical Engineering

基　　金：国家自然科学基金(52274182)。

摘　　要：提出了一种在矿山复杂情况下基于HSV色彩空间融合Retinex算法的全天候运动目标检测方法。该方法将运动目标检测转换为运动目标上特定标志物的检测,通过颜色空间转换提取目标原始灰度图像,由灰度分布直方图判断是否进行低照度图像恢复,并在场景中选取特定的标志物,将原图像进行HSV颜色空间变换,选择对应ROI区域,通过阈值获取二值化图像,进行平滑处理后使用Canny算子提取轮廓信息,从而提取标志物的特征参数。应用实例证实,通过该方法能精准获得标志物特征参数,进而解决复杂环境下机械设备间连接件脱落问题,具有一定的实际工程应用价值。An all⁃weather moving target detection method by using Retinex algorithm based on HSV color space conversion was proposed for complex situation in a mine,which changed the detection of moving target to the detection of special marker on the moving target,and extracted the original grey image of the target by color space conversion.The histogram of grey distribution was used to judge whether to restore the low⁃illumination image.And special markers were selected in the scene,and the original image was converted to HSV color space.Then,the corresponding ROI region was selected,and the binary image was obtained based on the threshold value.After smoothing of the data,the Canny operator was used to extract contour information,thus characteristic parameters of the marker was extracted.The practice proves that the characteristic parameters of the markers can be accurately obtained by this method,which can thus solve the problem of connector falling off from mechanical equipment under complex situation.It is shown that this method is of certain practical significance in engineering application.

关 键 词：运动目标检测 图像检测算法 HSV空间 标志物检测 视频图像 图像增强 边缘检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]