基于无人机巡线图像的地面油气管道识别方法  

Research on Recognition Method of Ground Oil and Gas Pipeline Based on UAV Patrol Line Image

作　　者：雍歧卫[1] 喻言家 YONG Qi-wei a YU Yan-jia b(a. The Military Supply Departmen b. Graduate Management School, Logistical Engineering University, Chongqing 401331, Chin)

机构地区：[1]后勤工程学院军事供油系,重庆401331 [2]后勤工程学院研究生管理大队,重庆401331

出　　处：《兵器装备工程学报》2017年第4期100-104,共5页Journal of Ordnance Equipment Engineering

摘　　要：限于成本,无人机搭载的任务设备主要为普通数码相机,采集的是可见光图像,针对此种情况,提出了一种利用彩色分割及形状检测识别油气管道的方法,首先需要设定感兴趣区域ROI,计算出协方差矩阵C和均值m,并使用欧氏距离、马氏距离对图像进行彩色聚类分割,然后对分割图像填色后进行边缘检测,最后根据边缘图像进行霍夫变换来检测直线特征,实现复杂环境下对管道位置的自动定位。测试图像库包含300幅图像,识别准确率达到80.3%,实验结果表明,在色彩差异较大背景中,基于颜色和形状特征的识别方法能有效进行管线跟踪定位。Limited to the cost,the unmanned aerial vehicles are equipped with ordinary digital cameras for the main task,collecting the visible light image. In this paper,we proposed a method to identify oil and gas pipelines by color segmentation and shape detection. Firstly,we needed to set the ROI of the ROI,and calculated the covariance matrix C and mean m,and had image color cluster segmentation with Euclidean distance and Mahalanobis distance. Then,the edge detection was performed after that the image is segmented. Finally,the Hough transform is used to detect the line features according to the edge image to realize the automatic positioning of the pipe position in the complex environment. The test image bank contains 300 images,and the recognition accuracy is 80. 3%. Experimental results show that the method based on color and shape features can effectively track and locate the pipeline in the background of large color difference.

关 键 词：油气管道 无人机巡检 彩色分割 形状特征 

分 类 号：TE832[石油与天然气工程—油气储运工程]