基于HOG特征和滑动框搜索的地面油气管道检测方法  

Ground Oil and Gas Pipeline Detection Method Based on HOG Characteristic and Sliding Box Search

作　　者：雍歧卫[1] 喻言家 陈雁[1] 

机构地区：[1]后勤工程学院军事供油系,重庆401331

出　　处：《重庆理工大学学报（自然科学）》2017年第11期192-197,共6页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金资助项目(51475469)

摘　　要：提出一种基于HOG特征和滑动框搜索的地面油气管道检测方法,能快速、高效、准确地检测高分辨率无人机巡线图像中的地面油气管道。该方法首先提取管道与非管道图像样本的HOG特征,由所得特征作为样本数据训练油气管道检测分类器。将训练好的分类器用在整张无人机巡线图像中自动检测油气管道,利用与样本管道图像尺寸一致的滑动框对整张无人机巡线图像进行扫描,提取滑动框中的HOG特征输入到已训练的分类器中判断该窗口是否为管道,并进行标记。为了验证提出方法的有效性,将此方法运用于235张高分辨率航拍图像上进行油气管道自动检测,检测精确率达到84.7%。A ground oil and gas pipeline detection method based on HOG features and sliding frame search is proposed,which can detect the ground oil pipeline in the high resolution unmanned aerial vehicle( UAV) patrol image rapidly,efficiently and accurately. The method firstly extracts the HOG features of the pipeline and non pipeline image samples,and uses the obtained features as the sample data to train the gas pipeline detection classifier. The trained classifier is used for automatic detection of UAV pipeline images. And using a sliding frame with a certain size to scan the whole patrol line image of the UAV,it extracts the HOG feature in the sliding box,and inputs it into the trainedclassifier to determine whether the window is a duct and mark it. In order to verify the effectiveness of the proposed method,this method is applied to automatic detection of oil and gas pipelines on 235 high resolution aerial images,and the detection accuracy is 84. 7%.

关 键 词：无人机 HOG特征 滑动框搜索 支持向量机 地面油气管道检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]