摘要
换流站阀厅设备在高压直流输电工程中起到重要的作用,目前已逐步实施由智能机器人代替人工操作对阀厅电力设备的工作状态进行日常巡检。机器人采集的巡检图像与设备的模板图像进行匹配,在巡检图像中定位设备区域是对设备状态识别的必要条件。本文根据阀厅设备在图像中呈现的特点,对SIFT特征匹配算法与RANSAC筛选算法进行了改进。首先,对图像中的光斑与阴影进行去除,增强图像细节信息,使SIFT算法避免了光照不一致的影响,可以在巡检图像与模板图像中提取出足够多的有效特征点。然后,采用靶心命中通过的方式代替RANSAC算法对特征点的筛选,缩短了计算时耗,提高了机器人的工作效率。经过机器人在阀厅现场的测试验证了本文算法的适用性与鲁棒性。
Converter valve is the important equipment in the process of direct current transmission. Regular inspection is necessary for ensuring the converter valve reliable working. At the present stage, a kind of developed rail type inspection robot is taking the equipment inspection work instead of people gradually. In this paper, an adaptive image matching algorithm is proposed for the valve hall inspection robot to improve the matching accuracy and decrease the compute time. In order to detect stable feature points from real-time image and template image by utilizing SIFT, at first, the brightness of these two images has to be equalized. Then, a target heart filter method is proposed to select effective feature points instead of RANSAC algorithm to increase the compute efficiency. Experimental results show good performance, it proves that the proposed method is both effective and accurate.
出处
《自动化技术与应用》
2016年第10期114-117,共4页
Techniques of Automation and Applications
