接触网定位点智能识别方法

Intelligent Recognition Method for Mast Position of Overhead Contact Line

将一定距离内接触网拉出值检测数据视为由线路公里标和拉出值组成的2值图像。首先,对拉出值曲线进行变换降噪处理,以降低接触线相邻点拉出值的差值对角点检测时轮廓支撑域选取的影响;然后,采用基于滑动矩形的角点检测方法,利用拉出值在大部分定位点处都具有以线路中心线的垂线为左右对称的特性,忽略拉出值曲线上被测点邻域内的夹角变化,仅考虑垂直方向上存在的角点,初步检测出候选定位点;最后,在提取所有候选定位点特征属性向量的基础上,基于SPRINT决策树算法,采用随机森林算法对其进行分类,智能识别正确定位点,并对此方法进行试验验证。结果表明:定位点智能识别方法可以识别正确的定位点,且在保证性能的前提下具有较高识别精度。

The inspection data of catenary stagger within a certain distance is regarded as a binary image composed of kilometer post and stagger.At first,the stagger curve is transformed for noise reduction to lower the influence of the stagger difference between the adjacent points of contact wire on the selection of contour support domain during corner detection.Then the candidate mast position is initially detected by corner detection method based on sliding rectangle,utilizing the characteristic of the stagger curve that the stagger is symmetrical to the perpendicular line of the central line at most mast positions,ignoring the variations of the angle in the neighborhood of the measured point on the stagger curve,and considering only the corner points in the vertical direction.Finally,on the basis of extracting the feature attribute vectors of all candidate mast positions,and based on SPRINT decision tree algorithm,the random forest algorithm is adopted for the classification and intelligent recognition of the correct mast positions.Accordingly,the method is tested and verified.Results show that the proposed intelligent recognition method for mast position can identify the correct mast positions,and it has higher recognition accuracy under the premise of guaranteeing performance.