基于机器视觉的高压断路器机械特性测量方法

Research on Mechanical Characteristic Measurement Method of High Voltage Circuit Breaker Based on Machine Vision

高压断路器是电力系统中关键的控制和保护设备,断路器机械特性的准确测量是其故障诊断和寿命预测的先决条件,对于保证系统的安全稳定运行具有重要意义。近年来,多种基于接触式传感器的机械特征监测方法被提出。但是,接触式传感器必须安装在断路器的本体结构上,存在诸多缺陷,例如传感器类型及安装位置对信号的影响、对机构正常工作的影响。针对传统接触式传感器的缺陷,该文提出一种基于机器视觉的机械特征提取方法。首先,使用高速图像采集系统采集机构的运动轨迹;然后,利用霍夫变换对视频图像中的关键角点进行自动定位;最后,基于Lucas-Kanade光流法跟踪关键角点的运动轨迹,得到高压断路器的机械特性。通过与行程传感器测量的数据对比,验证了上述方法的准确性。

High voltage circuit breaker(HVCB)is the key control and protection equipment in the power system.Accurate measurement of HVCB mechanical characteristics is a prerequisite for fault diagnosis and life prediction.Currently,several measurement methods using contact sensors have been proposed to extract the mechanical characteristics of HVCB.Contact sensors are mounted on the kinematic structure of the circuit breaker to extract the mechanical characteristics.However,the mounting imposes additional mass on the kinematic mechanism to interfere with its normal operation,and some mounting methods can cause damage to the HVCB.In response to the above problem,an improved method for tracking target trajectories based on machine vision with the optical flow is proposed in the presented paper.A high-speed camera is used to photograph the high-voltage circuit breaker operating mechanism,and then relevant information is extracted from the high-frame-rate video samples using relevant image processing algorithms.Firstly,the crank arm of the HVCB operating mechanism is photographed by using a high-speed camera(4kHz and 1080*1080).Because of the great advantage of tracking individual target points and can effectively reduce computational effort,Lucas-Kanade(LK)optical flow method is introduced to track and monitor the target points in the video.However,the following problems are prone to occur in the data extraction process:(1)Since optical flow tracking is based on image grayscale changes,the target points in the image with weak grayscale changes cannot be tracked accurately.(2)The analysis of multiple video samples requires excessive reliance on the manual positioning of target corner points at the same location,which is prone to errors and increases the workload of the experimenters.To solve the above problem,the Shi-Tomasi corner detection algorithm is applied to filter out the strong corner points in the image that can be easily tracked,and find an optimal corner point from them for later optical flow tracking.The selection of the optimal corner point requires that its motion trajectory be relatively long,because the longer the trajectory the more information it contains.Then the Hough transform algorithm with the adaptive thresholding technique in OpenCV is used to automatically locate the target corner points.In OpenCV,CV2.adaptiveThreshold()function is applied,the maximum value of the threshold is set to 255,and the threshold type is set to CV2.THRESH_BINARY.This method does not require manual participation in the process of mechanical feature extraction,which can effectively avoid the error brought by manually.Finally,the proposed method is compared with the displacement curves measured by acceleration and displacement sensors,and the results indicate that the proposed method is completely effective.The following conclusions can be drawn from the mechanical characteristics of HVCB extracted by the proposed method.(1)To solve the problem that some target points cannot be tracked,the Shi-Tomasi algorithm is applied to first filter the strong corner points that can be tracked.The results show that all the corner points derived from this method can be tracked accurately.(2)The use of the Hough transform algorithm can replace the manual positioning of target points,which can effectively reduce error and increase efficiency.