Automatic Recognition Algorithm For Steady Arms

Catenary detection is an important work to guarantee the electrified railway running reliably, the state of Catenary influence the safety of trains all the time. So in order to make sure the CRH run in order, improve the safety and reliability of power supply, a objective of high-speed rail power supply security detection monitoring system was proposed. The catenary suspended state detection monitoring device is one of subsystem. This device is used to detect catenary parameters and components state. Steady arms slope detection is one of the requirement of the system. So, how to find the Steady arms in pictures automatically becomes the basis of detection.To solve this problem, this paper proposes a steady arms automatic identification technology. This technology is based on Hough transform and AdaBoost algorithm. According to the eigenvalue of features, AdaBoost algorithm finds a threshold to separate the target object and others. Most components of catenary, such as stanchions, cantilever and positioners are all with linear characteristics, so we can extract the liner features of images and according to every liner feature defines an eigenvalue. And then, according to the liner features to train the AdaBoost algorithm, get some weak classifiers and a strong classifier. Finally, the strong classifier can be used to detect images and find where is the Steady arm in a image.But during the research, the size of pictures is very large and the size of Steady arms is smaller than the other components, so the Steady arm can’t be detected sometimes. To cure the above problems, we can find the stanchion in pictures, and then according to the location of the locating device on stanchions, cutting out a image region with locating device only. So the detection range is reduced obviously. After then we can identify Steady arms.The slope of the steady arm can be got, when the steady arm is detected. During a transformation we can get the slope in the world coordinate system.