Train track looseness non-contact real-time detection method based on camera shooting

The invention discloses a train track looseness non-contact real-time detection method based on camera shooting, and aims to solve the problem that current train track looseness mainly depends on manual inspection, and the efficiency is low. According to the method, a high-speed camera is used for acquiring a vibration video of a track when a train passes through or is artificially excited, a virtual feature point detection method based on an image pixel gray gradient and a feature clustering algorithm is provided by utilizing a characteristic that the train track and sleepers are arranged in a staggered manner, an optical flow algorithm is used for calculating optical flows of virtual feature points, and the time domain vibration of the track is measured. The natural vibration frequency of the track is obtained through FFT decomposition, and then whether the track is loosened or not is judged in real time according to the change of the natural vibration frequency. The method is simple in measuring device, hi